U.S. patent application number 13/776105 was filed with the patent office on 2013-11-07 for digital light processing hyperspectral imaging apparatus.
The applicant listed for this patent is Board of Regents, The University of Texas System. Invention is credited to Jeffrey A. CADEDDU, Robert P. FRANCIS, Edward LIVINGSTON, Rafael UFRET-VINCENTY, Karel J. ZUZAK.
Application Number | 20130296710 13/776105 |
Document ID | / |
Family ID | 41669597 |
Filed Date | 2013-11-07 |
United States Patent
Application |
20130296710 |
Kind Code |
A1 |
ZUZAK; Karel J. ; et
al. |
November 7, 2013 |
DIGITAL LIGHT PROCESSING HYPERSPECTRAL IMAGING APPARATUS
Abstract
A hyperspectral imaging system having an optical path. The
system including an illumination source adapted to output a light
beam, the light beam illuminating a target, a dispersing element
arranged in the optical path and adapted to separate the light beam
into a plurality of wavelengths, a digital micromirror array
adapted to tune the plurality of wavelengths into a spectrum, an
optical device having a detector and adapted to collect the
spectrum reflected from the target and arranged in the optical path
and a processor operatively connected to and adapted to control at
least one of: the illumination source; the dispersing element; the
digital micromirror array; the optical device; and, the detector,
the processor further adapted to output a hyperspectral image of
the target. The dispersing element is arranged between the
illumination source and the digital micromirror array, the digital
micromirror array is arranged to transmit the spectrum to the
target and the optical device is arranged in the optical path after
the target.
Inventors: |
ZUZAK; Karel J.; (Arlington,
TX) ; CADEDDU; Jeffrey A.; (Dallas, TX) ;
UFRET-VINCENTY; Rafael; (Dallas, TX) ; FRANCIS;
Robert P.; (Grand Prarie, TX) ; LIVINGSTON;
Edward; (Dallas, TX) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Board of Regents, The University of Texas System |
Austin |
TX |
US |
|
|
Family ID: |
41669597 |
Appl. No.: |
13/776105 |
Filed: |
February 25, 2013 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12538616 |
Aug 10, 2009 |
8406859 |
|
|
13776105 |
|
|
|
|
61087714 |
Aug 10, 2008 |
|
|
|
61087715 |
Aug 10, 2008 |
|
|
|
61168347 |
Apr 10, 2009 |
|
|
|
Current U.S.
Class: |
600/476 |
Current CPC
Class: |
A61B 5/413 20130101;
G06T 2207/10036 20130101; A61B 3/12 20130101; A61B 5/4842 20130101;
A61B 5/4064 20130101; G01J 3/0208 20130101; G01N 21/31 20130101;
A61B 5/0084 20130101; G01J 3/2803 20130101; A61B 5/0075 20130101;
A61B 5/14551 20130101; G01J 3/0229 20130101; G01J 3/10 20130101;
A61B 5/411 20130101; G01J 2003/1213 20130101; G01N 21/39 20130101;
G01J 3/0264 20130101; A61B 5/0059 20130101; G01J 3/2823 20130101;
G01J 2003/1282 20130101; A61B 5/0071 20130101; A61B 5/0086
20130101; G01J 3/02 20130101; G01J 3/021 20130101 |
Class at
Publication: |
600/476 |
International
Class: |
A61B 5/00 20060101
A61B005/00 |
Goverment Interests
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
[0002] This invention was made with government support under Grant
Number 5 U24 DK076169-02 awarded by the National Institutes of
Health. The government has certain rights in the invention.
Claims
1-45. (canceled)
46: A method of monitoring a tissue or an organ viability during
periods of ischemia comprising the steps of: forming a
hyperspectral image of said tissue or said organ, wherein said
hyperspectral image comprises a quantitative measure of oxygenation
of said tissue or said organ; calculating a length of time of said
tissue or said organ viability based on said quantitative measure
of oxygenation of said tissue or said organ.
47: The method according to claim 46, wherein said step of forming
a hyperspectral image is performed with an apparatus comprising: an
illumination source adapted to output a light beam, said light beam
illuminating said tissue or said organ; a dispersing element
arranged in said optical path and adapted to separate said light
beam into a plurality of wavelengths; a spatial light modulator
adapted to tune said plurality of wavelengths into a spectrum; an
optical device comprising a detector and adapted to collect said
spectrum reflected from said tissue or said organ and arranged in
said optical path; and, a processor operatively connected to and
adapted to control at least one of: said illumination source; said
dispersing element; said spatial light modulator; said optical
device; and, said detector, said processor further adapted to
output said hyperspectral image of said tissue or said organ,
wherein said dispersing element is arranged between said
illumination source and said spatial light modulator, said spatial
light modulator is arranged to transmit said spectrum to said
tissue or said organ and said optical device is arranged in said
optical path after said tissue or said organ.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is a Divisional of application Ser. No.
12/538,616, filed on Aug. 10, 2009, which application claims
priority to U.S. Provisional Patent Application No. 61/087,714,
filed Aug. 10, 2008, U.S. Provisional Patent Application No.
61/087,715, filed Aug. 10, 2008, and U.S. Provisional Patent
Application No. 61/168,347, filed Apr. 10, 2009, all of which
applications are incorporated herein by reference.
FIELD OF THE INVENTION
[0003] The present invention relates generally to the field of
spectroscopic imaging that results in collecting/measuring
spectroscopic image data and the associated algorithms that result
in chemically encoded images of tissue, including devices, methods
and systems for fluorescence imaging of tissues in real-time.
Additionally, the present invention relates generally to the field
of hyperspectral imaging of medical tissues. In particular, the
invention is directed to new devices, tools and processes for the
in vivo detection and evaluation of diseases and disorders.
Moreover, the present invention describes, but is not limited to a
spectrally conformable illumination system and associated
analytical algorithms, e.g., chemometrics, spectral and image
digital signal processing, used to reduce the number of image
frames required to generate a real-time processed image that is
color-coded based on matching the reflectance of each pixel in an
image to known reflectance spectra. Generally, the present
invention is applicable to surgical and clinical applications, and
is also applicable to a variety of other applications, including
but not limited to microscopy, pathology, food safety/inspection
and pharmacology.
BACKGROUND OF THE INVENTION
[0004] Without limiting the scope of the invention, its background
is described in connection with laparoscopic procedures, such as
cholecystectomy, in connection with a hyperspectral imaging system
used for medical tissues and in connection with new devices, tools
and processes for the in vivo detection and evaluation of diseases
and disorders.
[0005] The gallbladder is a small pear-shaped organ that stores and
concentrates bile. The gallbladder is connected to the liver by the
hepatic duct. It is approximately 3 to 4 inches (7.6 to 10.2 cm)
long and about 1 inch (2.5 cm) wide. The function of the
gallbladder is to store bile and concentrate. The bile emulsifies
fats and neutralizes acids in partly digested food. A muscular
valve in the common bile duct opens, and the bile flows from the
gallbladder into the cystic duct, along the common bile duct, and
into the duodenum (part of the small intestine).
[0006] The following are disorders of the gall bladder.
Cholelithiasis is a disorder of the extrahepatic biliary tract
related to gallstones resulting in the presence of stones in the
gall bladder. Gall stones are bodies formed within the body by
accretion or concretion of normal or abnormal bile components. Gall
stones of various shapes and sizes are formed within the gall
bladder. Cholecsytitis is the inflammation of the gall bladder.
Cholecystitis is often caused by Cholelithiasis, with choleliths
most commonly blocking the cystic duct directly which causes the
gallbladder's wall to become inflamed. Extreme cases may result in
necrosis and rupture, and could cause further infection and pain
resulting from such inflammation. Inflammation often spreads to its
outer covering, irritating surrounding structures such as the
diaphragm and bowel. Cholecystitis usually presents as a pain in
the right upper quadrant. Gall bladder Cancer is a relatively
uncommon cancer that occurs in the gall bladder. If detected early,
this cancer can be cured by removing the gall bladder.
[0007] Gall bladder diseases are marked with some or all of the
following symptoms: severe and constant pain in the upper right
abdomen which can last for days; increasing pain when drawing a
breath; and, radiating pain to the back or occurring under the
shoulder blades. About a third of patients have fevers and chills.
Nausea and vomiting may also occur. Complaints of gas, nausea, and
abdominal discomfort after meals are the most common, but they may
be vague and indistinguishable from similar complaints in people
without gallbladder disease. Moreover, gall bladder diseases may
result in jaundice (yellowish skin), dark urine, lighter stools, or
combinations thereof.
[0008] As the majority of patients incur gallstones along with a
gall bladder disease, the diagnosis can usually be confirmed
thorough ultrasound imaging, a safe, painless and non-invasive
technique that uses high frequency sound waves to create an image
of gallbladder and gallstones. Other diagnosis methods like X-Ray
and other scanning technology may be used.
[0009] Cholecystectomy is one of the most common operations
performed in United States. It is frequently used for the treatment
of symptomatic gallstones. Cholecystectomy is the surgical removal
of gall bladder. The two procedures utilized to surgically remove
the gall bladder are Open Cholecystectomy and Laparoscopic
Cholecsytectomy. The laparoscopic method is utilized more
frequently. The choice of the procedure is made on individual
basis. A Cholecystectomy is performed to treat Cholelithiasis and
Cholecystitis.
[0010] In conventional Cholecsytectomy, a surgeon makes an incision
approximately 6 inches long. The incision is made either
longitudinally in the upper portion of the abdomen or obliquely
beneath the ribs on the right side. During the procedure, drains
may be inserted into the abdomen, which are usually removed while
the patient is still in the hospital. Following a normal
cholecystectomy, the patient may be in the hospital from one to
three days post surgery. Normal activity can be resumed in about
four weeks. In complicated cases normal activity can be resumed in
about four to eight weeks. The procedure is very common and is
successful most of the time. However, any surgery involves risk
factor associated with it, which can cause complications. The most
common associated risk with Cholecsytectomy is the injury to the
common bile duct (CBD) which is hidden, as it lies below a layer of
fat. Thus, it is the surgeon's expertise and judgment for locating
the CBD and avoiding any injuries.
[0011] A laparoscope is a long and rigid tube that is attached to a
camera and a light source. Before the laparoscope is inserted, the
patient's abdomen is distended with an injection of carbon dioxide
gas, which allows the surgeon to see the internal organs of the
patient. With the help of the laparoscope and a video display,
which guides the surgeon for the procedure, the surgeon is able to
locate and perform a cholecystectomy. Other small incisions are
made in the abdomen; two of them on the right side below the rib
cage and one in the upper portion below the sternum or the breast.
Many other sophisticated instruments are used to perform the
procedure. For example, two instruments are used to grasp and
retract the gall bladder and a third is used to free the gall
bladder from its attachment. Once the gall bladder is free, the
surgeon then removes it from the patient.
[0012] Under normal conditions, patients recover within a day or
two; however, complications may still occur. The most common
complication that can occur is the injury to common bile duct
(CBD). This complication occurs in about 0.5% of cholecystectomies.
Even though this complication is rare, it demands a better imaging
technique that provides the surgeon with the precise location of
the CBD during a cholecystectomy. The imaging technique should also
be able to distinguish between the CBD and other tissue, providing
a good contrast image.
[0013] During a cholecystectomy, one of the important problems
faced by the surgeon is visibility of structures below the fat
layer. Thus, it is the surgeon's experience which determines the
approximate location of the bile duct. An imaging aid currently
available for intraoperative bile duct visualization is an
Intraoperative Cholangiography (IOC). This technique has hardly
undergone substantial changes since its introduction by Mirizzi in
1937. (See Mirizzi P L. 1937 Operative Cholangiography Surg
Gynaecol Obstet. 1937; 65; 702-710). Routine IOC performed with
every cholecystectomy is one of the strategies used to reduce bile
duct injuries. The rapidly advancing technology of nuclear imaging,
diagnostic ultrasound, MRI and CT have also been used for the
purpose of visualizing the bile duct. For diagnosis of
hepatobiliary disease imaging techniques such as ultrasound and
Magnetic Resonance Cholangiography (MRC) are frequently used.
Endoscopic Retrograde Cholangiography (ERC) is another standard for
visualization of the bile duct. Ultrasound is tolerated by patients
and is cost effective. MRC is superior in visualizing the biliary
system, does not require any contrast agent to visualize the bile
ducts and dilatation and gallstones in CBD are easily detected. For
patients with choedocholithiasis, endoscopic retrograde
cholangio-pancreatography (ERCP) can be utilized. Selective use of
preoperative ERCP has proven to be a good diagnostic tool, as well
as a way to allow clearance of CBD stones when present. (See Qasim
Al-qasabi et al. "Operative Cholangiography in Laproscopic
Cholecystectomy: Is it essential"). One of the faster and widely
available techniques is Computed Tomography Cholangiography (CTC).
A multi-detector CT reports a sensitivity of 65%-88% and a
specificity of 84%-97% to detect gallstones. With the development
of multi-detetcor CT, the resolution of CTC exceeds that of MR.
(See A Persson, N Dahlstrom, O Smedby and T B Brismar:
"Three-dimensional drip infusion CT cholangiography in patients
with suspected obstructive biliary disease: a retrospective
analysis of feasibility and adverse reaction to contrast
material").
[0014] Most histological evaluation of living tissues typically
involves fixation, sectioning, and/or staining to obtain samples
which exhibit high quality images under microscopy. While this
process is current the industry standard, the procedure is time
consuming. It requires removal of tissue from the patient,
processing time, and has inherent sampling error. In addition, the
major limitation of this process is the delay in providing the
surgeon with clinically relevant information at the time of
surgery. Thus, new methods have been developed to complement
existing modalities by providing the surgeon real-time information
that could be used intra-operatively to identify suspect
lesions.
[0015] One such method is hyperspectral imaging. Hyperspectral
imaging is a method of imaging spectroscopy that generates a
gradient map based on local chemical composition. Hyperspectral
imaging has been used in satellite investigation of suspected
chemical weapons production areas, geological features, and the
condition of agricultural fields, and has recently been applied to
the investigation of physiologic and pathologic changes in living
tissue in animal and human studies. Hyperspectral imaging has also
been used in medical applications and has been shown to accurately
predict viability and survival of tissue deprived of adequate
perfusion, and to differentiate diseased (e.g., tumor) and ischemic
tissue from normal tissue.
[0016] One such example can be seen in United States Patent
Application Publication No. 2007/0016079 (Freeman et al.). The '079
application describes methods and systems of hyperspectral and
multispectral imaging of medical tissues. In particular, the '079
application is directed to new devices, tools and processes for the
detection and evaluation of diseases and disorders such as diabetes
and peripheral vascular disease, that incorporate hyperspectral or
multispectral imaging.
[0017] Another example can be found in United States Patent
Application Publication No. 2007/0002276 (Hirohara et al.). The
'276 application describes spectral characteristics that reduce
variation depending on the frequency of received light intensity,
and is gentle on a subject's eye. The invention of the '276
application eliminates displacement between positions of respective
spectral images of the same area even if a change in alignment
occurs between the eye and apparatus during image capture. In the
'276 application, an apparatus for measuring spectral fundus image
includes: an illumination optical system having an illumination
light source that emits a light beam in a specified wavelength
range: a light receiving optical system for forming a fundus image
on the light receiving surface of a photographing section; a liquid
crystal wavelength tunable filter capable of choosing a wavelength
of a transmitted light beam in a specified wavelength range; a
spectral characteristic correction filter having a wavelength
characteristic for correcting the wavelength characteristic of the
emitted light intensity of the illumination light source and the
transmission wavelength characteristic of the wavelength tunable
filter so that the received light intensity on the light receiving
surface is kept within the specified range; and, a data measuring
section for taking the spectral fundus image data from the light
receiving surface while changing the wavelength of the light beam
passing through the wavelength tunable filter.
[0018] Yet another example is described in U.S. Pat. No. 7,199,876
(Mitchell) entitled "Compact Hyperspectral Imager". Here, the '876
patent details a hyperspectral imager including: a first optical
sub-system; at least one slit element; a second optical sub-system;
at least one reflective dispersive element located at a center
plane; and, at least one detecting element located at substantially
an image surface. During operation, the first optical sub-system
images, onto the slit element(s), electromagnetic radiation
originating at a source. The second optical sub-system
substantially collimates, at a center plane, electromagnetic
radiation emanating from the slit element(s). The second optical
sub-system also images, onto the image surface, the electromagnetic
radiation reflected from the reflective dispersive element(s). The
detecting element(s) detect the dispersed electromagnetic radiation
reflected from the reflective dispersive element(s).
[0019] Still yet another example is seen in U.S. Pat. No. 7,167,279
(Otten) entitled "High Efficiency Spectral Imager". The '279 patent
describes optical instruments having, inter alia, optics to process
wavelengths of electromagnetic radiation to produce an
interferogram. The instruments described in the '279 patent include
at least one optical path and optical elements positioned along
this path for splitting and recombining the wavelengths which
interfere with each other to produce a plurality of different
fringes of different wavelengths. In one group, the optics include
matched gratings which are positioned along the optical path
outside of the interferometer optics to produce first and second
sets of spectrally dispersed beams. The interferometer optics also
include a beam splitter and first and second mirrors. The gratings
may be positioned in a variety of locations along the optical path.
In another group, the optics include a beam splitter having a
plurality of surfaces, wherein each of the surfaces is either 100%
reflective, 100% transmissive or 50% reflective and 50%
transmissive. In a third group, the optics include the beam
splitter having a plurality of reflective and transmissive surfaces
and matched gratings. The instruments can all include a detector
for detecting the interferogram and means for processing the
detected interferogram to produce spectral information.
[0020] U.S. Pat. No. 6,198,532 (Cabib) discloses a spectral
bio-imaging method for enhancing pathologic, physiologic, metabolic
and health related spectral signatures of an eye tissue. The method
disclosed in the '532 patent includes the steps of: (a) providing
an optical device for eye inspection being optically connected to a
spectral imager; (b) illuminating the eye tissue with light via the
iris, viewing the eye tissue through the optical device and
spectral imager and obtaining a spectrum of light for each pixel of
the eye tissue; and, (c) attributing each of the pixels a color
according to its spectral signature, thereby providing an image
enhancing the spectral signatures of the eye tissue.
[0021] Another example can be found in U.S. Pat. No. 6,992,775
(Soliz et al.). The '775 patent discloses an ophthalmic instrument
for obtaining high resolution, wide field of area hyperspectral
retinal images for various sized eyes which includes a fundus
retinal imager (which includes optics for illuminating and imaging
the retina of the eye): apparatus for generating a real time image
of the area being imaged and the location of the hyperspectral
region of interest; a high efficiency spatially modulated common
path; a Fourier transform hyperspectral imager (a high resolution
detector optically coupled to the hyperspectral and fundus imager
optics); and, a computer (which is connected to the real time scene
imager, the illumination source, and the high resolution camera)
including an algorithm for recovery and calibration of the
hyperspectral images.
[0022] Finally, in U.S. Pat. No. 7,118,217 (Kardon el al.)
describes an optical imaging device of retinal function to detect
changes in reflectance of near infrared light from the retina of
human subjects in response to visual activation of the retina by a
pattern stimulus. The device of the '217 patent measures changes in
reflectance corresponding in time to the onset and offset of the
visual stimulus in the portion of the retina being stimulated. Any
changes in reflectance can be measured by interrogating the retina
with a light source. The light source may be presented to the
retina via the cornea and pupil or through other tissues in and
around the eye. Different wavelengths of interrogating light may be
used to interrogate various layers of the retina. Additionally,
various patterns and methods of stimulation have been developed for
use with the imaging device and methods.
[0023] The aforementioned hyperspectral imaging systems are slow
for routine clinical practice. In addition, there are no methods
for directly imaging the in vivo level of the biomolecules in live
humans or animals during clinical visits or during surgical (open,
endoscopic or laparoscopic) operations. Accordingly, there is a
need for an improved microscopy system and method that incorporates
superior speed for hyperspectral/multimodal imaging while offering
high spatial resolution and optimized signal sensitivity for fast
image acquisition. The present invention is directed to such a
need.
BRIEF SUMMARY OF THE INVENTION
[0024] In one embodiment, the present invention is a hyperspectral
imaging system that includes one or more optical tunable radiation
sources configured to illuminate one or more fluorescent targets:
one or more light dispersing elements positioned in the optical
path between the one or more optical tunable radiation sources and
one or more detectors; one or more spatial light modulators,
capable of tuning light from the optical radiation source into at
least one spectral band or the band spectrum, placed in the optical
path before the one or more fluorescent targets; an optical
microscope configured to collect the at least one spectral band or
the band spectrum reflected from the one or more fluorescent
targets positioned in the optical path between the optical tunable
radiation sources and the one or more detectors; and a processor
connected to the one or more optical tunable radiation sources, the
one or more detectors to process the at least one spectral band or
a band spectrum reflected from the fluorescent target into image
data, or both. In one aspect, the fluorescent target comprises
natural fluorescence of a tissue or fluid. In another aspect, the
one or more optical tunable radiation sources produce
electromagnetic radiations having wavelengths over the range of 250
nm-2,500 nm. In another aspect, the one or more optical tunable
radiation sources is a supercontinuum laser. In one aspect, the
hyperspectral imaging system is a microscope.
[0025] The system may further comprise at least one deconvolution
algorithm that normalizes each of the at least one spectral band or
the band spectrum at each pixel before image processing. In one
aspect, the spatial light modulator is a digital micromirror
device. In one aspect, the at least one spectral band or the band
spectrum are processed at a video rate. One or more dispersing
elements may be a grating, a prism, a tunable filter, an
electromechanical optical filter wheel, an acousto-optical tunable
filter, a liquid-crystal tunable filter, a digital micromirror
device, or any combination thereof. One or more detectors may be
selected from the group consisting of: a spectrometer, a
two-dimensional array detector, a multi-array detector, an on-chip
amplification CCD camera, a back-illuminated CCD, a liquid nitrogen
cooled CCD detector or a focal plane array, e.g., a CCD, VisGaAs,
InGas or ferromagnetic. In one aspect, the processor comprises an
image data acquisition software that tunes the spatial light
modulator, triggers the one or more detectors for collection of a
series of spectroscopic images formatted as a hyperspectral image
cube and processes the image data for visualization; and one or
more converters to digitize image data, wherein the processor is
connect to one or more displays. In another aspect, the processor
tunes the one or more spatial light modulators, triggers the one or
more detectors for collection of a series of spectroscopic images
formatted as a hyperspectral image cube and processes the image
data for visualization.
[0026] In another aspect, the fluorescence is natural fluorescence,
a fluorescent dye, a fluorescence resonance energy donor, a
fluorescence resonance energy acceptor, a fluorescence quencher or
combinations thereof. Non-limiting examples of one or more
fluorescence molecules for use with the present invention include
indocyanine green, 5-carboxyfluorescein (5-FAM),
6-carboxyfluorescein (6-FAM), fluorescein-5-isothiocyanate (FITC).
2'7'-dimethoxy-4'5'-dichloro-6-carboxyfluorescein (JOE); rhodamine
and rhodamine derivatives such as
N,N,N',N'-tetramethyl-6-carboxyrhodamine (TAMRA),
6-carboxyrhodamine (R6G), tetramethyl-indocarbocyanine (Cy3),
tetramethyl-benzindocarbocyanine (Cy3.5),
tetramethyl-indodicarbocyanine (Cy5),
tetramethyl-indotricarbocyanine (Cy7), 6-carboxy-X-rhodamine (ROX);
hexachloro fluorescein (HEX), tetrachloro fluorescein TET;
R-Phycoerythrin, 4-(4'-dimethylaminophenylazo) benzoic acid
(DABCYL), and 5-(2'-aminoethyl)aminonaphthalene-1-sulfonic acid
(EDANS) or combinations thereof.
[0027] In another aspect, the processor is defined further as: a
computer having image data processing software capable of
processing various fluorescence and digital signals; a digital
signal processor comprising algorithms for fluorescence analysis;
and, an algorithm library for visualizing chemistry on and/or
within the imaged one or more fluorescent targets for detection,
monitoring and diagnosis. In one aspect, the one or more
illumination optics comprises lenses, microscopes, dissection
microscopes, surgical microscopes, slit lamps, endoscopes,
laparoscopes, colonoscopes or any combination thereof. In another
aspect, the processor is in communication with a database
comprising a library spectrum of fluorescent targets. In yet
another aspect, the one or more fluorescent targets comprises a
tissue, in the bloodstream, in the lymphatic system, in bodily
secretions, in biopsies, on the skin, by fiber optic transdermal
spectroscopy on a blood vessel, on the skin or at the retina, by
laparoscopy, by endoscopy, in the central nervous system or
combinations thereof. Moreover, in still yet another aspect, the
one or more fluorescent targets comprises an organ. In another
aspect, the system can be configured for real-time in vivo analysis
of the one or more fluorescent targets.
[0028] Another embodiment of the present invention is an apparatus
and method for obtaining spectral image data from one or more
fluorescent targets in vivo comprising: generating a spectrum of
electromagnetic radiation; separating the electromagnetic radiation
into at least one spectral band or a band spectrum using a spatial
light modulator; illuminating the one or more fluorescent targets
with the at least one spectral band or the band spectrum;
collecting the reflected at least one spectral band or the band
spectrum resulting from the one or more fluorescent targets; and,
directing the collected at least one spectral band or the band
spectrum to one or more detectors that capture spectral image data.
The method may further comprise the steps of: tuning the spatial
light modulator; triggering the one or more detectors for
collection; formatting and digitizing the spectral image data as a
hyperspectral image; deconvoluting each pixel of the spectral image
data using a signal processing algorithm; and, displaying the
spectral image on one or more fluorescent targets. In one aspect,
the one or more fluorescent targets comprise tissue components that
have natural or native fluorescence.
[0029] Another embodiment of the present invention is a method for
hyperspectral surgery by capturing image data of one or more
fluorescent targets generating a wide spectrum electromagnetic
radiation comprising: separating the electromagnetic radiation into
at least one spectral band or a band spectrum using a spatial light
modulator; illuminating the one or more fluorescent targets with
the at least one spectral band or the band spectrum; collecting the
reflected at least one spectral band or the band spectrum resulting
from the one or more fluorescent targets; and, directing the
collected at least one spectral band or the band spectrum to one or
more detector to form the image data, wherein the hyperspectral
surgery is used in surgical procedures selected from a
cholecystectomy, amputation, burn, skin flap evaluation,
visualizing areas of angiogenesis, probes that bind antigens and
absorbs near-infrared during pathological evaluations and in vivo,
quality control of pharmaceuticals, monitoring vascular changes and
drug discovery in response to pharmaceuticals, monitoring diabetic
retinopathy, diseases such as cancer, diabetes, sickle cell,
anemia, bilirubin, raynauds, ulcers, burns, skin flaps, surgery,
gallbladder, brain, monitoring wound healing, and early detection
of wound infections.
[0030] Another embodiment of the present invention is a method for
hyperspectral surgery of a human subject by capturing image data of
one or more fluorescent targets generating a wide spectrum
electromagnetic radiation comprising: covering the human subject
with NIR transparent material to provide privacy, accessibility and
maintain body heat; separating the electromagnetic radiation into
at least one spectral band or the band spectrum using a spatial
light modulator; illuminating the fluorescent target with the at
least one spectral band or the band spectrum; collecting the
reflected at least one spectral band or the band spectrum resulting
from the one or more fluorescent targets; and, collecting at least
one spectral band or the band spectrum to one or more detector to
form the image data, wherein the hyperspectral surgery is used in
surgical procedures selected from a cholecystectomy, amputation,
burn, skin flap evaluation, visualizing areas of angiogenesis,
probes that bind antigens and absorbs near-infrared during
pathological evaluations and in vivo, quality control of
pharmaceuticals, monitoring vascular changes and drug discovery in
response to pharmaceuticals, monitoring diabetic retinopathy,
cancer, diabetes, sickle cell, anemia, bilirubin, raynauds, ulcers,
burns, skin flaps, surgery, gallbladder, brain, monitoring wound
healing, measuring oxygenation of retina, measuring optic nerve
oxygenation, measuring macular pigments, measuring pigments in
retinal photoreceptors and retinal pigment epithelium, diagnosis of
autoimmune retinitis, diagnosis of infectious retinitis, diagnosis
of infiltrative neoplastic conditions, evaluating disease
biomarkers, diagnosis of ocular trauma injuries, measuring
oxygenation of a kidney and early detection of wound
infections.
[0031] Most of the art discussed above can either collect all the
spectral information and scan the field of view, or collect all the
field of view and scan through the spectral range. Unlike the prior
art, the present invention can collect all the spectral information
and scan the field of view simultaneously by illuminating with a
spectrum, a spectroscopic wavelength, or multiple wavelengths and
collect the image data, along with chemometrics on the source side
and/or on the detector side.
[0032] One embodiment of the present invention is a hyperspectral
imaging system having one or more optical radiation sources
configured to illuminate one or more objects, illumination optics,
one or more dispersing elements, one or more spatial light
modulators that are capable of tuning light from the optical
radiation source into at least one spectral band or a band spectrum
(in other words, one or more single wavelength bands or a single
multi-wavelength band), and positioned in the optical path before
the one or more objects, an optical microscope configured to
collect the at least one spectral band or the band spectrum
reflected from the one or more objects, one or more detectors; and,
a processor to process the reflected at least one spectral band or
the band spectrum to provide an enhanced image. The present
invention system images may also include, but are not limited to,
organs and tissue components.
[0033] The optical radiation source of the system is capable of
producing photons having wavelengths in the range of 250 nm-2,500
nm. The spatial light modulator of the system may be a Digital
Micromirror Device (DMD). In certain embodiments, the dispersing
element of the present system may be a grating or a prism. The one
or more detectors can also include at least one detector selected
from: a spectrometer, a two-dimensional array detector, a
multi-array detector, an on-chip amplification CCD camera, a
back-illuminated CCD, a liquid nitrogen cooled CCD detector or a
focal plane array, e.g., a CCD, VisGaAs, InGas or
ferromagnetic.
[0034] In addition, the processor(s) of the present invention
include, but are not limited to, a computer having image processing
software that will tune the spatial light modulator and trigger the
one or more detectors for collecting a series of spectroscopic
images formatted as a hyperspectral image cube, one or more
displays, and a database of characterized objects.
[0035] The present system may also have one or more liquid crystal
tunable filters positioned in the optical path, and can even be
configured as a portable endoscopic system. In certain embodiments,
the present invention can be configured for real-time in vivo
analysis of the one or more objects.
[0036] The present invention also relates to a method for obtaining
spectral image data from an object by generating a wide spectrum
electromagnetic radiation, separating the electromagnetic radiation
into at least one spectral band or a band spectrum using a spatial
light modulator, illuminating the object with the at least one
spectral band or the band spectrum electromagnetic radiations,
collecting the reflected at least one spectral band or the band
spectrum electromagnetic radiations resulting from the object, and
directing the collected at least one spectral band or the band
spectrum electromagnetic radiation image to one or more detector
array for forming the spectral image data.
[0037] In addition, the reflected one or more single wavelength
electromagnetic radiations may include, but are not limited to:
electromagnetic radiation reflected, refracted, luminescence,
fluorescence, autofluorescence, Raman scattered, transmitted,
scattered, adsorbed, or emitted by the sample. In one aspect, the
present invention includes one or more slit lamps. The present
invention can also operate as a dissection microscope for clinical
and surgical applications. In another aspect, the present invention
includes a laparoscopic system for creating a library of
chemometric deconvolutions so that a clinician can simply switch
through various algorithms depending on applications.
[0038] In another embodiment, the present invention includes a
hyperspectral imaging system including: one or more optical tunable
radiation sources configured to illuminate one or more objects,
e.g., tissue components; one or more detectors; one or more
dispersing elements positioned in the optical path between the one
or more optical tunable radiation sources and the one or more
detectors; one or more spatial light modulators, capable of tuning
light from the optical radiation source into at least one spectral
band or a band spectrum; an optical microscope configured to
collect one or more single wavelength spectral band or a spectrum
band reflected from the one or more objects positioned in the
optical path between the optical tunable radiation sources and the
one or more detectors; one or more processors connected to the one
or more optical tunable radiation sources, the one or more
detectors, the one or more dispersing element, the one or more
spatial light modulator, the optical microscope, or any combination
thereof, wherein the one or more processors process the reflected
at least one spectral band or the band spectrum as image data and
to provide enhanced images. The one or more optical tunable
radiation sources are typically capable of producing
electromagnetic radiations having wavelengths over the range of 250
nm-2,500 nm and can be a supercontinuum laser.
[0039] In some aspects, the spatial light modulator is a digital
micromirror device, and the one or more dispersing element can be a
grating, a prism, a tunable filters, an electromechanical optical
filter wheel, an acousto-optical tunable filter, a liquid-crystal
tunable filter, a digital micromirror device, or any combination
thereof. In addition, the one or more detectors can have at least
one detector selected from: a spectrometer, a two-dimensional array
detector, a multi-array detector, an on-chip amplification CCD
camera, a back-illuminated CCD, a liquid nitrogen cooled CCD
detector or a focal plane array.
[0040] In another aspect, the processor of the present invention
can further include at least one processor having an image data
acquisition software that tunes the spatial light modulator,
triggers the one or more detectors for collection of series of
spectroscopic images formatted as a hyperspectral image cube and
processes the image data for visualization; one or more transducer
to digitize image data; and, one or more display. The processor can
further include a dedicated processor that tunes the spatial light
modulator, triggers the one or more detectors for collection of
series of spectroscopic images formatted as a hyperspectral image
cube and processes the image data for visualization. The processor
may be further defined as a computer having image data processing
software capable of processing various chemometric and digital
signals; a digital signal processor comprising algorithms for
chemometric analysis; and, an algorithm library for visualizing
chemistry on and within the imaged one or more objects for
detection, monitoring and diagnosis. In addition, the processor can
be connected to a database having a library spectrum of
characterized one or more objects. In another aspect, the one or
more illumination optics of the present invention can include
lenses, microscopes, dissection microscopes, surgical microscopes,
slit lamps, endoscopes, laparoscopes, colonoscopes or any
combinations thereof.
[0041] In some aspects, the system of the present invention can
further include one or more liquid crystal tunable filter and can
be configured as a portable endoscopic system. The present
invention can also be configured for real-time in vivo analysis of
the one or more objects.
[0042] Yet in another aspect, the present invention includes a
method for obtaining spectral image data from one or more objects
in vivo (e.g., tissue components) including: generating a spectrum
of electromagnetic radiation; separating the electromagnetic
radiation into one or more single wavelength bands or a spectrum
band using a spatial light modulator; illuminating the one or more
objects with the one or more single wavelength bands or the
spectrum band; collecting the reflected one or more single
wavelength bands or the spectrum bands resulting from the one or
more objects; and, directing the collected single wavelength bands
or the spectrum bands to one or more detectors that capture
spectral image data. The method can further include the steps of:
tuning the spatial light modulator; triggering the one or more
detectors for collection; formatting and digitizing the spectral
image data as a hyperspectral image; deconvoluting each pixel of
the spectral image data using a signal processing algorithm; and,
displaying the spectral image on one or more displays.
[0043] In some aspects, the method uses spectrum of electromagnetic
radiation comprises photons having wavelengths over the range of
250 nm-2,500 nm, or a supercontinuum laser. The spatial light
modulator of the present invention can include a grating, a prism,
a tunable filter, an electromechanical optical filter wheel, an
acousto-optical tunable filter, a liquid-crystal tunable filter, a
digital micromirror device, or any combination thereof. The
detector for the method can have at least one detector selected
from: a spectrometer, a two-dimensional array detector, a
multi-array detector, an on-chip amplification CCD camera, and a
back-illuminated CCD, a liquid nitrogen cooled CCD detector or a
focal plane array. It should be appreciated that the present
invention is not limited to foregoing detectors, but may include
any type of electro-optical detection elements. The reflected one
or more single wavelength band or the spectrum bands of the present
invention can include reflected radiation, luminescence,
fluorescence, autofluorescence, Raman scattered, transmitted,
scattered, adsorbed or emitted electromagnetic radiation by the one
or more objects.
[0044] In another aspect, the present invention includes a method
for hyperspectral surgery by capturing image data of one or more
object generating a wide spectrum electromagnetic radiation
including, by not limited to the following steps: separating the
electromagnetic radiation into one or more single wavelength bands
or a spectrum band using a spatial light modulator; illuminating
the object with the one or more single wavelength bands or the
spectrum band; collecting the reflected one or more single
wavelength bands or the spectrum band resulting from the one or
more object; and, directing the collected or more single wavelength
bands or spectrum band to one or more detector to form the image
data, wherein the hyperspectral surgery is used in a surgical
procedure selected from the group consisting of: a cholecystectomy,
amputation, burn, skin flap evaluation, visualizing areas of
angiogenesis, probes that bind antigens and absorbs near-infrared
during pathological evaluations and in vivo, quality control of
pharmaceuticals, monitoring vascular changes and drug discovery in
response to pharmaceuticals, monitoring diabetic retinopathy,
diseases such as cancer, diabetes, sickle cell, anemia, bilirubin,
raynauds, ulcers, burns, skin flaps, surgery, gallbladder, brain,
monitoring wound healing, measuring oxygenation of retina,
measuring optic nerve oxygenation, measuring macular pigments,
measuring pigments in retinal photoreceptors and retinal pigment
epithelium, diagnosis of autoimmune retinitis, diagnosis of
infectious retinitis, diagnosis of infiltrative neoplastic
conditions, evaluating disease biomarkers, diagnosis of ocular
trauma injuries, measuring oxygenation of a kidney and early
detection of wound infections. The method can further include the
step of covering the human subject with NIR transparent material to
provide privacy, accessibility and maintain body heat.
[0045] The present invention further includes a hyperspectral
imaging system having an optical path, where the system includes an
illumination source adapted to output a light beam, the light beam
illuminating a target, a dispersing element arranged in the optical
path and adapted to separate the light beam into a plurality of
wavelengths, a digital micromirror array adapted to tune the
plurality of wavelengths into a spectrum, an optical device having
a detector and adapted to collect the spectrum reflected from the
target and arranged in the optical path and a processor operatively
connected to and adapted to control at least one of: the
illumination source; the dispersing element; the digital
micromirror array; the optical device; and, the detector, the
processor further adapted to output a hyperspectral image of the
target. The dispersing element is arranged between the illumination
source and the digital micromirror array, the digital micromirror
array is arranged to transmit the spectrum to the target and the
optical device is arranged in the optical path after the
target.
[0046] In some embodiments, the present invention hyperspectral
imaging system further includes collimating optics adapted to
transmit the light beam from the dispersing element to the digital
micromirror array as a collimated light beam. In other embodiments,
the present invention hyperspectral imaging system further includes
beam shaping optics adapted to transmit the light beam from the
digital micromirror array to the target so that the light beam
substantially illuminations all of the target.
[0047] The present invention further includes a method of obtaining
a hyperspectral image of a target including the steps of:
generating a beam of light; dispersing the light beam with a
dispersing element; separating the dispersed light beam into a
first complex spectrum using a spatial light modulator;
illuminating the target with the first complex spectrum, wherein
the first complex spectrum subsequently reflects off of the target
as a first reflected light beam; collecting the first reflected
light beam; directing the collected first reflected light beam to a
detector to capture a first spectral image data; separating the
dispersed light beam into a second complex spectrum using the
spatial light modulator; illuminating the target with the second
complex spectrum, wherein the second complex spectrum subsequently
reflects off of the target as a second reflected light beam;
collecting the second reflected light beam; directing the collected
second reflected light beam to the detector to capture a second
spectral image data; separating the dispersed light beam into a
third complex spectrum using the spatial light modulator;
illuminating the target with the third complex spectrum, wherein
the third complex spectrum subsequently reflects off of the target
as a third reflected light beam; collecting the third reflected
light beam; directing the collected third reflected light beam to
the detector to capture a third spectral image data; and, forming
the hyperspectral image from the first, second and third spectral
image data.
[0048] In some embodiments, the first, second and third reflected
light beams each includes reflected, luminescence, fluorescence,
autofluorescence, Raman scattered, transmitted, scattered,
adsorbed, or emitted electromagnetic radiation. In other
embodiments, the detector includes a processor having an image data
acquisition software adapted to tune the spatial light modulator,
trigger the detector for collection of the first, second and third
spectral image data formatted as a hyperspectral image cube and
process the hyperspectral image cube for visualization, a digital
signal process algorithm for analyzing chemometrics of the target
and a display device adapted to display the hyperspectral image. In
still other embodiments, the method is adapted for use in
hyperspectral surgery, and the hyperspectral surgery is used in
surgical procedures selected from the group consisting of: a
cholecystectomy; an amputation; a burn; a skin flap evaluation;
visualizing areas of angiogenesis; probes that bind antigens and
absorbs near-infrared during pathological evaluations; in vivo
quality control of pharmaceuticals; monitoring vascular changes and
drug discovery in response to pharmaceuticals; monitoring diabetic
retinopathy and diseases such as cancer, diabetes, sickle cell,
anemia, bilirubin, raynauds, ulcers, burns, skin flaps, surgery,
gallbladder, brain; monitoring wound healing; measuring oxygenation
of retina; measuring optic nerve oxygenation; measuring macular
pigments; measuring pigments in retinal photoreceptors and retinal
pigment epithelium; diagnosis of autoimmune retinitis; diagnosis of
infectious retinitis; diagnosis of infiltrative neoplastic
conditions; evaluating disease biomarkers; diagnosis of ocular
trauma injuries; measuring oxygenation of a kidney; and, early
detection of wound infections. In still yet other embodiments, the
first, second and third complex spectrum each include a plurality
of intensities of a plurality of wavelengths.
[0049] These and other objects and advantages of the present
invention will be readily appreciable from the following
description of preferred embodiments of the invention and from the
accompanying drawings and claims.
BRIEF DESCRIPTION OF THE DRAWINGS
[0050] The nature and mode of operation of the present invention
will now be more fully described in the following detailed
description of the invention taken with the accompanying drawing
figures, in which:
[0051] FIG. 1 is a fluorescent imager for detecting, for example,
ICG and bilirubin;
[0052] FIG. 2 is a short pass filter transmission characteristics
used as an excitation filter for bilirubin;
[0053] FIG. 3 is a long pass filter transmission characteristics
used as an excitation filter for bilirubin;
[0054] FIG. 4 is a diagram of a 2.times.2 binning process;
[0055] FIG. 5 is a block diagram depicting a method for computing
the spatial resolution of the system;
[0056] FIG. 6A is a percent contrast against spatial resolution
without an emission filter at the detector;
[0057] FIG. 6B is a percent contrast against spatial resolution
with an emission filter at the detector;
[0058] FIG. 7 is a block diagram depicting the set up used for
measurement of absorption spectrum of ICG;
[0059] FIG. 8 is a block diagram depicting the set up used for
measuring the best concentration of ICG in Water;
[0060] FIG. 9A is a digital image of beef fat with an ICG in water
filled capillary taken from a digital image;
[0061] FIG. 9B is an image of beef fat with an ICG in water filled
capillary taken using a present invention surgical fluorescence
imager;
[0062] FIG. 10 is a SNR and contrast to background calculation
wherein a profile is taken from each row of the image and SNR and
contrast to background are then calculated for each row and
averaged;
[0063] FIG. 11A is a plot depicting the signal to noise ratio
against penetration depth in intralipid of Aqueous ICG with
constant exposure;
[0064] FIG. 11B is a plot depicting the signal to noise ratio
against penetration depth in intralipid of Aqueous ICG with
variable exposure;
[0065] FIG. 12 is a plot depicting the signal to noise ratio
against penetration depth in intralipid of ICG with Human Bile with
Constant Exposure;
[0066] FIG. 13A is a plot depicting the contrast to background
ratio against penetration depth in intralipid of Aqueous ICG with
constant exposure;
[0067] FIG. 13B is a plot depicting the contrast to background
ratio against penetration depth in intralipid of Aqueous ICG with
variable exposure;
[0068] FIG. 14 is a plot depicting the contrast to background ratio
against penetration depth in intralipid of ICG with human bile
under constant exposure;
[0069] FIG. 15 is a block diagram depicting a set up for
measurement of fluorescence from bilirubin;
[0070] FIG. 16 is a fluorescence image of bilirubin with excitation
from 400 nm-500 nm compared to a blank capillary;
[0071] FIG. 17A is a plot depicting percent contrast against
spatial resolution, when no emission filter is used;
[0072] FIG. 17B is a plot depicting percent contrast against
spatial resolution, when an emission filter is used;
[0073] FIG. 18 is a hyperspectral imager for use with the present
invention;
[0074] FIG. 19 is an illustration of hyperspectral image cube
visualization;
[0075] FIG. 20 is an embodiment of the present invention
hyperspectral imaging system;
[0076] FIG. 21 is a diagram of an example binning in a CCD
camera;
[0077] FIG. 22 is another diagram of an example binning in a CCD
camera;
[0078] FIG. 23 is yet another diagram of example binnings in a CCD
camera;
[0079] FIG. 24A is a graph showing a comparison of actual measured
hemoglobin versus the predicted hemoglobin using pure hemoglobin
samples and the actual measured hemoglobin using the present
invention;
[0080] FIG. 24B is a graph showing a comparison of actual measured
hemoglobin versus the predicted hemoglobin using pure hemoglobin
samples and the actual measured hemoglobin using the present
invention;
[0081] FIG. 25 is a schematic diagram of a microscopic system;
[0082] FIG. 26 is a diagram of the visible low resolution LCTF
calibration setup;
[0083] FIG. 27 is a plot of visible low resolution LCTF calibration
curve;
[0084] FIG. 28 is a plot of bandpass and wavelength;
[0085] FIG. 29 is a schematic of tune-wait example setup;
[0086] FIG. 30 is a graph of visible LCTF tune-wait with normalized
intensity and its relationship with wavelength;
[0087] FIG. 31A is an image of a portion of the resolution
target;
[0088] FIG. 31B is a graph of the reflected intensity corresponding
to the target;
[0089] FIG. 32 is a graph of percent contrast and its relationship
with spatial resolution with relay optics;
[0090] FIG. 33 is a graph of percent contrast and its relationship
with spatial resolution without relay optics;
[0091] FIG. 34 is an image of color coded hyperspectral image:
[0092] FIG. 35 is a schematic of an embodiment of the present
invention;
[0093] FIG. 36 is an image of oxyhemoglobin contribution taken with
reflectance hyperspectral imaging system of the present
invention;
[0094] FIG. 37 is an image of small vessels within a human
conjunctive of an eye taken with hyperspectral imaging system;
[0095] FIG. 38 is an example image taken on an occluded fingers
using a DLP HSI two-shot method:
[0096] FIG. 39 is a graph that shows one hundred twenty six (126)
separate wavelengths using an LCTF to separate the bands prior to
illumination, after which an image is captured with each bandpass
or with a single frame including up to one hundred twenty six (126)
images;
[0097] FIG. 40 is a graph showing a band spectrum in which a
digital micromirror array was used to create spectral illumination
that allows for a lower number of images per frame:
[0098] FIG. 41 is a graph of a comparison of data obtained with the
LCTF and the digital Micromirror array illumination using the
single bandwidths of FIG. 39;
[0099] FIG. 42 is a graph of a comparison of data obtained with the
LCTF and the digital Micromirror array illumination using the
single bandwidths of FIG. 40;
[0100] FIG. 43 is a flow chart of the basic 2-shot algorithm:
[0101] FIG. 44 is a flow chart of the processing of the data cube
obtained using the basic 2-shot method;
[0102] FIG. 45 is a flow chart of the acquisition method of the
basic 2-shot method;
[0103] FIG. 46 is a sample of images obtained for a finger
occlusion at different times using the present invention;
[0104] FIG. 47 is a comparison of images obtained using visible
light and Near infrared (NIR) of the reperfusion of a foot
following removal of the shoe;
[0105] FIG. 48 is the in vivo hyperspectral imaging of human
tissue, spatial variation of percentage of HbO.sub.2 and surface
temperature in response to burn;
[0106] FIG. 49 is a schematic illustration of the different
elements of a DLP.RTM. hyperspectral imaging system of the present
invention;
[0107] FIG. 50A is the normalized absorbtion spectra of HbO.sub.2
and Hb used as references for multivariate deconvolution of
spectral sweep absorbance cubes;
[0108] FIG. 50B is the positive and negative subtraction of spectra
used for "3 shot" illumination;
[0109] FIG. 51 is a block diagram that shows the experimental
procedure used to capture images using the "3 shot" method of the
present invention and the MATLAB.RTM. algorithm used to process the
images captured using the "3 shot" method of the present
invention;
[0110] FIG. 52A is an illustration of the "3 shot" illumination
method of the present invention used for visualizing blood
oxygenation showing the normalized absorbance spectra in the 520
nm-645 nm wavelength range for HbO.sub.2 and Hb subtracted from
each other, and the positive areas become the two illumination
spectra;
[0111] FIG. 52B is an illustration of the "3 shot" illumination
method of the present invention used for visualizing blood
oxygenation showing relative intensity of each illumination
spectrum is stretched from 0 to 100 to maximize the overall light
intensity and match the required OL-490 input format;
[0112] FIG. 52C is an illustration of the "3 shot" illumination
method of the present invention used for visualizing blood
oxygenation showing the absorbers and scatterers that are not
HbO.sub.2 and Hb;
[0113] FIG. 53A is a graph illustrating the comparison of optical
output measured by a spectrometer and desired optical output for
first illumination spectrum of the "3 shot" method before applying
center wavelength calibration and intensity adjustment;
[0114] FIG. 53B is a graph illustrating the comparison of optical
output measured by a spectrometer and desired optical output for
first illumination spectrum of the "3 shot" method after applying
center wavelength calibration and intensity adjustment;
[0115] FIG. 54 is a timing diagram of hyperspectral acquisition
where N is the total number of slices to be acquired, for a
spectral sweep, N=126, and for a "3 shot", N=3;
[0116] FIG. 55A is a diagram illustrating the processing algorithms
for visualizing the images of blood oxygenation captured using the
spectral sweep method comparing measured spectrum to reference
spectra via multivariate least squares analysis to quantify
relative concentration HbO.sub.2;
[0117] FIG. 55B is a diagram illustrating the processing algorithms
for visualizing the images of blood oxygenation captured using the
"3 shot" method of the present invention which subtracts the image
representing Hb absorbance from the image representing HbO.sub.2
absorbance and dividing the broadband absorbance to quantify
relative concentration of HbO.sub.2;
[0118] FIG. 56A is the visualization of an ischemia induced by
occluding blood flow to a finger imaged with the DLP HSI in
spectral sweep mode;
[0119] FIG. 56B is the visualization of an ischemia induced by
occluding blood flow to a finger imaged with the DLP HSI in "3
shot" mode:
[0120] FIG. 57A is the average of five spectral sweep oxyz output
images captured as `Control`;
[0121] FIG. 57B is the average of five spectral sweep oxyz output
images captured while `Occluded`;
[0122] FIG. 57C is the average of five spectral sweep oxyz output
images captured for `Reperfusion`;
[0123] FIG. 58A is a color-coded "3 shot" output images captured
immediately before cutting the rubber band;
[0124] FIG. 58B is a color-coded "3 shot" output images captured 10
seconds after cutting the rubber band;
[0125] FIG. 58C is a color-coded "3 shot" output images captured 2
minutes after cutting the rubber band;
[0126] FIG. 59 is the real-time progression of reactive hyperemia
after cutting rubber band from finger shown as a plot of average
pixel values from "3 shot" images acquired during removal of
tourniquet from finger which shows reactive hyperemia in at 3
frames per second;
[0127] FIG. 60 is an image of a hyperspectral imaging setup in
animal lab surgical suite at UTSW for imaging porcine partial
nephrectomy;
[0128] FIG. 61A is an image of a kidney showing a `Control`;
[0129] FIG. 61B is an image of a kidney showing `Occluded`;
[0130] FIG. 61C is a spectral sweep image of a kidney showing a
`Control`;
[0131] FIG. 61D is a spectral sweep image of a kidney showing
`Occluded`;
[0132] FIG. 61E is a "3 Shot" image of a kidney showing a
`Control`;
[0133] FIG. 61F is a "3 Shot" image of a kidney showing
`Occluded`;
[0134] FIG. 64 is a graph plotting monitoring the time progression
percentage of oxyhemoglobin perfusing the kidney before and during
renal AO and AV occlusion;
[0135] FIG. 63A is a raw image of a bilateral ischemia exhibited in
a pig kidney having two renal arteries taken immediately after
clamping one renal artery;
[0136] FIG. 63B is an image of a bilateral ischemia exhibited in a
pig kidney having two renal arteries taken with the "3 shot" method
after clamping one renal artery showing that the lower pole is
ischemic, but the upper pole is still highly oxygenated;
[0137] FIG. 63C is an image of a bilateral ischemia exhibited in a
pig kidney having two renal arteries taken with the "3 shot" method
after clamping the second renal artery showing that the upper pole
becomes ischemic as well;
[0138] FIG. 64A is a raw image of a human kidney having partial
nephrectomy;
[0139] FIG. 64B is an image of a human kidney having partial
nephrectomy taken using the "3 Shot" method before clamping:
[0140] FIG. 64C is an image of a human kidney having partial
nephrectomy taken using the "3 Shot" method immediately after
clamping;
[0141] FIG. 64D is an image of a human kidney having partial
nephrectomy taken using the "3 Shot" method after the tumor is
removed and still clamped;
[0142] FIG. 64E is an image of a human kidney having partial
nephrectomy taken using the "3 Shot" method immediately after
clamping;
[0143] FIG. 64F is an image of a human kidney having partial
nephrectomy taken using the "3 Shot" method after dissection;
[0144] FIG. 65 is a graph showing the progression of relative
oxygenation measured from output images in FIGS. 23B-23E;
[0145] FIG. 66 is a digital image of a full leg with a superimposed
spectral sweep image mapping the surface oxygenation of a patient
with lower limb neuropathy;
[0146] FIG. 67 is a graph showing the spatial progression of
surface oxygenation as measured along the centerline of the leg in
the hyperspectral images in FIG. 66 wherein the dashed line
represents the limit of neuropathy as determined by a physician
administering a clinical nervous response exam on the leg;
[0147] FIG. 68A is a control image of a rabbit brain taken with the
"3 shot" method with the black sample area showing normal tissue
and blue sample area showing damaged tissue;
[0148] FIG. 68B is an image of a rabbit brain taken with the "3
shot" method after oxygen supply is cut off to the rabbit wherein
the normal tissue becomes ischemic and there is no change in the
damaged tissue;
[0149] FIG. 69A is an image of a human hand having burned regions
thereon;
[0150] FIG. 69B is a hyperspectral image of the human hand of FIG.
69A chemically encoded to show oxyhemoglobin;
[0151] FIG. 69C is a hyperspectral image of the human hand of FIG.
69A chemically encoded to show water;
[0152] FIG. 70A is an representation of a portion of a digital
micromirror device showing a single column of micromirrors in an
"on" position;
[0153] FIG. 70B is a representation of an illumination spectrum
produced by the digital micromirror device of FIG. 70A;
[0154] FIG. 71A is an representation of a portion of a digital
micromirror device showing a single column of micromirrors in an
"on" position;
[0155] FIG. 71B is a representation of an illumination spectrum
produced by the digital micromirror device of FIG. 71A;
[0156] FIG. 72A is an representation of a portion of a digital
micromirror device showing a plurality of columns of micromirrors
in an "on" position where respective columns have different numbers
of micromirrors in an "on" position; and,
[0157] FIG. 72B is a representation of a complex illumination
spectrum produced by the digital micromirror device of FIG.
72A.
DETAILED DESCRIPTION OF THE INVENTION
[0158] At the outset, it should be appreciated that like drawing
numbers on different drawing views identify identical, or
functionally similar, structural elements of the invention. While
the present invention is described with respect to what is
presently considered to be the preferred aspects, it is to be
understood that the invention as claimed is not limited to the
disclosed aspects.
[0159] Furthermore, it is understood that this invention is not
limited to the particular methodology, materials and modifications
described and as such may, of course, vary. It is also understood
that the terminology used herein is for the purpose of describing
particular aspects only, and is not intended to limit the scope of
the present invention, which is limited only by the appended
claims.
[0160] Unless defined otherwise, all technical and scientific terms
used herein have the same meaning as commonly understood to one of
ordinary skill in the art to which this invention belongs. As used
herein, DLP and Digital Micromirror Device (DMD) are used
interchangeably. A DMD chip has on its surface several hundred
thousand microscopic mirrors arranged in a rectangular array which
correspond to the pixels in the image to be displayed. The mirrors
can be individually rotated .+-.10-12.degree., to an on or off
state. In the on state, light from the projector bulb is reflected
into the lens making the pixel appear bright on the screen. In the
off state, the light is directed elsewhere (usually onto a heat
sink), making the pixel appear dark. It can be purchased
commercially from Texas Instruments. As used herein.
"hyperspectral" in reference to an imaging process or method is
meant the acquisition of an image at more than one wavelength or
bands of wavelengths. Furthermore, as used herein, "spectra" is
defined as illumination with spectra of light with multiple
wavelengths within a spectral range. Although any methods, devices
or materials similar or equivalent to those described herein can be
used in the practice or testing of the invention, the preferred
methods, devices, and materials are now described. Still further,
as used herein, "hyperspectral image" is defined as an image
derived from hyperspectral image data, e.g., hyperspectral data
cube, which includes but is not limited to, a chemically encoded
image, a principle component analysis (PCA) image and wavelength
dependent images.
[0161] A spectral illuminator using DLP technology may be used to
excite fluorescence markers, which are native to the tissue (tissue
fluorescence) or target, or markers that were injected or painted
onto the tissue or target. An optical filter, Liquid Crystal
Tunable Filter (LCTF) or other element can be used to pass emitted
fluorescence light to a detector. For imaging bile ducts, e.g.,
ICG, indocyanine green can be injected or the native fluorescence
properties of bile can be used. With improved imaging, surgical
procedures can focus on the target for treatment (or the adjacent
tissue) by visualization, e.g. the biliary tree. The system can be
used for other surgical applications to perform in vivo pathology
during surgery, i.e., in vivo hyperspectral pathology. The present
invention can also be used in other methods or techniques,
including but not limited to, in vivo pathology of cancer in
real-time (e.g. before, during or after surgery), in the body
(e.g., laparoscopically), without having to biopsy and/or fix the
tissue or target to a slide. The example being a probe that binds
to antigens. The techniques described infira are also applicable to
fluorescence microscopy. Finally, chemometrics and digital signal
processing may be used to produce chemically encoded images to
enhance visualizations. Furthermore, it has been found that using a
digital light processor (DLP), the present inventors were able to
achieve video rate hyperspectral image capture and processing. Such
video rate capture and processing is not possible with current
illumination technology: however, is made possible by the use of
DLP technology as set forth herein. The video capture rate also
benefits from the chemometric illumination.
[0162] In addition to the above described disadvantages of present
imaging apparatus and techniques, the following describes still
further disadvantages. ERC is an invasive, user-dependent modality
which may also induce pancreatitis. Though ultrasound is cost
effective, the images are not clearly understood by clinicians. MRC
is also often inconclusive in patients with air in the biliary
system. Proponents of routine intraoperative cholangiography (IOC)
claim that this practice lowers risk of CBD injuries and leads to
fewer retained bile duct stones. (See Flowers J L, Zucker K L,
Graham S M, et al. "Laparoscopic cholangiography: results and
indications." Ann Surg 1992; 215:209-16; and, Gerber A. Apt MK.
"The case against routine operative cholangiography. AM J Surg"
1982; 143:734-6). The disadvantages of routine IOC include
increased costs, operative time and false positive findings,
leading to unnecessary efforts to clear the CBD stones. Injecting
contrast media for a CTC can led to adverse effects such as
anaphylaxis, urticaria and respiratory distress. (See A Persson. N
Dahlstrom, O Smedby and T B Brismar: "Three-dimensional drip
infusion CT cholangiography in patients with suspected obstructive
biliary disease: a retrospective analysis of feasibility and
adverse reaction to contrast material"). These can be reduced if
the contrast agents are diffused instead of injected. Possible
explanations of infrequent use of CTC might be the low resolution
of a single detector helical CT and reports of an unacceptable high
number of adverse events after the injection of meglumine
iotroxate. (See A Persson, N Dahlstrom, O Smedby and T B Brismar:
"Three-dimensional drip infusion CT cholangiography in patients
with suspected obstructive biliary disease: a retrospective
analysis of feasibility and adverse reaction to contrast
material"). It has been observed that selective preoperative ERCP
was successful in showing CBD stones.
[0163] Fluorescence is the optical phenomenon of luminescence in
which the molecular absorption of a photon triggers the emission of
another photon of longer wavelength. The energy difference between
the absorbed and the emitted wavelength ends up as molecular
vibration or heat. A material that exhibits fluorescence is called
a fluorophore. Different fluorophores have different absorption and
emission wavelengths. Fluorescence imaging has found a number of
applications in the field of biochemistry and medicine. Typically a
fluorophore molecule, in the ground state S.sub.0, absorbs energy
h.upsilon. provided by the excitation light. This energy takes the
molecule to an excited state S.sub.1. This is an unstable state;
hence, the molecule returns to the ground state S.sub.0 by emitting
energy equivalent to h.upsilon.. The return to the ground state can
happen through many paths.
[0164] In the current research Indocyanine Green (ICG) and
bilirubin are used as fluorophores. ICG has already been
established as an injectible fluorophore for retinal angiography.
Bilirubin on the other hand exists as a component of bile, thus,
avoiding injection of an additional fluorophore.
[0165] One of the major hurdles during a cholecystectomy procedure,
open or laparoscopic, is the visualization of the bile duct, which
is hidden underneath fat layers, to avoid any bile duct injuries.
As discussed above, the known imaging techniques are not fully
satisfactory.
[0166] As described supra, the critical issue while performing a
Cholecystectomy is injury to the Common Bile Duct (CBD) which can
cause post procedural complications. Since the CBD is under a fat
layer it is important to image and locate the CBD during the
procedure. The present invention fluorescence imager helps
visualize the CBD by using the fluorescence properties of
Indocyanine Green, which is injected into the body, or by using the
fluorescence properties of bilirubin, already existing in bile. An
instrument using ICG as a fluorophore has been characterized for
its penetration depth to view fluorescence using an intralipid
model, thereby determining the best possible fluorescence
concentration that would give maximum fluorescence photons at the
same time remaining within the limits of the concentration
prescribed for human dosage. The system incorporates unique short
pass-long pass filter combinations at the source and detector ends,
respectively, to provide excitation illumination and detect
emission. The filters are designed in accordance with the
absorption and emission characteristics of individual fluorophores
in different mediums.
[0167] The best fluorescence concentration was found to be 0.015
mg/ml. Depth analysis was performed for ICG mixed with water, going
deeper in a 1% intralipid solution used as a model to mimic tissue
and fat properties. A vernier height gauge was coupled to a
capillary holder, which held the capillary containing the
fluorophore. Two separate analyses were carried out, ICG mixed with
human bile and aqueous ICG solution, both having an approximate
concentration of 0.015 mg/ml. Contrast to background and signal to
noise ratios were computed at each depth to find the maximum depth
the system can visualize. The maximum depth of penetration was
found to be 11 mm below the surface of the intralipid solution when
ICG was mixed with bile and 20 mm below the surface of the
intralipid solution for aqueous ICG. The threshold for contrast to
background was set based on beef fat measurement.
[0168] Thus, the present invention includes devices and methods for
fluorescence imaging that allows the user to visualize the bile
duct either by injection of a fluorophore into the bile duct or by
exciting the fluorophore inherently present in the bile duct. As
described above, this can be accomplished by the use of
fluorophores either existing inside the bile duct (bilirubin) or
through external injection into the bile duct (Indocyanine Green).
A unique design of short pass-long pass filters enables viewing all
emitted photons due to fluorescence. The excitation fluorescence
filter is coupled to a broadband QTH source, which provides the
energy for excitation of the fluorophore. The fluorophore absorbs
this energy and emits photons of a higher wavelength than the one
being absorbed. These photons are detected by the FPA which is
coupled to the emission fluorescence filter, designed uniquely to
prevent the transmission of any excitation light. The FPA produces
an image of the fluorescence of the fluorophore, which is then used
for further analysis and to determine the location of the bile
duct.
[0169] The present invention includes a novel fluorescence imagery
system for visualizing the biliary tract, in-vivo, using
Indocyanine Green and Bilirubin as fluorophores. The present
invention imaging system includes a focal plane array (FPA)
sensitive in the near infrared region for imaging Indocyanine green
(ICG). The present invention imaging system helps surgeons
visualize the bile duct real-time by exploiting the fluorescence
properties of ICG.
[0170] The system includes of a broadband illumination source of
light from a 250 W Quartz Tungsten Halogen (QTH) lamp placed in a
housing (Oriel, Stratford, Conn.). The source is powered by a
radiometric power supply to maintain a stable lamp output with
minimum light ripple, making it an excellent long term, stable
illuminator. The condensing lens assembly includes a molded
Pyrex.RTM. aspheric which focuses light through the filters and to
the optical couplers. A liquid light guide (Oriel, Stratford,
Conn.), which also acts as an UV filter, i.e., attenuates
wavelengths below 420 nm, is coupled to the condenser. The other
end of liquid light guide is coupled to a Pyrex.RTM. aspheric
expander which converts a fiber optic light beam to a collimated
light beam. The broadband illumination is attenuated by a low pass
filter mounted in front of the aspheric expander. The low pass
filter (Omega Optical. Brattleboro, Vt.) was selected taking the
ICG absorption characteristics into account. The lowpass filter
(LPF) has a cut off (50% Transmittance) wavelength at 795 nm. The
detector includes a high pass filter (HPF) coupled to a 50 mm,
f/1.4 lens (Nikon, Tokyo, Japan) which is placed in front of the
focal plane array.
[0171] The foregoing embodiment is represented in FIG. 1. Lamp 100
transmits light through collector optics 102 and into light guide
104. The light is then transmitted from light guide 104 through
beam shaping optics 106 and filter 108. After exiting filter 108,
the light reflects off of surface 110 and the reflected light is
subsequently passes through filter 112 and lens 114 which in turn
transmits the light to camera 116.
[0172] The high pass filter (Omega Optical, Brattleboro, Vt.) was
selected according to the emission characteristics of ICG. The HPF
has a cut off (50% Transmittance) wavelength at 805 nm. The CCD
includes a Focal Plane array. The system utilizes a PIXIS 400 BR
(Princeton Instruments, Trenton, N.J.). The PIXIS 400 BR FPA is a
fully integrated system with a permanent vacuum and deep cooling
arrangement. It uses a high performance, back illuminated,
spectroscopic format CCD. The CCD incorporates deep depletion
technology to extend the sensitivity in the near infrared (NIR).
These special devices are thermoelectrically cooled (air) down to
-75.degree. C. to provide the lowest dark charge and thereby
reduction of dark current noise. The FPA has a 1340.times.400 pixel
array having an 8 mm chip height and 27 mm spectral coverage. This
arrangement makes the FPA ideal for multistripe spectroscopy and
maximum light collecting area. Each pixel has a size of 20
.mu.m.times.20 .mu.m with a total imaging area of 26.8 mm.times.8.0
mm. The device's sensitivity ranges from 220 nm to 1100 nm, with a
peak efficiency of about 85% at .about.800 nm. This device has a 16
bit Analog to Digital Converter (ADC) coupled to the FPA which
provides a maximum of 2.sup.16 (655356) shades of gray. A high end
laptop (e.g., Dell Latitude D610. Austin, Tex.) is connected to the
camera for image rendering and analysis. The foregoing example
system is helpful for a surgeon to visualize the common bile dust
during cholecystectomy; however, this system is presented only as
an example and does not limit the scope of the claimed
invention.
[0173] The fluorescence imaging system using bilirubin as a
fluorophore includes a Focal Plane Array which is sensitive in the
visible region. The system also includes a broadband illumination
of light from a 250 W Quartz Tungsten Halogen (QTH) lamp placed in
a housing (Oriel, Stratford, Conn.). The source is powered by a
radiometric power supply to maintain a stable lamp output with
minimum light ripple, making it an excellent long term, stable
illuminator. The condensing lens assembly contains a molded
Pyrex.RTM. aspheric which focuses light toward the filters and
optical couplers. A liquid light guide (Oriel. Stratford, Conn.),
which also acts as an UV filter, i.e., attenuates wavelengths below
420 nm, is coupled to the condenser. The other end of the liquid
light guide is connected to a beam expander/collimator. A low pass
filter (Omega Optical, Brattleboro, Vt.) with a cut off (50%
Transmission) at 500 nm, selected on the basis of the absorption
properties of bilirubin, is coupled to the collimating assembly. A
high pass filter (Omega Optical, Brattleboro. VT) with a cut off
(50% Transmission) at 515 nm, selected on the basis of the emission
properties of bilirubin, is coupled to a 50 mm, f/1.4 lens (Nikon,
Tokyo, Japan) which is placed in front of the focal plane
array.
[0174] The present invention fluorescence system includes CoolSnaps
(Photometrics, Tucson, Ariz.) which contains a Sony ICX 285 focal
plane array. Higher sensitivity is one of the distinct advantages
of this CCD which enables it to have a reduced time for data
collection. The analog to digital converter (ADC) is 12 bits with a
speed of 20 Mpixels/s, thereby decreasing the acquisition time by a
factor of 20 when digitizing the same number of pixels.
[0175] Again, the foregoing embodiment is represented in FIG. 1.
Lamp 100 transmits light through collector optics 102 and into
light guide 104. The light is then transmitted from light guide 104
through beam shaping optics 106 and filter 108. After exiting
filter 108, the light reflects off of surface 110 and the reflected
light is subsequently passes through filter 112 and lens 114 which
in turn transmits the light to camera 116.
[0176] The specific system components described infra are for
illustrative purposes only and are not intended to limit the scope
of the claimed invention. The QTH source (Lamp and Housing) along
with the radiometric power supply are manufactured by
Spectra-Physics, a division of Newport Corporation. The research
Grade lamp housing (Model: 66884) houses a 250 W QTH lamp source
(Model: 6334) and the Radiometric power supply (Model: 69931)
drives the source.
[0177] The Oriel Radiometric Power Supply is a highly regulated
source of constant current or constant power for QTH lamps. The
control and monitoring features of the power supply include
powering the supply on and off, set the current/power preset and
limit, monitor the current, voltage, power and operating time. The
power supply is run in current mode with a current setting at 10.42
Amps, which drives the QTH lamp. (See `The Newport Resource`
2008-2009. Newport Corporation pp 140-141). The Research grade lamp
housing holds the QTH lamp. The housing includes condensing optics
to produce a collimated or focused beam. Asphere condensers are
used for superior uniformity. The housing also incorporates rear
reflectors to collect the lamp's back radiation, external lamp and
reflector adjustments to fine position the filament and a power
regulated fan to cool the lamp and the housing. (See `The Newport
Resource` 2008-2009. Newport Corporation pp 137-138). Re-imaging
onto the filament does increase the collimated output a little and
changes the power balance of the system.
TABLE-US-00001 TABLE 1 Specification of Radiometric Power Supply
(Model 69931) Parameters Model 69931 Power Factor >0.99 Input
Voltage 90-264 VAC Input Current 5 A Input Frequency 47-63 Hz
Output Power 40-300 W Output Current 3-24 A Output Voltage Range
0-45 VDC Line Regulation 0.01% Output Voltage Ripple <0.05%
r.m.s. Light Ripple <0.05% r.m.s. Meter Accuracy (% of full
<0.05% scale) Digital Meter Resolution, 0.1 VDC Voltage Digital
Meter Resolution, 0.01 A Current Digital Meter Resolution, 1 W
Power Safety Interlock Voltage 12 VDC/GND Operating Mode Constant
current or constant power Ambient Operating 0-45.degree. C.
Temperature Weight 20 (9) Dimensions (W .times. D .times. H) [in.
12.0 .times. 16.0 .times. 5.18 (mm)] (305 .times. 406 .times.
132)
[0178] Quartz Tungsten Halogen lamps are popular visible and near
infrared sources because of their smooth spectral curve and stable
output. ICG has a molecular formula of
C.sub.43H.sub.47N.sub.2NaO.sub.6S.sub.2. The filter designs have to
be selected based on the emission and absorption spectrum of ICG.
ICG has a bimodal absorption spectrum from 650 nm-800 nm, with
peaks occurring at 685 nm and 775 nm in distilled water. In plasma,
ICG has two absorption peaks at around 710 nm and 805 nm. The
absorption spectrum changes according to the concentration of ICG
and solvent. For example, higher concentrations have a maximum
absorption at 685 nm, while lower concentrations have a maximum
absorption at 775 nm in an aqueous solution. (See M. L. J.
Landsman, G. Kwant. G. A. Mook, W. G. Zijlstra, "Light-Absorbing
Properties, Stability and Spectral Stabilization of Indocyanine
Green", Jour. Applied Physiology, Vol. 40. No. 4. April 1978).
Thus, separate filters were designed for ICG with water as a
solvent and ICG with blood as a solvent. Instead of a conventional
band pass filter combination for emission and excitation filters,
unique low pass and high pass filters are used as excitation and
emission filters.
[0179] The filter combination (excitation and emission) of low pass
and high pass filters are used. The low pass filter has a cut off
(50% Transmission) wavelength at 790 nm. This filter coupled with
the source (mentioned above) is used to excite ICG with all
possible absorption wavelengths of ICG including the peak
absorption wavelength. The high pass filter (coupled with the
detector) was used to collect the fluorescence photons coming from
ICG. In an aqueous solution ICG, has an emission peak at 820 nm.
(See R. C. Benson, H. A. Kues, "Fluorescence properties of
Indocyanine Green as Related to Angiography", Vol. 23, No. 1,
159-163, Phys. Med. Biol. 1978). Thus, a high pass emission filter
with a cut-off (50% Transmission) at 805 nm is used. This enables
us to avoid any excitation light reflected back. These curves were
obtained by using a standard calibrated USB2000+ spectrometer
(Ocean Optics, Dunedin, Fla.) and using the filter in the
transmission mode and illuminating with the broad band source
discussed above. During the entire study all the parameters
including distance, exposure time and other pre-processing remained
the same.
[0180] ICG in blood follows a similar pattern as that for ICG in an
aqueous solution with changes in the Excitation and Emission
cut-off filters. The low pass excitation filter has a wavelength
cut off (50% Transmission) at 810 nm. This filter coupled with the
source enables illuminating the target (ICG) with its entire
absorption wavelength. ICG emission occurs at 830 nm when in blood.
(See R. C. Benson, H. A. Kues, "Fluorescence properties of
Indocyanine Green as Related to Angiography", Vol. 23, No. 1,
159-163, Phys. Med. Biol. 1978). Thus, the high pass emission
filter has a filter cut-off (50% Transmission) at 815 nm allowing
the viewing of only the fluorescence photons coming into the
detector.
[0181] Bilirubin stands as one of the important constituents of
bile. The pH of bile ranges from 7.5-9.5, which indicates that bile
is alkaline. This is due to the presence of bicarbonates. Bilirubin
is a major product of heme catabolism. Bilirubin is a yellow
tetrapyrrole pigment which is water soluble as it possesses two
propionic side chains, which might be expected to render it highly
polar. Bilirubin may be water insoluble as the bilirubin molecule
can adapt to various configurations. Bilirubin is mostly found in
the conjugated form in bile. Unconjugated bilirubin is normally 5%
of total bilirubin. Bilirubins are also susceptible to oxidation
and are photosensitive thereby leading to a variety of derivatives.
(See Francesco Baldini, Paolo Bechi, Fabio Cianchi. Alida Falai,
Claudia Fiorillo, Paolo Nassi "Analysis of Optical Properties of
Bile" J. Biomedical Optics 5(3). 321-329, July 2000). The quantum
yield of free bilirubin at room temperature is low (<10'). (See
M. A. Rosci, "Fluorescence of free bilirubin at room temperature",
Experimentia Vol: 39 (1983)). Thus, fluorescence of free bilirubin
is difficult to observe. However fluorescence of bilirubin is
enhanced in the presence of albumin. (See Humra Athar, Nisar Ahmad,
Saad Tayyab, Mohammad A. Qasim "Use of Fluorescence enhancement
technique to study bilirubin-albumin interaction" Int. J.
Biological Macromolecules, 25, 353-358. 1999).
[0182] Bilirubin in bile fluoresces as a result of its interactions
with other bile components and albumin. Thus, imaging this
fluorescence is an indicator for the position of bile duct, which
contains bile in it. Bilirubin has a strong absorption of visible
light between 400 nm-500 nm. (See Francesco Baldini. Paolo Bechi.
Fabio Cianchi, Alida Falai, Claudia Fiorillo, Paolo Nassi "Analysis
of Optical Properties of Bile" J. Biomedical Optics 5(3), 321-329,
July 2000). Thus, a filter designed to allow ideally all the
wavelengths below 500 nm would become an excellent excitation
filter for fluorescence of bilirubin in bile. This was the basis of
the excitation filter selection in the present example. The
transmission characteristics of such a filter are shown in FIG. 2.
The fluorescence of bilirubin in bile is observed to peak around
528 nm. (See Humra Athar, Nisar Ahmad, Saad Tayyab, Mohammad A.
Qasim "Use of Fluorescence enhancement technique to study
bilirubin-albumin interaction" Int. J. Biological Macromolecules,
25, 353-358. 1999). Thus, a filter capable of eliminating the
excitation wavelengths and able to measure or see the fluorescence
was selected. Such a filter was implemented through a long pass
filter with its cut off (50%) at 515 nm. The transmission of such a
filter is shown in FIG. 3.
[0183] The emission filter is coupled to a 50 mm, f/1.4 Nikon Lens
manufactured by Nikon. This lens helps in focusing the target onto
the CCD. This lens is fast enough for shooting virtually all types
of light. It produces distortion free images with high resolution
and color rendition. This lens incorporates a wide variety of
f-stops or apertures. The maximum aperture is f/1.4, while the
minimum aperture is f/1.6. The back end of the lens is coupled to a
C-F mount, which is used to fit a variety of lenses onto the
CCD.
[0184] The Charged Coupled Device or Focal Plane Array is an analog
shift register that enables analog signals to be transported
through successive stages controlled by a clock signal. An image is
projected by a lens onto the photoactive region of the CCD which
includes photodiodes and capacitors that accumulate charges based
on the intensity of light at that location. This image is then
transferred to the transmission region, which includes shift
registers. Once the array has been exposed to the image, a control
circuit causes each capacitor to transfer its content to its
neighbor. The last capacitor dumps the charge to a charge
amplifier, which converts the charge into a voltage and is finally
read out.
[0185] The Focal Plane Array used for imaging Indocyanine Green
(ICG) is the PIXIS 400 BR. The specifications for PIXIS 400BR is
given in Table 2 below.
TABLE-US-00002 TABLE 2 Specification for PIXIS 400 BR Specification
PIXIS 400 BR CCD Format 1340 .times. 400 imaging pixels 20 .times.
20-.mu.m pixels 100% fill factor 26.8 .times. 8.0-mm imaging area
System Read Noise 5 e-rms @ 100-kHz digitzation (max) 16 e-rms @ 2
MHz digitization (max) Spectrometric Well 300 ke- (High
Sensitivity- Typical) Capacity 1 Me- (High Capacity- Typical)
Deepest Cooling -70.degree. C. (High Sensitivity) Minimum
Temperature -75.degree. C. (High Sensitivity) Typical Dark Current
@ -75.degree. C. 0.25 e-/p/s (Typical) High Sensitivity 0.5 e-/p/s
(Maximum) Dynamic Range 16 Bits Vertical Shift Rate 30 .mu.sec per
row Operating Environment +5 to +30.degree. C. non-condensing
[0186] PIXIS 400 BR has controlling software that controls and
maintains the set temperature allowing a deviation of
.+-.0.05.degree. C. from the set temperature by controlling the
camera's cooling circuit. Dark charge is thermally induced into the
FPA over time. This statistical noise is called the dark noise.
Dark noise tends to change depending on the exposure time,
temperature and gain. Dark noise could be measured when there is no
light passing through the FPA. The longer the exposure time and
warmer the temperature, the background becomes larger and less
uniform. (See PIXIS User Manual). The camera has controller gain
software which allows setting three different gain levels. Level 1
(low) is used for high signal intensities. Level 2 (mid) is used
for mid intensity levels and Level 3 (high) is used for low
intensities.
[0187] From the specification table one can see that the camera has
a dual digitization rate available (100 KHz/2 MHz). The 2 MHz
digitization rate is used for the fastest possible data collection,
while the 100 KHz is used when noise performance is of greatest
concern. Thus, multi digitization allows complete freedom between
`Slow Operation` for low noise and high SNR and `Fast Operation`
for rapid spectral acquisition. (See PIXIS User Manual).
[0188] Megapixel resolution and smaller pixel size allows imaging
very fine details. Sensitivity can be improved by binning, but at
the expense of resolution. The camera has a 1340.times.400 pixel
CCD array that provides superior resolution over the industry
standard `1024` pixel format. Binning increases the frame rate. The
camera has a dynamic range of 16 bits allowing bright and dim
signals to be quantified in a single image with 2.sup.16 shades of
gray.
[0189] The PIXIS 400 BR has a back illuminated FPA. This means that
light enters from a back surface through a thinned (etched) silicon
layer. This has an advantage that no light absorption or reflection
takes place at the polysilicon gate structure enabling the CCD to
have higher quantum efficiency (almost twice the efficiency). Due
to this etching, the layer becomes transparent to NIR wavelengths
causing fringe effects for NIR spectroscopy. To overcome the
disadvantages of etaloning, the CCD is made of thicker silicon
(roughly twice the thickness of a normal back-illuminated CCD).
This contributes significantly to the absorption of NIR light,
reducing the amount of light that survives a round trip path to
cause interference and increasing the quantum efficiency. This
reduces the amount of light into the CCD that is reflected back
from the polysilicon side of the back surface. It also increases
the quantum efficiency by increasing the amount of light into the
CCD and reducing stray light in the spectrometer.
[0190] The CoolSnap.sub.ES camera is used as a detector for
measuring the fluorescence of bilirubin. This camera has high
quantum efficiency in the visible region which makes it a perfect
choice for fluorescence measure of bilirubin. This camera is used
in applications requiring high speed and high spatial resolution in
the visible region. It is manufactured by Roper Scientific (Now
Princeton Instruments). The CoolSnap.sub.ES offers higher
sensitivity and lower read noise to produce high quality 12-bit
monochrome images. (See CoolSnapES user manual). Exemplary
specifications are provided is Table 3 (below).
TABLE-US-00003 TABLE 3 Specifications of CoolSnap.sub.ES FPA
Specification CoolSnap.sub.ES CCD Format 1392 .times. 1040 imaging
pixels 6.45 .times. 6.45-.mu.m pixels 8.77 .times. 6.6-mm imaging
area (Optically Centered) System Read Noise <8 e-rms @ 20 MHz
Well Capacity 16000 e- (Single Pixel) 30000 e- (2 .times. 2 Binned
Pixel) Cooling Thermoelectric, 5.degree. C. below ambient
Temperature Dark Current 1 e-/p/s Dynamic Range 12 Bits @ 20 MHz
Dimensions 4.5'' .times. 5.0'' .times. 2.5'' (1.9 lbs) Operating
Environment 15.degree. C. to 30.degree. C. ambient Frame readout 91
ms/frame
[0191] The CoolSnap.sub.ES incorporates a SONY ICX285AL silicon
chip array with interline transfer capability. The interline
transfer CCD has a parallel register that is subdivided into
alternate columns of sensor and storage areas. The image
accumulates in the exposed area of the parallel register and during
CCD readout the entire image is shifted under interline mask into a
hidden shift register and then proceeds in normal CCD fashion.
Since the signal is transferred in microseconds, smearing is
undetectable for typical exposures. However, a drawback to
interline transfer CCD's has been their relatively poor sensitivity
to photons since a large portion of each pixel is covered by the
opaque mask. As a way to increase the detector's fill factor, high
quality interline transfer devices have microlenses that direct the
light from a larger area down to the photodiode. The quantum
efficiency is about 60% in the region of bilirubin-albumin
fluorescence (500 nm-550 nm), hence these CCDs are used as
detectors for measuring the fluorescence of bilirubin-albumin.
[0192] Both the above mentioned detectors are driven by
Photometrics Virtual Camera Access Method (PVCAM) software. The
PVCAM application programming interface for high performance
digital cameras is a set of software library routines that
implement a camera's operations in a hardware independent, platform
independent suite of function calls. This software is used to
control and acquire data from the camera. The data collection is
done using Vpascal programming integrated in V++ which internally
communicates with PVCAM for controlling the detectors.
[0193] Binning is a process of combining charge from adjacent
pixels in a CCD during readout. This process is performed prior to
digitization in the on-chip circuitry of the CCD by specialized
control of the serial and parallel registers. The two primary
benefits of binning are improved signal-to-noise ratio (SNR) and
the ability to increase frame rate, at the expense of spatial
resolution. Binning 1.times.1 has the maximal spatial resolution.
No charges are combined in this case. In the case of 2.times.2
binning, during parallel readout, the charges from two rows of
pixels, rather than a single row, is shifted into the serial
register. Next, charge is shifted from the serial register, two
pixels at a time, into the summing well. It then goes to the output
amplifier. This process is iterated until the entire array is read.
One of the prime advantages of binning is high SNR. In normal
operation, CCD read noise will be added to each pixel, whereas
during binning, the CCD read noise is added to each super pixel
thereby increasing the SNR. A 2.times.2 binning process is shown in
the FIG. 4.
[0194] The following characterization of indocyanine green
fluorescence imager is provided for clarification. Indocyanine
green (ICG) (C.sub.43H.sub.47N.sub.2O.sub.6S.sub.2Na), having a
molecular weight of 775, is a trycarbocyanine type of dye with
infrared absorbing properties. ICG has little or no absorption in
the visible spectrum. ICG may be used for recording dye dilution
curves, in particular for determination of cardiac output. The
principle advantages of using ICG as a dye are its absorption
maximum at the isobestic point of hemoglobin and oxy-hemoglobin,
the confinement to the vascular compartment through plasma protein
binding, low toxicity and rapid excretion into the bile. (See M. L.
J. Landsman, G. Kwant. G. A. Mook, W. G. Zijlstra, "Light-Absorbing
Properties, Stability and Spectral Stabilization of Indocyanine
Green", Jour. Applied Physiology. Vol. 40, No. 4, April 1978).
Moreover, ICG is readily soluble in water. Following intravenous
injection, ICG binds to plasma proteins with albumin as its
principle carrier. ICG undergoes no significant extra hepatic or
entero hepatic circulation. ICG is taken up by plasma almost
exclusively by the hepatic parenchymal cells and is rapidly
excreted into the bile. Due to its properties, ICG is used
extensively for the study of hepatic function. ICG is FDA approved
and can be used with proper dosages as per FDA regulations.
IC-Green.TM. is available in vials from Akorn Inc. Dosages for ICG
can also be taken from the same.
[0195] It has been demonstrated that ICG can be used as a contrast
agent in animals for visualizing biliary tract during Laparoscopic
Cholecystectomy. (See Araki, K. Namikawa, J. Mizutani, M. Doiguchi.
H. Yamamoto, Arai T. Yamaguchi, et al. "Indocyanine Green Staining
for Visualization of Biliary System during Laparoscopic
Cholecystectomy"; and, D. Persemlidis, A. Barzilai, et al.
"Enhanced Laparoscopic Visualization of Extrahepatic Bile duct with
Intravenous ICG"). The present invention uses the fluorescence
properties of ICG for viewing the biliary tract during
cholecystectomy.
[0196] The spatial resolution of the fluorescence imager was
established by computing the percent contrast by imaging a standard
USAF 1951 resolution target. The block diagram for the set up is
given in FIG. 5. The distance between the source and the target was
set as 22 inches. The results are plotted for multiple bins. The
percent contrast, C, where I.sub.max is the maximum intensity
reflected by a line of the resolution target (i.e., a white bar)
and I.sub.min is the minimum intensity from the non reflecting area
between the white bars (i.e. dark bars).
[0197] Percent contrast is calculated by using equation (1):
C = ( I max - I min I max + I min ) ( 1 ) ##EQU00001##
[0198] This percent contrast measurement is carried out with and
without the emission filter on the detector end, to observe any
effects on spatial resolution due to the filter. FIG. 6A shows the
percent contrast measurements carried out without the emission
filter on the detector end, while FIG. 6B shows the percent
contrast measurements carried out with the emission filter on the
detector end.
[0199] It was found that there is a slight reduction in percent
contrast when the emission filter is placed in the path. This can
be attributed to the way the filters are manufactured. The surface
of the filters is not always smooth and hence there can be a
reduction in the spatial resolution. With the filter in place, the
spatial resolution of the system (not for fluorescence) turns out
to be 0.4 mm when used with the lowest bin value (Bin 1).
[0200] The following describes the quantification of the absorption
spectrum of ICG. The first step in determination of ICG as a
potential fluorophore for visualizing biliary tract during
Cholecystectomy is to find its absorption spectrums and wavelengths
needed for ICG to get excited for fluorescence. ICG has a bimodal
absorption peak when in an aqueous solution (e.g., distilled
water). The absorption spectrum changes with different solutes of
ICG and different concentrations. This observation leads to the
conclusion that ICG does not obey Beer Lamberts law. There is
however a low concentration region where it does obey Beer Lamberts
law. (See M. L. J. Landsman, G. Kwant, G. A. Mook, W. G. Zijlstra,
"Light-Absorbing Properties, Stability and Spectral Stabilization
of Indocyanine Green", Jour. Applied Physiology, Vol. 40, No. 4,
April 1978). The set up for measuring the absorption spectrum of
ICG is shown in FIG. 7.
[0201] A QTH lamp, as discussed above, along with all its optical
components is coupled to a low pass filter (e.g. an excitation
filter having a cutoff of 790 nm), which acts as an excitation
filter for the fluorophore ICG. Cardio Green (Sigma Aldrich, St.
Louis. MO) was used. Three different concentrations of ICG (0.078
mg/ml, 0.015 mg/ml and 0.005 mg/ml) were prepared. An initial
solution of 0.078 mg/ml was prepared and the other concentrations
were obtained by dilution of this initial solution. Thus, any error
in the measurement of ICG is consistent. These were filled inside
capillaries having a 1.5 mm bore diameter. The detector (PIXIS 400
BR) is coupled to a NIR Liquid Crystal Tunable Filter (LCTF)
(Cambridge Research & Instrumentation. Boston, Mass.) which is
already calibrated. The distance between the source and the target
was set to 22 inches for all concentrations. It is possible to tune
the LCTF for different wavelengths and get images at each of those
wavelengths. The wavelengths and interval between wavelengths is
user defined through software. In this embodiment, the LCTF was
tuned for wavelengths from 650 nm to 900 nm with increments of 2
nm. Images were obtained at every 2 nm increment of wavelength
starting from 650 nm and ending at 900 nm. Thus, a 3-D
hyperspectral image cube is obtained with two spatial dimensions
and one spectral dimension. These cubes were obtained for each
three different concentrations and also for spectralon (i.e., a
material having 100% reflectance). Each of these concentration
cubes were then divided by the spectralon cube to obtain
absorptions of the different ICG concentrations. The resultant cube
was then filtered using a savitsky-golay filter and the spectrum
was plotted by taking a sample area from the cube that matches the
position of the capillary. ICG in distilled water has a bimodal
peak absorption spectrum with maximum absorption occurring at 705
nm and 775 nm, respectively. One important observation is the
change in the shape of the curves with changes in concentration. As
the concentration of ICG in distilled water increases, the peak at
775 nm tends to become a shoulder and only one peak at 705 nm
remains. Contrarily, as the concentration decreases the peak at 705
nm tends to become a shoulder and only one peak at 775 nm remains.
This effect can be attributed to the aggregate formation of ICG
when the concentration is higher. (See M. L. J. Landsman, G. Kwant,
G. A. Mook, W. G. Zijlstra, "Light-Absorbing Properties. Stability
and Spectral Stabilization of Indocyanine Green", Jour. Applied
Physiology, Vol. 40, No. 4, April 1978).
[0202] Determination of the fluorescence concentration that
produces the best emission was required. The set up for
determination of best fluorescence concentration is shown in FIG.
8. A QTH source is coupled to a low pass filter having a cutoff of
790 nm, which provides all excitation wavelengths for ICG.
Different concentrations of ICG in distilled water were used. An
ICG solution having a concentration of 0.03 mg/ml was first made by
carefully measuring from an Adventurer.TM. (Ohaus). This solution
was further diluted to form the other concentrations (0.02 mg/ml.
0.015 mg/ml. 0.010 mg/ml and 0.005 mg/ml). These solutions were put
in different capillaries. A NIR detector (PIXIS 400 BR) coupled to
an emission filter having a cutoff of 805 nm was used to measure
the fluorescence. The five capillaries were kept parallel to each
other and were excited by the QTH source coupled to the excitation
filter as mentioned previously. Care was taken to make sure that
the capillaries each received approximately the same intensity of
illumination by keeping the capillaries closer to each other.
Fluorescence photons emitted by ICG were then detected using the
detector mentioned above. The best concentration was decided based
on the maximum fluorescence photons detected by the detector.
Higher concentrations were eliminated as there was a reduction in
fluorescence emission and due to restrictions on the dosage.
[0203] Fluorescence for 0.03 mg/ml, 0.02 mg/ml and 0.015 mg/ml ICG
are almost identical with the peak fluorescence at 0.03 mg/ml and
0.015 mg/ml. Thus, either concentration can be used for injecting
into humans. However, dosage regulations restrict the concentration
to 0.015 mg/ml. ICG forms aggregate at higher concentrations. The
total dose of dye injected should be kept below 2 mg/kg. Thus, for
an average adult having a weight of 70 kg, the concentration of dye
with blood as a solvent is 0.025 mg/ml.
[0204] Maximum depth of fluorescence inside tissue was then
determined. Human tissue has both scattering and absorbance
properties. Thus, light falling on tissue gets partly scattered and
partly absorbed. This depends on the absorbing properties of the
tissue. Different tissues have different absorption and scattering
properties. As described above, one of the issues related to
viewing the bile duct is that the CBD lies underneath a thick fat
layer. Thus, it is important to determine the depth that the
fluorescence imager can see, to enable the user to accurately
determine the position of bile duct.
[0205] An intralipid model was then considered, which model mimics
tissue properties. A 1% intralipid solution was prepared by
diluting a 20% intralipid solution (obtained from Fresenius-Kabi)
with distilled water. This 1% intralipid solution mimics the
properties of skin tissue. (See I Driver, J W Feather, P R King, J
B Dawson "The optical properties of aqueous suspensions of
Intralipid, a fat emulsion" Phys. Med. Biol., 1989 Vol:34, No 12,
1927-1930). 0.03 mg of ICG was carefully weighed using an
Adventurer.TM. (Ohaus) and then mixed with 10 ml of distilled
water. This solution was then diluted to get a concentration of
0.015 mg/ml. This fresh ICG solution was filled into capillaries
having 1.5 mm bore diameters using a 1 cc syringe. The capillary
was then fitted to a stand, which was coupled to a vernier height
gage such that each capillary could be moved with precision in the
vertical direction. The vernier height gage enables data to be
produced at each millimeter. The fresh intralipid solution,
previously prepared, was then filled into a tub. The setup was
arranged in such a way that the center point of the bore of each
capillary was at the surface of the 1% intralipid solution. A QTH
source coupled with the low pass excitation filter having a cut off
of 795 nm was used as an excitation source for the ICG in the
capillary. The detector (PIXIS 400 BR) was coupled to a high pass
emission filter. The distance between the source target and the
detector is also illustrated in the figure. A distance of 22 inches
between the source and target was fixed because of clinical
constraints during cholecystectomy surgery. The detector was
focused on the surface of the intralipid solution. At each step,
each capillary was carefully moved into the intralipid solution in
increments of 1 mm using the vernier scale. Care was taken not to
disturb the alignment of the capillary tube with respect to the
detector. Two different studies were done using the same setup: (1)
with a constant exposure time; and, (2) with variable exposure
time. In the constant exposure method, the exposure time and
aperture of the detector was adjusted to measure maximum
fluorescence without getting saturated, when each ICG filled
capillary was at the surface of intralipid solution. The same
exposure time and aperture were used for all depths. Background
(i.e., plain intralipid solution using both filters) images were
also collected using the same exposure and aperture. In the
variable exposure time method, the exposure time was varied at each
depth to get maximum fluorescence photons out of the ICG. Thus,
each depth had a different exposure time. This difference in
exposure time was accounted for during the analysis of the images.
Background (i.e. plain intralipid solution using both filters)
images were also collected using the same aperture but different
exposure times to obtain maximum counts from the intralipid
solution. The entire study was conducted in the dark to maximally
avoid any stray light from the background into the detector.
[0206] Data acquisition was also taken from beef fat. Data was
taken with capillaries having aqueous ICG, at 2 mm from the surface
of the fat and at 2 cm from the surface of the fat. The second data
point can be used as a threshold to determine the penetration depth
in the intralipid model as shown in FIGS. 9A and 9B.
[0207] Images at different depths were collected in both methods.
Images collected using method 1 (i.e., with constant exposure),
were ratioed with the background image obtained with all parameters
being the same. However, in the case of method 2 (i.e., variable
exposure time), each image was first divided by its exposure time
to obtain an Intensity/sec at each pixel. The background image was
also divided by the corresponding exposure time and then each image
was ratioed with the processed background image. Ratioing with the
background image removes any source patterns associated with the
original images. Thus, the fluorescence signals can then be
processed. Beef fat was obtained and a capillary filled with ICG
having the optimal concentration obtained above at approximately 2
mm to 3 mm below the surface of fat. The images obtained are shown
in FIGS. 9A and 9B.
[0208] FIGS. 9A and 9B show images of beef fat with an ICG in water
filled capillary taken from two cameras. FIG. 9A shows a digital
image of beef fat with a capillary having ICG with water in it.
FIG. 9B shows a digital image taken using the Surgical Fluorescence
Imager of the same beef fat with the same capillary. The
fluorescence glow can be clearly seen.
[0209] In both methods, the processed images were then cropped for
the region of interest, and then a row profile of each row was
obtained. This profile was median filtered with a filter window of
3. The signal to noise and contrast to background at each row were
then averaged and plotted along with their error bars. FIG. 10
shows SNR and contrast to background calculations. This profile is
taken from each row of the image. SNR and contrast to background
are then calculated for each row and then averaged.
[0210] SNR (Signal to Noise Ratio) is the ratio of signal power to
the noise power. From the profile obtained above, signal to noise
was calculated by selecting the noise region, getting the standard
deviation and then dividing the signal by this noise. This is
depicted in FIG. 10, and described by equation (2):
S N R = ( Sig max - Sig min StdDev_of _Noise _Region ) ( 2 )
##EQU00002##
[0211] where Sig.sub.max is the maximum fluorescence signal;
and,
[0212] Sig.sub.min is the baseline fluorescence signal.
[0213] The plots for the SNRs using both of the above mentioned
methods are shown in FIG. 11A and FIG. 11B. SNR was calculated for
each row from the cropped, processed image. A Mean SNR was
calculated along with the standard deviation. FIG. 11A shows the
signal to noise ratio against penetration depth in the intralipid
solution of aqueous ICG with constant exposure, while FIG. 11B
shows the signal to noise ratio against penetration depth in the
intralipid solution of aqueous ICG with variable exposure. The
square data points depict aqueous ICG in beef fat at two different
depths.
[0214] Moreover, FIGS. 11A and 11B show that the signal to noise
decreases as the capillary goes deeper. As discussed supra,
intralipid has both absorption and scattering properties. When the
fluorescence signal goes deeper, scattering and absorption takes
place in the intralipid leading to lower signal and higher noise.
Another factor that could affect the signal is the excitation
source. The penetration depth of the excitation light itself is
lower due to intralipid absorption and scattering thereby leading
to lower fluorescence signal and a reduction in SNR. Typically the
threshold for SNR is taken as five (5). As shown in the figures, a
SNR of five (5) is reached at 18 mm below the surface with constant
exposure and at 21 mm below the surface with variable exposure.
[0215] Another measure of depth analysis, with the same set up was
taken for ICG with bile. Care was taken to maintain the
concentration of ICG in bile to be 0.015 mg/ml. This was carried
out with the constant exposure method. FIG. 12 is a plot depicting
the signal to noise ratio against penetration depth in intralipid
of ICG with Human Bile using the constant exposure method.
[0216] When bile is mixed with ICG, the penetration depth reduces
due to various effects of the bile, e.g., quenching. Thus, from the
plot, we can see that an SNR of five (5) is reached at 11 mm below
the surface. Beyond this depth, it is difficult to distinguish
between signal and noise, and thus data analysis was
terminated.
[0217] Contrast to Noise was calculated using the same value of
signal as for SNR and dividing it by the mean of the noise region.
This parameter can also be used as a deciding factor for finding
the penetration depth of ICG in intralipid. The plots for contrast
to background are shown in FIGS. 13A and 13B. FIG. 13A shows a plot
depicting the contrast to background ratio against penetration
depth in intralipid of aqueous ICG using the constant exposure
method, while FIG. 13B shows a plot depicting the contrast to
background ratio against penetration depth in intralipid of aqueous
ICG using the variable exposure method. Square data points show
data with aqueous ICG in beef fat at two different depths.
[0218] Similar to signal to noise ratio, contrast to background
ratio was also computed for ICG with Bile in an Intralipid
solution. FIG. 14 is a plot depicting the contrast to background
ratio against penetration depth in intralipid of ICG with Human
Bile using the constant exposure method.
[0219] From FIG. 14, it is clear that the contrast to background
ratio decreases with penetration depth. From FIGS. 11A-14, it can
be concluded that the penetration depth of ICG with water in a 1%
intralipid solution is approximately 21 mm from the surface of
intralipid. The dashed line in the plot (FIG. 13B) determines the
threshold value of contrast to background taken from aqueous ICG in
beef fat. However, since a 1% intralipid solution is a homogenous
medium and ICG in aqueous solution is also homogenous, the
penetration depth is high. Analysis of beef fat revealed a contrast
to noise threshold of 1.0.
[0220] From FIG. 14 it is clear that the penetration depth is
approximately 11 mm below the surface of intralipid beyond which
the contrast to background ratio is below 1.
[0221] As described above, free bilirubin does not have high
fluorescence and conjugated bilirubin has a better quantum yield.
Bilirubin has a strong and main absorption spectrum in the visible
region with wavelengths ranging from 390 nm to 460 nm and a molar
absorption coefficient of approximately 50.times.10.sup.3
mol.sup.-1 cm.sup.-1 in aqueous solvents. The position of maximum
absorption, the shape of the spectrum and the molar absorption
coefficient depends greatly on the conformational structure.
[0222] The fluorescence of bilirubin increases in the presence of
albumin. (See Humra Athar, Nisar Ahmad, Saad Tayyab, Mohammad A.
Qasim "Use of Fluorescence enhancement technique to study
bilirubin-albumin interaction" Int. J. Biological Macromolecules,
25, 353-358, 1999). Bilirubin dianions combine reversibly with
human albumin in neutral or alkaline solution. (See Jorgen
Jacobsen, Rolf Broderson "Albumin-Bilirubin Binding Mechanism"
Jour. Of Biological Chemistry, Vol 258, No. 10, Issue of May 25,
pp. 6319-6326, 1983). A bilirubin albumin solution has an
absorption spectrum from 500 nm-600 nm, with an emission occurring
at around 528 nm.
[0223] Measurement of fluorescence was done using the source
detector combination as shown in FIG. 15. FIG. 15 is a block
diagram depicting a set up for measurement of fluorescence from
bilirubin.
[0224] Bile was taken from a patient with the patient's consent.
The raw bile was to used for filling the capillary through a
syringe. The capillary was kept over a black cloth to avoid any
reflection as this would affect the fluorescence. A source coupled
to an excitation filter (i.e. a short pass filter having a cut off
of 500 nm) was used to excite the bile inside the capillary. The
source was a broad band QTH source as discussed above. The
source-filter combination was at a distance of 15 inches from the
capillary. The detector used was CoolSnap.sub.ES (Princeton
Instruments). The detector was coupled to the bilirubin emission
filter (i.e., a long pass filter having a cut off of 515 nm) along
with a 50 mm, f/1.4 Nikon lens. The detector and the source were
arranged in reflective mode geometry as shown in FIG. 15. The
detector was further connected to the laptop for data acquisition.
In such a mode, light emitted through fluorescence was collected at
the detector which then amplified and provided an image. FIG. 16 is
a fluorescence image of bilirubin with excitation from 400 nm-500
nm. The fluorescence is clearly seen as compared to the blank
capillary.
[0225] The above described image was obtained by subtracting the
raw image against its background (i.e., image without the bile
capillary). From the image, the fluorescence of bilirubin is
clearly seen. This can be used as one of the parameters in the
detection of biliary duct. From this experiment, it has been found
that the bilirubin in the bile fluoresces. This provides another
method of visualizing the bile duct in a cholecystectomy
procedure.
[0226] The spatial resolution of the bilirubin fluorescence imager
was established in a similar manner as it was done for determining
percent contrast using the ICG fluorescence imager. However, in
this embodiment, the detectot was the CoolSNAP.sub.ES. Excitation
filters having a cut off of 500 nm and emission filters having a
cut off of 515 nm were used. The results are shown in FIGS. 17A and
17B. FIGS. 17A and 17B are plots depicting percent contrast against
spatial resolution, when no emission filter is used (FIG. 17A), and
when an emission filter is used (FIG. 17B). FIGS. 17A and 17B show
that there is a slight decrease in the percent contrast when the
emission filter is used in the optical path. This is possible as,
there can be non-uniformity generated due to the filter from
certain manufacturing issues. Additionally, the spatial resolution
with the emission filter is approximately 0.25 mm (at a percent
contrast of 26%). This does not reflect the spatial resolution when
measuring fluorescence and just reflects the spatial resolution of
the CCD coupled to the filter.
[0227] It has been demonstrated that Indocyanine Green and
Bilirubin can be used as a latent contrast agent for visualizing an
anteriorly placed biliary structure. This is evident from the
analysis of penetration depth in an intralipid model. However. ICG
mixed with bile resembles closer to real data imaging. Thus, it has
been shown that the penetration depth of Indocyanine Green is
approximately 11 mm below the surface of a 1% intralipid. Plain
aqueous ICG was used for depth analysis in a 1% intralipid
solution, and the penetration depth is almost doubled. This
demonstrates that when bile is mixed with ICG, aggregates are
formed. Moreover, bile itself absorbs some of the incident
radiation that falls on it, thereby reducing the excitation energy
for ICG. However, this forms a more realistic model that can be
used for clinical studies.
[0228] Another potential method is to perform multimodal imaging
using a DLP hyperspectral imager that is capable of imaging
fluorescence signals from ICG, bile and NIR reflectance radiation.
Moreover, the present invention comprises a method of performing
multimodal imaging using a DLP hyperspectral imager which includes
but is not limited to visualization of biomarkers and/or
biochemical markers that absorb, reflect and/or fluoresce by using
digital signal processing and chemometric algorithms, or
alternatively, complex spectral illumination can be used to collect
complex spectral data thereby providing chemically encoded
information. Accordingly, an embodiment of the present invention
includes a method of confirming the location of the bile ducts by
video rate or near video rate imaging of the area and switching
between various methods such as the hyperspectral imaging method
using NIR spectral illumination and fluorescing ICG and bile.
[0229] FIG. 18 is yet another embodiment of the hyperspectral
imager of the present invention. Briefly, the hyperspectral imager
includes Xenon lamp assembly (Optronic Labs) 200, broadband light
202 emitted from lamp 200, passes through sub-millimeter slit
(Optronic Labs) 204 and in turn passes through collimating lens
206. Collimated light 208 then reflects off of diffraction grating
209 which creates diffracted light 210, i.e., light separated into
a plurality of bandpasses of light. It should be appreciated that
slit 204, in combination with collimating lens 206 and grating 209
may also be a prism, a tunable filters, an electromechanical
optical filter wheel, an acousto-optical tunable filter, a
liquid-crystal tunable filter, a digital micromirror device, an
electro-optical filter, a holographic filter, and combinations
thereof. The diffracted light strikes a DLP.RTM. digital
micromirror array 211, which reflects projected light 212.
Projected light 212 includes intensities of wavelengths of a
complex spectrum reflected back into the optical path by array 211.
In other words, diffracted light 210 falls on array 211 in such a
way that the columnar bands of discrete wavelengths of light each
fall on a specific column of micromirror on array 211. Then, by
selecting particular rows within the columns of micromirrors, the
spectral content of projected light 212 may be controlled forming a
complex spectral illumination. For example, as shown in FIG. 18,
rows within the columns of micromirrors corresponding to
wavelengths ranging from 380 nm to 1600 nm are "turned on" while
rows within other columns of micromirrors are "turned off". Thus,
projected light 212 includes varying light intensity ranging in
wavelength from 550 nm to 600 nm. Subsequently, projected light 212
passes through beam shaping optics 214 and strikes tissue sample
216. Reflected light 218 from tissue sample 216 is received by CCD
Focal Plane Array detector 220 and the detected data is
communicated to computer 222 for data processing. Computer 222 can
also be arranged to control digital micromirror array 210 and the
detector 220.
[0230] As described supra, the present invention includes a
spectral illuminator using DLP technology to excite fluorescence
markers (which can be native to the tissue (tissue fluorescence) or
target, or that was injected or painted onto the tissue or target)
and an optical filter, LCTF or other element used to pass emitted
fluorescence light to a detector. With the improved imaging
provided by the present invention, surgical procedures can focus on
the target for treatment (or its adjacent tissue) by visualization,
e.g., the biliary tree. The present invention system can be used
for other surgical applications to perform in vivo pathology during
surgery, or alternatively, in vivo hyperspectral pathology. The
present invention can also be used in methods, including, in vivo
pathology of cancer in real-time (e.g., before, during or after
surgery), in the body (e.g., laparoscopically) without having to
biopsy and/or fix the tissue or target to a slide. Other examples
include a probe that binds to antigens. The techniques are also
applicable to fluorescence microscopy. Finally, chemometrics and
digital signal processing may be used to enhance the visualization.
Furthermore, it has been found that using a digital light processor
the present inventors were able to achieve video rate hyperspectral
image capture and processing. Video rate capture and processing is
not possible with current illumination technology and made possible
by including DLP. The video capture rate also benefits from the
chemometric illumination.
[0231] Clinical results have shown visible reflectance
hyperspectral imaging is capable of visualizing chemical changes
using inherent chromophores within the microvasculature. These
results have created the opportunity to apply this cutting-edge
imaging modality in a variety of clinical applications. As an
example, the surgeon needs a system for helping determine the
degree of amputation and monitoring vascular healing after surgery
in order to reduce the risk of an amputation failure. Similarly,
plastic and reconstructive surgeons have a need for a reliable
system indicating tissue viability of skin flaps and transplanted
tissue, as well as, determining how much tissue to extract when
removing a cancerous skin tumor. Moreover, the laparoscopic gastric
bypass surgeon needs a system for detecting hemorrhage and ischemic
tissue during gastric bypass procedures to avoid catastrophic
complications. To that end, hyperspectral imaging is useful in
monitoring and assessing vascular healing after performing a lower
limb amputation, in determining the degree of amputation and tumor
removal, monitoring tissue viability of skin flaps and transplanted
tissue and detecting strictures and leaks during gastric bypass
surgery. There were approximately 82,000 lower-limb amputations
reported in the year of 2002; 105,000 bariatric gastric bypasses;
500,000 cholecystectomy surgeries were performed in 2003; and,
approximately 59,350 new cases of skin cancer, all of which are
expected to increase over time.
[0232] As described supra, hyperspectral imaging is an optical
imaging technique that captures the spatial and spectral
information from the source target, typically capturing hundreds,
of contiguous wavelength bands for each pixel. The hyperspectral
data would be rendered useless, until the recorded data is
precisely analyzed and processed using tools borrowed from spatial
image processing, chemometrics and spectroscopy to yield
information that can be presented in an image form. Thus, the
hyperspectral imaging approach provides a "data cube" consisting of
two dimensional images with each desired wavelength being
represented by one of the images from the stack of images
collected. The image cube thus obtained, has three dimensions: two
(2) spatial (X and Y) and one (1) spectral dimension (wavelength).
A basic hyperspectral imaging system consists of a broadband light
source, with the desired spectral band of interest; optics arranged
to focus light on the source target; an electronically tunable
filter to spectrally discriminate the imaged light from the source
target; and, a sensitive array detector or focal plane array, which
collects the light, converts it into a two dimensional gray scaled
image and transfers it to a computer. The image acquisition process
continues until all images at desired wavelengths have been
collected, generating the hyperspectral image cube. An overview of
a hyperspectral image cube is depicted in FIG. 19.
[0233] Hyperspectral imaging has been used in satellite technology
to identify features on the earth surface: however, there has been
little application to biology and almost none to clinical medicine.
The present inventors address this with the next generation of
hyperspectral imaging systems that are "source based".
[0234] The present invention incorporates DLP.RTM. technology and
increases the speed of image data acquisition. The DLP.RTM. is used
in conjunction with a source lamp, and is capable of tuning
different wavelengths of light within microseconds versus
milliseconds required by current systems in the industry. In
addition, the placement of the liquid crystal tunable filter (LCTF)
within the slit lamp beam path and the spectral deconvolution are
unique features of this invention.
[0235] The current invention source based system can collect data
60 to 100 times faster that previous known detector based systems.
With the increased speed, the present invention may be applied to
practical clinical and surgical uses. Additionally, the increased
speed reduces movement artifacts, producing real time video images,
with the capability of displaying both normal and hyperspectral
video simultaneously.
[0236] The present invention includes integrating DLP.RTM.
technology with Focal Plane array technology, collecting and
storing digital hyperspectral image data and analyzing the data
using chemometric analysis methods. Visualizing tissue chemistry
are features unique in the field of clinical hyperspectral imaging.
This technology enables clinicians to visualize the levels of
chemicals within the tissue, for example, visualizing anteriorly
placed or hidden structures during open, endoscopic and
laparoscopic surgery detecting retinal disease early, guiding laser
therapy, monitoring patients during pharmacologic therapies for
preventing blindness, helping guide and monitor wound healing and
amputations as well as skin flaps during recovery. Other
applications include research grade devices for animal research and
in quality control of pharmaceuticals and food.
[0237] Another example application of the present invention is to
visualize tissues and organs in the human body using chemometric
techniques and sophisticated signal processing methods. The system
has a wide range of clinical applications such as enabling surgeons
to see through tissues before they cut into them during
cholecystectomies, chromoendoscopy facilitating cancer detection,
plastic surgical skin flap viability and burn wound healing
evaluation, skin cancer detection, retinal blood perfusion in
diabetic retinopathy, assessing and monitoring wound healing in
lower limb amputations, monitoring anastomotic viability in gastric
bypass operation, dental evaluation, measuring oxygenation of the
kidneys and monitoring pressure sores. The present invention can
also be used as an integral component to the smart bed and smart
hospital for measurement of vital signs and monitoring a patient's
tissue oxygenation.
[0238] The present invention consists primarily of a spectral light
engine with DLP.RTM. technology providing the spectroscopic
illumination, a digital camera with a scientific grade CCD for
imaging, and software designed and developed to manage the data
acquisition and the chemometric visualization. The data acquisition
software automatically tune the DLP.RTM. technology and trigger the
camera for collecting a series of spectroscopic images formatted as
a hyperspectral image cube. Next, the spectroscopic image data are
deconvoluted using chemometric analysis methods. The resulting gray
scale or color encoded images provides the clinician with a
non-invasive visualization of the chemical state within the micro
vasculature perfusing the tissue while the patient is in the clinic
or surgery.
[0239] One of the many goals of the present invention is to
demonstrate a DLP.RTM.-based Visible Hyperspectral Imaging Systems
can be used routinely for a variety of medical applications. In
certain embodiments, a DLP spectral light engine can be used for
illuminating the tissue with different wavelengths, colors of light
and light reflected back to the detector are measured as a spectrum
at each image pixel and using different chemometric methods to
determine relative levels of inherent chromophors within the tissue
can be visualized. In addition, the chemometric methods can be
performed on the detector instead of on the source, or the
combination of both. Furthermore, the present invention can
illuminate the object to be analyzed with spectroscopic light,
tuned by DLP.RTM.. The speed of the present invention reduces the
current acquisition time from minutes to seconds making
hyperspectral imaging a practical everyday surgical and clinical
tool for imaging.
[0240] In some embodiments, the present invention may be used in
the field of ophthalmology. Here, a hyperspectral slit lamp, which
involves the integration of electro-optics and a clinical slit
lamp, enables clinicians to image psychopathological aspects of the
retina, including the oxygenation status and the level of macular
pigments. This in turn allows the clinician to assess the need for
treatments earlier than is currently possible. Additional examples
of the clinical applications of hyperspectral imaging include: (a)
the determination of areas of ischemia in diabetic retinopathy and
retinal vascular occlusive diseases. This helps the clinician to
decide if laser therapy or intraocular injections of pharmacologic
agents would be helpful to prevent or treat retinal
neovascularization or macular edema. In the case of laser therapy,
this imaging technique would also help delineate the specific areas
of the retina that require treatment; (b) the measurement of
macular pigment levels to help in the early diagnosis and
therapeutic interventions of diseases like parafoveal
telangiectasis and, most importantly, age-related macular
degeneration; and, (c) monitor patients for determining the best
pharmacological therapy for maximizing favorable patient outcomes.
Consequently, hyperspectral imaging may help prevent or decrease
the vision loss from these highly prevalent diseases.
[0241] In some embodiments, the present invention capitalizes on
different levels of chemicals, for example the relative amount of
oxyhemoglobin, deoxyhemoglobin, carboxyhemoglobin, and met
perfusing the retinal tissue. Imaging not only the vasculature
structure but the relative amount of chemicals within such
structure to help detect and diagnose eye disease early. The
present invention can directly imaging the level of the above
mentioned molecules particularly oxygenation of the retina. This
enables monitoring effectiveness of a therapy of a patient during
treatment.
[0242] This embodiment enables clinicians to visualize the levels
of chemicals (oxyhemoglobin, deoxyhemoglobin and carboxyhemoglobin
within the retina, to detect retinal disease early, to guide laser
therapy and to monitor patients during pharmacologic therapies for
preventing blindness. In this embodiment, a liquid crystal tunable
filter is placed into the beam path of a slit lamp source
illuminating the eye with different wavelengths of light and light
reflected back to the detector are measured as a spectrum at each
image pixel. This is a unique configuration designed specifically
for measuring the chemical levels within the retina.
[0243] Typically, the wavelengths to be used depend on the tissue
type due to the penetration of tissue depth. For example, in
visible spectra. HbO.sub.2. Hb and HbCO.sub.2 can be measured. In
near-infrared (NIR) spectra, HbO.sub.2, Hb, water and lipids can be
measured.
[0244] In another embodiment, the present invention can be used for
detecting, diagnosing and monitoring disease in live human patients
visiting the clinic or during open and closed (endoscopic and
laparoscopic) surgical procedures. The system can also be used for
animal research applications. A hyperspectral imaging system
integrating DLP technology to illuminate the area of interest with
wavelength bands of light ranging over a spectral range from the
visible to the near-infrared that is interfaced with a digital
focal plane array for acquisitioning and storing a series of
digital spectroscopic images (a hyperspectral data cube) that is
analyzed using chemometric methods providing chemical information
within the area of interest aiding the clinician in detecting,
diagnosing and monitoring disease. Clinicians can also monitor
patients with this technology for determining the best
pharmacological therapy and maximize favorable patient outcomes.
The present invention utilizes DLP technology with a focal plane
array and chemometric deconvolution for visualizing the biochemical
levels within a human non-invasively for clinical and surgical
applications. The present invention can also be illuminated with
spectrum of light.
[0245] In yet another embodiment, the present invention enables the
clinician to visualize anatomical structure and the chemistry
within live human patients or research animals within seconds
during a visit to the doctor or during open, laparoscopic and
endoscopic surgery. These images are useful to the clinician or
surgeon for detecting, diagnosing, and monitoring disease and the
effectiveness for identifying hidden structures during surgery and
help clinicians determine the best therapy. The present invention
provides unique capability to image relative levels of inherent
chromophors within the tissue noninvasively without contrast agents
in live humans during clinical visits or during surgery. A DLP
spectral light engine may be used for illuminating the tissue with
different wavelengths spectrum of light, and the light reflected
back to the detector are measured as spectrum at each image pixel
and using different chemometric methods to determine relative
levels of inherent chromophors within the tissue can be
visualized.
[0246] In certain embodiments, the system of the present invention
has multiple modalities that include UV light, visible light, near
infrared light, infrared light, and fluorescence excitation. In the
DLP.RTM.-based Visible Hyperspectral Imager (DVHI) embodiment, the
present invention may be coupled with standard surgical and
clinical devices, for example a laparoscope, slit lamp, or with
lens for microscopic imaging. In particular, a quick release
mechanical coupling between the DVHI and a standard laparoscope is
developed allowing the system to image both laparoscopically during
closed surgical procedures and removing the laparoscope for open
surgery imaging. The DVHI system is calibrated at NIST and to NIST
standards. For visible to 2500 nm applications, the present
invention may use TI Discovery 1100 electronics board with an
Accessory Light Processor (ALP) electronic board.
[0247] In another embodiment, the DVHI is used in vivo using an
animal model during open surgical procedures and during closed
laparoscopic procedures. Spectral light illuminating the animal
tissue are evaluated for safety by measuring the temperature of the
tissue. A variety of tissues can be imaged and identified by
measuring spectroscopy and the chemometrics, as well as, imaging
before and after inducing ischemia and hemorrhaging the tissue. The
system can be assessed by comparing hyperspectral and chemometric
images with an evaluation made by the attending veterinarian using
standard clinical methods.
[0248] Yet in another demonstration, the DVHI is translated to
human surgery and clinic establishing its efficacy for a variety of
human applications. Here, the present inventors translate Visible
DLP.RTM. Hyperspectral Imaging for imaging during live human
surgery, and monitoring the progression of clinical disease for
determining feasibility. Also, surgical and clinical hyperspectral
data is monitored to observe the progression of disease and changes
in tissue chemistry during wound healing due to trauma,
amputations, tumor removals, plastic surgery skin flaps and
visualizing ischemia and hemorrhage during gastric bypass. For
these purposes, the present inventors utilized the existing
chemometric analysis algorithms and develop new visualization
methods, chemometric and statistical analysis. In addition, the
present inventors continue collected clinical hyperspectral image
data using existing system and existing hospital protocols for
building our clinical hyperspectral image database that can be
populated over time specifically on spectral properties of
composite structures for example the gallbladder and extrahepatic
bile ducts, determined additional clinical applications, and
refined new chemometric algorithms. The end results provide better
imaging algorithms for better clinical visualizations that can be
implemented into the software of the new DLP.RTM. Visible
Hyperspectral Imaging System.
[0249] The DLP.RTM. Hyperspectral Imaging and Chemometric
deconvolution can be applied to all products that use light, for
example, the fields of clinical endoscopy, clinical chemistry,
microscopy, surgical microscopy, drug discovery, microarray
scanners and microplate readers.
[0250] In certain embodiments, the multi-modal microscopic
reflectance Hyperspectral Imaging system developed provides
enhanced diagnosis in a clinical environment. The system developed
has the capability to measure relative contributions of
oxyhemoglobin in the microvasculature in the visible as well as the
near infrared region, hence it is multi-modal. An example was
formulated to measure contributions of to oxyhemoglobin in the
microvasculature perfusing the dermal tissue of the palm in the
near infrared region. The system was also used to image the human
eye measuring the relative contributions of oxyhemoglobin perfusing
the scleral surface. The instrumentation used for this imaging
system is shown in the FIG. 20.
[0251] FIG. 20 depicts one embodiment of the above mentioned
hyperspectral imaging system of the present invention having
charged coupled device 300 (CCD), beam splitter 302, eye pieces
304, magnification knob 306, light source 308, joystick 310,
intensity control knob 312, illumination mirror 314, subject
position 316, liquid crystal tunable filter (LCTF) 318, and common
center for rotation 320. In another embodiment, the LCTF may be
replaced with a digital micromirror device, and can be located
anywhere along the optical path of the present apparatus.
[0252] In certain embodiments, the instrumentation shown in FIG. 20
may contain a slit lamp (light source 308) that illuminates the
target using an in-built 12V/30 W illumination Halogen lamp source.
The light from the source is focused onto the target which is
placed on the Y-shaped headrest. The diffused reflected light from
the target is passed back through the slit lamp microscope optics
onto a beam splitter 302 which allows a part of light into eyepiece
304 for the observer and the remaining part of light to the relay
optics which further pass light into the LCTF 318, which is an
electronically tunable filter discriminating the reflected light
into individual wavelengths passing it onto the camera lens which
then focuses it onto Charged Coupled Device or HQ2 (CCD) camera
300. It should be appreciated that the foregoing arrangement, which
includes a LCTF, may alternatively include a DLP illumination
device as described throughout this application. Thus, the LCTF can
be removed, and the DLP illumination device coupled to this system
at the location shown for light source 308. The images formed on
the CCD are then digitized and stored in a laptop computer for
further analysis. In some embodiments, the system can be applied to
collect data in the near infrared region or in the visible region.
The visible and near infrared application differ in the use of a
unique CCD camera, i.e., the CooLSNAP.sub.ES camera instead of the
PIXIS camera, and a unique LCTF, visible LCTF instead of near
infrared LCTF, or visible versus a NIR DLP.RTM. spectral
illuminator.
[0253] The slit lamp is a microscope having an illumination system
and an observation system, which are mounted on a rotated drum with
a common centre of rotation. The slit lamp is binocular; that is,
it has two eyepieces 304, giving the binocular observer a
stereoscopic (i.e., three dimensional) view of the eye. These
devices are sometimes referred to as Biomicroscope or
stereomicroscopes. The Galilean magnification changer allows
increasing the magnifying power between three stages, e.g.,
10.times., 16.times., 25.times., by simply turning a knob which is
designed with a flat area so that the operator can change the power
by feel, without looking at the markings, since the higher
magnification is achieved when the flat portion is on top, while
lowers magnification is achieved when the flat portion is at the
bottom. The actual magnification of what is seen through the slit
lamp is derived by multiplying the power of the eyepieces, which
are fixed, with the power of the objective lens. Thus, if the
eyepieces 304 are 10.times. and the objective lens is 1.6.times.,
the total magnification is 16.times.. The non-radiometric light
source of the slit lamp is a 12V/30 W halogen lamp providing
broadband white light. The light is controlled by a transformer
which provides continuous light intensity with the help of rheostat
placed next to the joystick 310. The slit width and the slit height
can be continuously varied from 1-14 mm. A Beam splitter 302 is
effectively a semi-silvered mirror which is placed in-between the
eyepiece and the microscope optics to partially transmit light to
the eyepiece and partially reflect light to the attached CCD camera
300. In another embodiment, the detector can be a focal plane
array. Typically, for macroscopic and endoscopic embodiments, the
present invention uses a non-radiometric source system. This
non-radiometric source provides significant advantages for the use
in standard clinical operations due to its safety characteristics.
However, in the embodiments of the hyperspectral imaging system
which include DLP, a radiometric power supply is used due to it
stable current as a light source.
[0254] Attaching LCTF 318 and CCD camera 300 onto beam splitter 302
of the slit lamp does not use a lot of CCD chip area, i.e., the
spot created by the reflected light from the slit lamp falling onto
the CCD after passing through the LCTF is very small. Hence there
is a need to magnify the reflected light coming out of the beam
splitter before passing onto the LCTF. Thus, a relay optic module
was designed which is composed of two optics namely an eyepiece and
a photo eyepiece placed in a machined metal cylinder. An eyepiece
lens is an optical system which is used to magnify images in
optical instruments such as telescopes, microscopes, etc. A photo
eyepiece performs the function of focusing the images onto a plane.
These two optical components are housed in a metal cylinder 18 cm
in length.
[0255] Light passes through a Liquid Crystal Tunable Filter (LCTF)
318 having solid-state construction and no-moving parts or
vibrations, which works by applying voltage to its liquid crystal
elements, to select a transmitted wavelength range, while blocking
the rest of the wavelengths. The 60 mm, f/2.8 D micro-Nikon camera
lens focuses the light filtered by LCTF 318 onto CCD camera
detector 300. Typically, the 60 mm is the dimension for a slit-lamp
in spectral detection. The 50 mm lens is used for macroscopic
imaging spectral detector, and no lens is used with the LCTF as the
spectrum illuminator. It has the ability to produce distortion-free
images with superb resolution, sharpness and contrast. The lens has
a good depth of field and a small working distance.
[0256] The charge coupled devices (CCD's) or focal plane arrays
(FPA) perform three essential functions: photons are transduced to
electrons, integrated and stored, and finally read out. They can be
roughly thought of a two-dimensional grid of individual photodiodes
(called pixels), with each photodiode connected to its own charge
storage "well." Each pixel senses the intensity of light falling on
its collection area, and stores a proportional amount of charge in
its associated "well." Once charge accumulates for the specified
exposure time, the time between start acquisition and stop
acquisition (also known as integration time), the pixels are read
out serially.
[0257] As described in part above, the CoolSNAP.sub.ES monochrome
camera typical incorporates a SONY ICX-285 silicon chip with
Interline-transfer capability. The interline-transfer CCD has a
parallel register subdivided into alternate columns of sensor and
storage areas. The image accumulates in the exposed area of the
parallel register and during CCD readout the entire image is
shifted under the interline mask into a hidden shift register.
Readout then proceeds in normal CCD fashion. Since the signal is
transferred in microseconds, smearing is undetectable for typical
exposures. However, a drawback to interline-transfer CCDs has been
their relatively poor sensitivity to photons since a large portion
of each pixel is covered by the opaque mask. As a way to increase a
detector's fill factor, high-quality interline-transfer devices
have microlenses that direct the light from a larger area down to
the photodiode. Blooming is the migration of electronic charge to
the adjacent pixels; however, the incorporated Sony chip by Roper
Scientific provides protection against blooming by having built in
drains that remove any excessive charge generated from an
overexposed pixel. This Sony interline chip provides anti-blooming
for optical signals greater than 1000 times the full well capacity.
The spectral response for the CoolSNAP.sub.ES camera within the
visible region is of interest for the visible system application.
The quantum efficiency is always beyond 45% in this region, peaking
to over 60% between wavelength regions of 475 nm to 625 nm.
[0258] The CCD cameras, PIXIS 400 BR (or 1024 BR). CoolSNAP.sub.ES
or HQ2 are driven by PVCAM software which stands for Photometrics
Virtual Camera Access Method. HQ2 has a 20 MHz analogue to digital
digitizer which is faster. This software is used to control the
cameras and acquire data from them. The data collection process is
automated using Vpascal programming language integrated in V++, a
precision digital image processing and enhancement software. V++
has been designed to control any PVCAM-compatible feature using the
V++ interface and the VPascal built-in programming language. The
HQ2 system typically has a 20 MHx A/D digitizer that is faster than
the others. Another computer codec has been written to configure a
portable computer to provide mobile data collection. The software
further controls and synchronizes the speed of illumination, data
collection, analysis and/or deconvolution.
[0259] In an embodiment, the software performs the following steps
for analysis and deconvolution. First, ratio data for each pixel is
calculated.
RD ij = Log 10 ( BKG ij - DF ij SD ij - DF ij ) ( 3 )
##EQU00003##
[0260] where
[0261] RD.sub.ij is the ratio data for each pixel i at wavelength
j
[0262] BKG.sub.ij is the reflectance of a 100% reflectance
standard
[0263] DF.sub.ij is a dark field (the value being read when no
light is coming into the camera)
[0264] SD.sub.ij is the reflectance from the sample
The ratioed spectra are then filtered using a Savitsky-Golay
filter, which filter is well known in the art and therefore is not
discussed in detail herein. Next, the spectrum at each pixel is
normalized.
ND ij = ( RDF ij - min ( RDF ij ) max ( RDF ij ) - min ( RDF ij ) )
( 4 ) ##EQU00004##
where ND.sub.ij is the normalized spectrum at each pixel
[0265] RDF.sub.ij is the ratioed data
The effect of such normalization is that each spectrum will then
range from 0 to 1. Then, the data is further modified by a
multivariate least squares deconvolution. It should be appreciated
that such a deconvolution requires two or more reference spectra in
order to be performed. In short, a least squares fit is performed,
at each pixel, of the measured spectrum by performing linear
combinations of the reference spectra to obtain a best fit curve.
Then, the resulting relative linear contributions are scaled to
produce a gray scale image. The following equations are used to
performed the foregoing multivariate least squares
deconvolution.
S.sub.i.lamda.=(A.sub..lamda.i.sup.TA.sub..lamda.i).sup.-1A.sub..lamda.i-
.sup.T (5)
[0266] where S is the pixel sensitivity matrix
[0267] A represents the matrix of i pure component spectra composed
of .lamda. wavelengths
Then, substitution of the sensitivity matrix, S, into the
multivariate regression determines C, comprising linear
contributions of the pure components in the unknown sample,
namely:
C.sub.ij=S.sub.i.lamda.R.sub.j.lamda.+e.sub.ij (6)
[0268] where R is the matrix of the experimentally determined
sample spectra j
[0269] e.sub.ij represents the residuals.
[0270] In another embodiment, the software performs the following
steps for determining the illumination spectra and subsequent
analysis. First, at least two reference spectra, as measured at the
detector, are obtained for targets of interest, e.g., 100%
oxyhemoglobin and 100% deoxyhemoglobin. It should be appreciated
that the targets of interest may include but are not limited to
reference spectra, a patient, a portion of a patient, a purified
substance, etc. Additionally, a reference spectrum, as measured at
the detector, is obtained for 100% illumination from the
illumination source. In other words, a measurement of the spectral
content of the illumination source is determined. Then, each
spectrum is ratioed according to the following equations:
ratioed_spectrum 1 = ( spectrum 1 - spectrum 2 spectrum 100 ) ( 7 )
ratioed_spectrum 2 = ( spectrum 2 - spectrum 1 spectrum 100 ) ( 8 )
##EQU00005##
[0271] where
[0272] ratioed_spectrum.sub.1 is the ratioed spectrum from the
first target of interest
[0273] ratioed_spectrum.sub.2 is the ratioed spectrum from the
second target of interest
[0274] spectrum.sub.1 is the measured spectrum of the first target
of interest
[0275] spectrum.sub.2 is the measured spectrum of the first target
of interest
[0276] spectrum.sub.100 is the measured spectrum of the
illumination source
The ratioed spectra are then used to illuminate an unknown sample,
e.g., a kidney. In other words, the spectra calculated by equations
(7) and (8) are used to illuminate the unknown sample and the
reflected illumination are measured with the detector. Then, pixel
by pixel, the second image, i.e., the image obtained from the
illumination of ratioed_spectrum.sub.2, is subtracted from the
first image, i.e., the image obtained from the illumination of
ratioed_spectrum.sub.1. The resulting image is a chemically encoded
image providing a quantitative assessment of the targets of
interest. It should be appreciated that each of the foregoing
measurements occur on a pixel by pixel basis and therefore
differences in illumination and/or pixel response can be accounted
for.
[0277] As described above, binning is the practice of merging
charge from the adjacent pixels in a CCD prior to digitization in
the on-chip circuitry of the CCD by specific control of the serial
and parallel registers. Binning reduces the readout time and the
burden on computer memory, increases the signal to noise ratio, but
at the expense of image resolution. To comprehend the binning
practice in the Roper scientific cameras, consider the examples
shown below in FIG. 21 (binning 1.times.1 where no charges are
summed provides the maximal resolution) against FIG. 4 (binning
2.times.2 where charges from 4 neighboring pixels are summed). FIG.
21(1) shows the CCD at the end of an exposure, wherein the capital
letters represent different charge accumulated on the CCD pixels.
Readout of the CCD begins with the parallel readout phase. FIG.
21(2) shows simultaneous shifting of all pixels in a bottom row
towards the serial register followed by the serial readout phase,
FIG. 21(3) and FIG. 21(4) shown shifting of charge in the serial
register into the summing well which is then digitized. Only after
all the pixels in the bottom row are digitized is the second row
from the bottom moved into the serial register. Thus, for example
above, the order of shifting is therefore A1, B1, C1, D1, A2, B2,
C2, D2, A3, . . . D6.
[0278] The charge that has integrated during the exposure is shown
as capital letters in FIG. 22(1). Readout begins with a parallel
readout, as shown in FIG. 22(2); however, since binning of
2.times.2 is required, charge from two rows of pixel, rather then a
single row during 1.times.1 binning, is shifted into the serial
register. Next, charge is shifted from the serial register, as
shown in FIG. 22(3) and FIG. 22(4), two pixels at a time, into the
summing well rather then a single pixel as in binning 1.times.1.
The result is that each readout event from the summing well
contains the collected charge from four pixels on the CCD, i.e., a
superpixel. This procedure is iterated until the entire array has
been read and the formation of superpixels shown below for binning
of 2.times.2. FIG. 23 shows similarly superpixel formations for
2.times.2 binning and 4.times.4 binning.
[0279] In certain embodiments, hyperspectral data analysis provides
spatially distributed contributions of oxyhemoglobin, which are
obtained based upon oxy- and deoxyhemoglobin reference spectrums.
FIGS. 24A and 24B show a comparison of actual measured hemoglobin
versus the predicted hemoglobin using pure hemoglobin samples and
the actual measured hemoglobin using the present invention. The
system requires two or more reference spectra, followed by a least
squares fit, at each pixel, of the measured spectrum by performing
linear combinations of the reference spectra to obtain best fit
curve. The resulting relative linear contributions are scaled to
produce a gray scale image.
[0280] Pure HbO.sub.2 and Hb solutions were prepared at NIH by
standard methods using blood collected from a healthy individual
and reference spectra were obtained from the original imaging
system developed at NIH for the Visible region (400 nm-700 nm). A
region of interest (520 nm-645 nm) containing the peaks for oxy-
and deoxyhemoglobin was selected and used for imaging purposes in
the visible region, see FIG. 24A. Similar methods are used for
HbO.sub.2, Hb, H.sub.2O, and lipids by scanning the NIR regions.
References can be taken for HbO.sub.2, Hb, HbCO and HbNO in the
visible regions. In addition, an unsupervised method can also be
used, e.g., data can be pre-processed using a filter, and then
normalized thereafter. Here, each individual spectrum can be
normalized first. The typical steps include, but are not limited
to, measuring a spectrum, filtering the spectrum, normalizing the
spectrum data, deconvoluting the spectrum data, and calibrating the
corresponding data. The present invention does not concentrate or
only measure the spectral amplitude; the present invention observes
the changes in spectral broadening or narrowing in addition to
amplitude change.
[0281] In some embodiments, the protocol used may involve the
following: An example involving seven subjects, for imaging the
Human eye after seeking approval from the Institutional Review
Board, IRB. In this example, the sclera of the eye was illuminated
using the light source from the slit lamp for five (5) seconds
while acquiring the Hyperspectral image cube. The imaging was
achieved in two steps, firstly on entering the lab, the subjects
were asked to rest their chin on the chin rest of the slit lamp
with closed eyes and light from the slit lamp was shone onto one of
the closed eyelid which was brought in focus of the slit lamp.
Secondly, upon getting the eyelid in focus using the slit lamp
optics, the subjects were then asked to open their eyes, look in
the opposite direction from where the light was shone into the eye,
and the slit lamp was made to focus on the scleral vessels upon
which, hyperspectral image data cube was obtained for five (5)
seconds.
[0282] The setup for an embodiment of the present invention
noninvasive non-DLP* microscopic reflectance hyperspectral imaging
system consisting of a source, optics, filter and detector as shown
in FIG. 25. The slit lamp contains a 12V/30 W halogen source
providing broad band illumination, that illuminates the target and
the reflected light from the target is guided by the microscopic
optics onto a beam splitter, a semi-silvered mirror transmitting
light to the eyepiece and to the attached hyperspectral imaging
system. The radiation from the beam splitter is directed towards
the LCTF using relay optics. The LCTF collects reflected light
provided by the source into individual band passes of wavelengths
that are placed on the focal plane array (FPA) detector with the
use of a 60 mm Nikon lens fitted in front of the FPA. The FPA has
an analog to digital, A/D, converter for digitizing the data, which
is transferred to a high end laptop PC for post processing. A
computer program automatously manages the data collection by
synchronizing the timing between individual hardware components,
for example, tuning the LCTF and triggering the FPA, and setting
parameters (e.g., image size, exposure time, spectral range and
resolution, image binning, gain, filename) using a GUI that was
built in V++. To sum up the data acquisition, various parameters
are inputted initially using the GUI developed in V++, after which
the LCTF initializes and the desired voltages calculated for the
selected spectral range are stored in the palette. Upon
initialization, the LCTF tunes to the first wavelength specified in
the palette, the camera is triggered, the CCD is exposed for the
duration of the exposure time, and ADC digitizes the image
information storing it onto the laptop. This process is repeated
for the remaining wavelengths. For the visible application, the
focal plane array (FPA) used was the HQ2 CCD in combination with
the visible low resolution LCTF. The raw hyperspectral image data
was collected over a spectral range of 520 nm-602 nm with a
spectral resolution of 2 nm increments with the magnification set
at 10.times. and CCD binning of 2.times.2 for increasing readout
speed. Thus an entire reflectance hyperspectral image cube was
acquired and saved to hard disk in around five (5) seconds.
[0283] The microscopic hyperspectral imaging system was then
characterized. The characterization of the system involved the
characterizing the LCTF, the FPA and the slit lamp source. The LCTF
was characterized for its spectral band-pass, tune delay, and the
tuning wavelength capability of the filter. The FPA was
characterized for its ability to differentiate between objects,
i.e. spatial resolution.
[0284] The spectral capabilities of the near infrared LCTF were
determined using a calibrated Perkin Elmer Spectrometer. The
spectrometer was scanned in increments of 1 nm with the near
infrared LCTF that was tuned to a specific wavelength placed in the
collimated optical path of the spectrometer. This procedure was
repeated for a series of sequential wavelengths spanning the range
of the near infrared LCTF, 650 nm to 1000 nm, at every 50 nm and
the corresponding transmittance spectra of the LCTF were noted. The
transmission spectra from the spectrometer were then analyzed in
Matlab to measure the center wavelength the LCTF actually tuned
to.
[0285] The desired wavelengths that were electronically sent to the
LCTF controller are plotted on the X axis and the measured
wavelengths, the spectrometer transmission spectral analysis to
which the LCTF tuned, are plotted on the Y axis. Example data shows
a linear regression curve fit (Y=1.00.times.-0.20, R2=1). From the
example data, it is believed that an error exists between the
desired tuning and the actual tuning of the LCTF. To rectify this
small error, a look-up table in the V++ software was established
from the relationship that was derived above which makes sure that
the LCTF gets tuned to the wavelength which the operator
desires.
[0286] The visible low resolution LCTF similarly was calibrated,
however, a high resolution LCTF was used having a band-pass lying
between 0.19 nm at 500 nm and 0.75 nm at 700 nm. As shown in FIG.
26, the input light from the source is passed through the low
resolution LCTF, which is to be calibrated, tuned to a particular
wavelength onto the high resolution LCTF which is scanned with an
increment of 1 nm over its spectral range of 480 nm-720 nm, and
focused using the 60 mm Nikon camera lens onto the CCD camera
generating a series of images, i.e. a hyperspectral image cube.
This procedure was repeated until the low resolution LCTF was tuned
in the wavelength range 500 nm-700 nm at every 10 nm and a
hyperspectral cube captured. The hyperspectral cubes generated were
then analyzed in Matlab to locate the center wavelength the LCTF
was actually tuned to. As in the near infrared LCTF calibration
plot, here also the desired wavelengths sent to the LCTF controller
were plotted on the X axis and the measured wavelengths from the
hyperspectral analysis were plotted on the Y axis, see FIG. 27.
Points in FIG. 27 are example values and the dashed line represents
a linear regression curve fit (Y=0.9984.times.-0.5013. R2=0.9999).
From this relationship, a small look-up table was generated in the
V++ software tuning the LCTF to the desired wavelength that the
operator desires.
[0287] Similarly the spectral band-pass for the visible low
resolution filter varies from 3.7 nm at 400 nm to 12.2 nm at 700
nm, see FIG. 28. The conclusion from these Figures is that the LCTF
band-pass is wavelength dependent. This bandwidth is more than
sufficient to spectroscopically resolve HbO.sub.2 and Hb which have
spectral characteristics that are several times broader than the
broadest band-pass.
[0288] In the hyperspectral imaging system images are collected
sequentially at wavelengths differentiated continuously by the LCTF
that tunes from one wavelength to another. LCTF takes typically 50
to 150 ms to switch from one wavelength to another, i.e. the
response time. Any imaging application collecting sequential image
data through the LCTF must account for the response time required
to tune the LCTF. In the V++ program developed to control the
acquisition of the hyperspectral imaging system, a tune-delay time
is introduced that accounts for the LCTF tune time. An example was
done to determine this tune-delay or tune-wait time for the LCTF's;
the setup for which is shown in FIG. 29. In this example, light
focused on a spectralon target is reflected onto the FPA through
the LCTF and a hyperspectral cube was collected. Hyperspectral
cubes were collected for varying tune-delay time starting from 0 ms
to 100 ms in increment of 10 ms.
[0289] Similarly, spectral analysis was performed on the visible
LCTF, and as shown in FIG. 30 there is little or no difference
between 0 ms, 30 ms and 50 ms, thus suggesting a tune-delay of 0 ms
for the visible LCTF.
[0290] The tune-delay time increases the exposure time for
collecting the hyperspectral, hence keeping it to a minimum is
advantageous. Tune-wait/Tune-delay time depends on various factors
such as the liquid crystals and the electronic components in the
circuitry. Thus, tune-wait time is typical for each LCTF.
[0291] The following describes the calibration of the FPA. The
spatial resolution of the microscopic hyperspectral imaging system
was characterized, which is defined as the ability to distinguish
between two closely spaced objects within an image. The spatial
resolution of the system was established by computing the percent
contrast which depends on various factors such as the focal plane
array, filter, slit lamp magnification, type of camera lens,
f-stop, depth of field, and degree of pixel binning. Modifications
of any of the factors alter the spatial resolution of the system.
In view of the foregoing, spatial resolution was evaluated with
various degrees of binning and magnification. The percent contrast.
C, was determined using the expression equation (9) shown below,
and plotted as a function of spatial resolution in millimeter.
C = ( I max - I min I max - I min ) .times. 100 ( 9 )
##EQU00006##
[0292] where: [0293] I.sub.max is the maximum intensity reflected
by a line of the resolution target (white bar) of the resolution
target are shown in FIG. 31; and, [0294] I.sub.min is the minimum
intensity from the nonreflecting area between the lines (dark bar)
of the resolution target are shown in FIG. 31.
[0295] A 1951 quartz USAF resolution target is depicted in FIG. 31
where A depicts a portion of the target and the corresponding
reflected intensity taken along a row of pixels is depicted in B
was used to determine the spatial resolution. It should be
appreciated that a line pair is defined as the distance between the
first edge of a white bar and the second edge of a dark bar.
Generally, the spatial frequency, or number of line pairs,
increases (the number of pairs/unit length) as the percent contrast
decreases. The quartz target was placed in the imaging path and a
single image was collected which was then analyzed to determine the
percent contrast by using equation (7). This procedure was followed
for determining spatial resolution of the near infrared imaging as
well as visible imaging system. Binning and magnification were
changed keeping other parameters the same, and images were
collected and analyzed to obtain the percent contrast. Percent
contrast calculations were done for the near infrared microscopic
hyperspectral imaging system and plotted as a function of spatial
resolving power (mm) in FIG. 32. The graph of FIG. 32 includes
percent contrast as the dependent variable (y-axis) and spatial
resolution in millimeter as the independent variable (x-axis) and
it has been determined from the regression curve fit models when
various binning embodiments cross the 20% contrast threshold set by
the Rayleigh criterion, i.e. the horizontal dashed line.
[0296] Percent contrast of the visible microscopic hyperspectral
imaging system is plotted as a function of spatial resolving power
(mm) in FIG. 32. Each embodiment shown in FIG. 32 has a polynomial
regression curve fit model. For a magnification of 100.times. and
binning of 1.times.1 (diamond shaped data points), the regression
equation is y=-996.04.times.2+548.86x+5.7296. Keeping the
magnification same but changing the binning to 2.times.2 (square
shaped data points), the regression equation is
y=-919.92.times.2+544.64x+0.8112. For a binning of 3.times.3
(triangle shaped data points), and 4.times.4 (cross shaped data
points), the respective regression equations are
y=-660.48.times.2+498.39.times.-4.0232 and
y=-531.26.times.2+458.81.times.-5.5901. Again, in all the above
embodiments, the dependent variable (y-axis) is percent contrast
and the independent variable (x-axis) is spatial resolution in
millimeter, and it is determined from the regression curve fit
models when it crosses the 20% contrast threshold set by the
Rayleigh criterion, i.e., the horizontal dashed line. The spatial
resolution for the visible hyperspectral imaging system with the
magnification at 10.times. is 0.027 mm at binning 1.times.1, 0.038
mm at binning 2.times.2, 0.052 mm at binning 3.times.3 and 0.06 mm
for binning 4.times.4. As mentioned above, the binning decreases
the spatial resolution which is verified by the spatial resolution
numbers obtained.
[0297] Another interesting comparison was made using the visible
microscopic hyperspectral imaging system by obtaining percent
contrast measurements for the visible system without the presence
of the relay optics in the light path. These measurements are
plotted in FIG. 33. Regression equations were obtained and analyzed
keeping the Magnification at 100.times., and changing the binning
on the FPA to find out the spatial resolution for the system
without the relay optic, yielding spatial resolutions of 0.072 mm
for binning of 1.times.1, 0.104 mm for binning of 2.times.2, 0.129
mm for binning of 3.times.3 and 0.169 mm for binning of 4.times.4.
The spatial resolution for the system without the relay optics was
found to be worse then that for the system with the relay optics in
the light path. The relay optics were introduced in the system to
increase the magnification of the system and allow the system to
occupy a larger area of the CCD chip thus improving the overall
spatial resolution of the system as verified by the spatial
resolution numbers for the system with and without the relay
optics.
[0298] The following describes the system performance of an
embodiment of the present invention. Hyperspectral imaging is
emerging as a powerful imaging tool to exemplify spectral as well
as the spatial characteristics. An embodiment of a present
invention microscopic near infrared hyperspectral imaging system
was used to image, non-invasively, microvascular perfusion of the
dermal tissue, in vivo, for assessing oxyhemoglobin contribution
during a resting condition for ten subjects. The hyperspectral data
deconvoluted for the contribution of oxyhemoglobin is gray scale
encoded, meaning the greater the pixel intensity the greater the
oxyhemoglobin contribution as depicted by the gray scale bar
associated with an image.
[0299] The microscopic hyperspectral system was further applied
towards imaging the microvasculature structure present in the
anterior region of the human eye, especially the sclera, which
required imaging in the visible spectrum, i.e., 520 nm-602 nm. A
hyperspectral image cube was obtained using the visible slit lamp
microscopic imaging system, with parameters such as setting the
magnification to 10.times., setting the binning to 2.times.2 and
varying the wavelength from 520 nm-602 nm, incremented at 2 nm.
These parameters successfully reduced the acquisition time for the
hyperspectral image cube to within 5 seconds which is critical for
imaging the eye since the subjects involved in the data collection
are not allowed to blink or move their eye since that would create
artifacts in the images.
[0300] The hyperspectral cube was deconvoluted for oxyhemoglobin
contribution values that are color encoded and spatially depicted
in FIG. 34A. A small region over a high oxyhemoglobin contribution
region, shown within the small black box, as suggested from the
color bar, i.e., red areas, was selected and pixels under the
region were averaged to obtain a spectrum shown in FIG. 34C which
resembles the oxyhemoglobin spectrum. Similarly, the spectrum under
the area of the small pink box was selected and pixels under the
region were averaged to obtain a spectrum shown in FIG. 34B, and
represents a deoxyhemoglobin spectrum.
[0301] Seven healthy subjects were imaged to obtain scleral tissue
oxyhemoglobin contributions for assumed arterial and venous
structures in the image as displayed by the spectroscopic
information. The average oxyhemoglobin contribution values for
these seven subjects were obtained. The regions from under which
the above average values evaluated were not random but visually
selected depending on the spectroscopic information the region
yielded.
[0302] Qualitative values with quantitative evaluations pending
were obtained. These values demonstrate that there exist marked
differences between arterial and venous structure oxyhemoglobin
contributions and that the system is capable of distinguishing
between arterial and venous structure in the tissue being
imaged.
[0303] The following description sets forth embodiments of a
present invention microscopic visible hyperspectral imaging
systems. After demonstrating capabilities of the microscopic
hyperspectral imaging system using dermal microvasculature imaging,
other applications of the present invention are now discussed, in
particular, imaging of the human eye, especially the sclera. The
sclera is the white portion of the eye, covered by the episclera,
tenon's capsule, and the conjunctiva. The blood vessels seen on the
scleral surface are actually found between these various regions.
Scleral disease such as scleritis and episcleritis, are uncommon in
ophthalmology patients, however they present symptoms for serious,
painful and life threatening systemic diseases such as rheumatoid
arthritis, syphilis, spondylitis, etc.
[0304] The anterior region of the human eye was imaged using the
microscopic visible hyperspectral imaging system. The visible
system uses the CoolSNAP.sub.ES CCD in conjunction with the visible
low resolution LCTF and the slit lamp microscope. The
CoolSNAP.sub.ES CCD's spatial resolving power at a binning of
2.times.2 and a magnification of 10.times. is 0.038 mm, which is
better then the resolving power of the PIXIS 400 BR with the same
parameters, i.e., 0.083 mm. Since speed is an important trade off
with the imaging of the eye, the 20 MHz readout rate is preferred
against 2 MHz readout rate of the PIXIS FPA. Also to decrease the
acquisition time of the system the spectral range and spectral
resolution were reduced to 520 nm-602 nm at increments of 2 nm,
instead of the previously reported macroscopic version of the
visible hyperspectral system that used a spectral range of 520
nm-645 nm, at increments of 1 nm. These were the two major trade
offs considered to reduce the data acquisition time to 5 seconds,
which acquisition time is clinically suitable for eye imaging
applications. Inputting these example parameters with the slit lamp
magnification set to 10.times., the sclera of the subject was
imaged to obtain a hyperspectral data cube that upon deconvolution
with the visible range reference spectra of oxyhemoglobin and
deoxyhemoglobin rendered an image that contained spatially
distributed contributions of oxy-hemoglobin.
[0305] FIG. 35 is an illustration of the basic setup of a DLP
hyperspectral imaging system of the present invention. The setup
for noninvasive DLP.RTM. microscopic reflectance hyperspectral
imaging system consisting of a spectrally-tunable source, optics, a
detector, a computer, and a DLP.RTM.. The radiometric radiation
source can be a 12V/30 W halogen source and/or a QTH (quartz
tungsten halogen) source providing broad band illumination. The
radiation is directed towards the DLP, using relay optics. The DLP,
differentiates the broadband reflected light into individual band
passes or provides a spectrum of wavelengths and is directed to the
object to be analyzed. Major components include, but not limited
to: white laser source 350 and spatial light modulator 352, such as
a DLP.RTM., used to create one or more light wavelengths or a band
spectrum which are captured as raw images 354. Raw images 354 are
then processed to detect known chemometrics 356, which are then
output as visual image 358. The user is then able to determine if
they wish to save or discard the image, which image may be stored
on storage device 360, e.g., hard drive, CD-ROM, etc.
[0306] Another example of the present invention is demonstrated as
follows. The hyperspectral slit imager acquisitions hyperspectral
data that is gray scale encoded as the percentage of oxyhemoglobin
at each pixel of the image along with its associated spectrum.
Spectra were selected from darker pixels indicating a venous
structure, the purple spectrum, versus bright pixels indicating an
artery, the red spectrum, as shown in FIG. 36. FIG. 36 demonstrates
retinal imaging of oxyhemoglobin contribution using visible
reflectance hyperspectral imaging system. Other part of the eye,
for example the conjunctiva, can also be imaged and can also be
color encoded, the redder the pixel, the greater the percentage of
oxyhemoglobin, and a variety of image processing techniques may be
applied, for example, feature extraction or optical biopsy, for
future evaluation (see FIG. 37). FIG. 37 is a hyperspectral imaging
of small vessels within human conjuctiva of the eye.
[0307] FIG. 38 is an image taken using one embodiment of the
present invention by using a DLP system with a "2 shot"
illumination scheme. The results were obtained using illumination
with a spectrum of light. This embodiment includes illumination
with the difference of two (2) principle component spectra (one
spectrum that is consistent with the positive values and the other
in which the negatives are multiplied by negative one, illuminated
and then the image multiplied by negative one again). The resulting
images are then normalized and processed.
[0308] In another embodiment, the present invention includes using
optical filters as a method for fluorescing indocyanine green dye
(ICG). The embodiment enables the identification of organs such as
the biliary structures when ICG is injected into a blood vessel
such as the femoral artery of an animal or human. This embodiment
enables fluorescence of biliary structures. Here, the illuminator
can be a Texas Instrument's DLP illuminator and the fluorescence
image of the ICG can more precisely illuminate the exact spectrum
for maximizing the fluorescence of the ICG and imaging it in the
biliary structures. The present Applicants have completed the
optical filter system and have successfully imaged ICG and bile
fluorescence in capillary tubes. This embodiment can further be
modified to be used in animals and humans or clinical trials.
[0309] In certain embodiments, the present invention can be used in
conjunction with using a supercontinuum laser instead of a plasma
source to illuminate the DMD chip. The laser used in this
embodiment does provide greater intensity, smaller band passes, and
larger spectral range. Supercontinuum or white-light laser sources
are recognized in the art. Briefly, they are typically suitable for
applications in spectroscopy and microscopy. It can include a pump
laser and a microstructured fiber (either a photonic-crystal fiber
or a tapered fiber). Although 80 fs pulses from a Ti:sapphire laser
and 200 fs pulses from an ytterbium (Yb):glass laser have been used
to generate supcrcontinuum light, researchers have demonstrated a
more-compact and lower-cost portable femtosecond supercontinuum
source with a small footprint. Moreover, in other embodiments, the
present invention can include a light emitting diode (LED) based
illumination source. The source may include a single LED or a
plurality of LEDs.
[0310] In some embodiments, the present invention includes image
analysis of a patient wearing a silk gown. Near infrared light is
known to see through silk. Wearing a silk gown allows the patient
to cover them selves while letting the NIR camera see through to
the tissue for analysis. In this embodiment, the patient can lay on
a piece of clear material, e.g. plexiglass or glass. A layer of oil
can help negate index or refraction effects and improve imaging
similar to an oil immersion microscope objective lens. This
embodiment provides maximum privacy to the person being imaged,
preserves body heat and provides accessibility for the surgeon.
[0311] In another embodiment, the present invention can also
include at least one deconvolution algorithm. The algorithm is
typically used for normalizing each spectrum at each pixel as a
step in the pre-processing. This algorithm is counter-intuitive to
the skilled artisan in the art or for the general spectroscopist.
The skilled artisan would not consider and would question this
unique and unexpected algorithm.
[0312] In yet another embodiment, the present invention can be used
during kidney surgery. The kidney is highly vascularized and tends
to bleed copiously during surgery, as such, surgeons minimize this
bleeding by clamping off the blood supply. However, clamping off
this organ also limits greatly the time permitted to perform the
surgery without affecting long-term kidney function. The present
invention allows a small amount of blood to leak past the clamp to
perfuse the kidney, thereby giving the surgeon more time to perform
the surgery before permanent kidney tissue damage takes place. The
hyperspectral imager of the present invention aids in visualizing
and determining when the kidney is at risk of permanent damage
before damage occurs by providing the surgeon real-time information
regarding kidney cell and tissue status.
[0313] The present invention can be applied in various field, e.g.,
in cholecystectomy, amputations, burns, skin flap evaluation,
visualizing areas of angiogenesis, probes that bind antigens and
absorbs NIR during pathological evaluations and in vivo, quality
control of pharmaceuticals, monitoring vascular changes and drug
discovery in response to pharmaceuticals, monitoring diabetic
retinopathy, diseases such as cancer, diabetes, sickle cell,
anemia, bilirubin, raynauds, ulcers, burns, skin flaps, surgery,
gallbladder, brain, monitoring wound healing, and early detection
of wound infections.
[0314] FIG. 39 is a graph that shows one hundred twenty six (126)
separate wavelengths using a LCTF to separate the bands prior to
illumination, after which an image is captured with each bandpass
or with a single frame including up to one hundred twenty six (126)
images. FIG. 40 is a graph that shows a band spectrum in which a
digital micromirror array was used to create spectral illumination
that allows for a lower number of images per frame. FIG. 41 shows a
comparison of data obtained with the LCTF and the digital
micromirror array illumination using the single bandwidths of FIG.
39. FIG. 42 shows a comparison of data obtained with the LCTF and
the digital micromirror array illumination using the single
bandwidths of FIG. 40.
[0315] FIG. 43 is a flow chart of the basic "2-shot" algorithm.
FIG. 44 is a flow chart of the processing of the data cube obtained
using the basic "2-shot" method.
[0316] FIG. 45 is a flow chart of the acquisition method of the
basic "2-shot" method. Briefly, normalized absorption spectra in
the 520 nm-645 nm wavelength range for HbO.sub.2 and Hb are
subtracted from each other, and the positive areas become the two
illumination spectra. The relative intensity of each illumination
spectrum is stretched from 0 to 100 to maximize the overall light
intensity and match the required OL-490 input format. Each data
cube (M.times.N.times.2) consists of one M.times.N pixel image
taken in the first image capture and one M.times.N pixel image or
shot captured, taken for the second image or shot captured.
[0317] FIG. 46 shows sample images obtained for finger occlusion at
different times using the present invention. FIG. 47 compares the
images obtained using visible light and near infrared (NIR) of the
reperfusion of a foot following removal of the shoe. FIG. 48 shows
the in vivo hyperspectral imaging of human tissue, spatial
variation of percentage of HbO.sub.2 and surface temperature in
response to a burn.
[0318] The present invention describes a "3 shot" illumination
method in an integrated DLP hyperspectral imaging (HSI) system
which reduces the number of image frames required to generate a
processed image that is color-coded based on matching the
reflectance of each pixel in an image to known reflectance spectra.
The "3 shot" method of the present invention can replace
traditional hyperspectral imaging which requires many frames to be
acquired at contiguous wavelengths throughout a wavelength range of
interest.
[0319] In the "3 shot" method of the present invention, a sample is
illuminated with three complex spectra which have varying levels of
intensity at each wavelength across the wavelength range of
interest. Traditional hyperspectral imaging captures images for
discrete wavelengths with narrow bandwidths. Contrarily, in the "3
shot" method, a series of narrow bandwidths is replaced with a
broadband spectrum, in which the relative intensity of each
wavelength is attenuated according to reference spectra. Without
explicitly measuring the reflectance spectrum of a tissue, the "3
shot" method implicitly determines how closely the sample matches
one of a pair of reference spectra.
[0320] After illuminating a sample with three broadband components
of reference spectra, the "3 shot" method of the present invention
subtracts and divides the resulting three images, and outputs
several processed images per second. Present technology, on the
other hand, illuminates or filters multiple narrow bandwidths,
measures the sample reflectance for each of those discrete
bandwidths, compares the reflectance spectrum to known reference
spectra, and outputs, at most, one processed image per several
seconds.
[0321] The advantages of the "3 shot" method of the present
invention over traditional hyperspectral imaging makes the present
invention ideal for real-time visualization of the molecular
components of tissue for clinicians and surgeons. In addition the
"3 shot" method of the present invention can be implemented in any
hyperspectral imaging device which has the ability to illuminate
with a variable intensity spectrum.
[0322] The present invention describes a "3 shot" illumination
method in an integrated DLP HSI instrument. The "3 shot" method
generates an image of the target very quickly in comparison to the
more commonly used and previously described spectral sweep
illumination method.
[0323] The data acquisition software automatically tune the DLP:
technology and trigger the camera for collecting a series of
spectroscopic images formatted as a hyperspectral image cube. Next,
the spectroscopic image data are deconvoluted using chemometric
analysis methods. The resulting gray scale or color encoded images
provide the clinician with a non-invasive visualization of the
chemical state within the micro-vasculature perfusing the tissue
while the patient is in the clinic or surgery.
[0324] Hyperspectral imaging is the equivalent of reflectance
spectroscopy at many discrete points arranged in a spatial pixel
array. Reflectance spectroscopy involves collecting the reflectance
of light at each wavelength, in the spectral dimension. Projecting
the spectral dimensions to each pixel in an image results in a 3D
data cube that contains the intensity spectrum for each pixel in a
2D spatial array. A hyperspectral data cube can be acquired in two
fundamental ways: scanning a point spectrometer over a spatial
area, or sweeping the wavelengths of light incident on an array
detector. With the 2-dimensional array detectors available today,
the reflectance of light can be collected in two spatial and one
spectral dimension.
[0325] The first way to capture a hyperspectral image cube is by
moving a sample on a mechanical stage to raster-scan a point
spectrometer through the image field of view (FOV). (See Sellar R
G, Boreman G D, Classification of imaging spectrometers for remote
sensing applications. 2005;44:1-3). A point spectrometer couples a
very small area of broadband light into a prismatic grating. The
grating spatially disperses polychromatic light into its
monochromatic components, which are then uniquely detected to
quantify the intensity spectrum as a function of wavelength. The
spectrum measured by a point spectrometer is a measurement of all
the light coupled into the spectrometer. To generate a 3D
hyperspectral image cube, the point spectrometer must capture an
intensity spectrum at each spatial pixel in the image FOV.
[0326] The second fundamental method of hyperspectral imaging is
acquiring all spatial information at a wavelength of interest and
scanning through a spectral range using optical or electro-optical
filters. A liquid crystal tunable filter (LCTF) or acousto-optic
tunable filter (AOTF) is used to cut off all broadband light except
for a precisely tuned narrow bandpass. (See Sellar R G, Boreman G
D, Classification of imaging spectrometers for remote sensing
applications. 2005; 44:1-3). A scientific grade camera captures an
image of a full 2D scene (e.g., tissue sample) at each narrow
bandpass through a sweep of contiguous wavelengths, resulting in a
3D hyperspectral image cube.
[0327] The instrumentation in the present invention consists
primarily of a spectral light engine with DLP.RTM. technology
providing the spectroscopic illumination, a digital camera with a
scientific grade CCD for imaging, and software designed and
developed to manage the data acquisition and the chemometric
visualization. The hyperspectral imaging system comprises an OL 490
spectral light engine created with Texas Instruments' "Digital
Light Processor" (DLP.RTM.) technology. In the OL 490,
polychromatic visible light is dispersed onto the micromirror array
of a DLP chip so that each column of micromirrors corresponds to a
narrow band of monochromatic light. Programming the mirrors
individually allows the user to precisely define the intensity of
each wavelength in the light engine's optical output spectrum. The
OL 490 can illuminate with narrow bandpasses of light as in
traditional hyperspectral imaging, but the novelty of the DLP HSI
is using the OL 490 to illuminate with complex broadband
illumination spectra. Regardless of the illumination scheme, the
light from the OL 490 is reflected from a tissue sample of interest
and detected by a scientific grade CCD focal plane array. The data
acquisition software automatically tunes the DLP.RTM. technology
and triggers the camera for collecting a series of spectroscopic
images formatted as a hyperspectral image cube. Then, the
spectroscopic image data are deconvoluted using chemometric
analysis methods. The resulting gray scale or color encoded images
provide the clinician with a non-invasive visualization of the
chemical state within the micro-vasculature perfusing the tissue
while the patient is in the clinic or surgery. A simplified
schematic of the instrumentation described above is presented in
FIG. 49.
[0328] One of the many goals of the present invention is to
demonstrate that the "3 shot" illumination method in a
DLP.RTM.-based visible hyperspectral imaging system can be used
routinely for a variety of medical applications. The speed afforded
by the "3 shot" illumination method of the present invention
reduces the current acquisition time from minutes to seconds making
hyperspectral imaging a practical everyday surgical and clinical
tool for imaging.
[0329] The DLP.RTM. HSI with "3 shot" illumination and chemometric
deconvolution can be applied to all products that use light, for
example, the fields of clinical endoscopy, clinical chemistry,
microscopy, surgical microscopy, drug discovery, microarray
scanners and microplate readers.
[0330] As described above, the present invention can be used during
kidney surgery. The kidney is highly vascularized and tends to
bleed copiously during surgery, as such, surgeons minimize by
clamping off the blood supply. However, clamping off this organ
also limits greatly the time permitted to perform the surgery
without affecting long-term kidney function. The present invention
allows a small amount of blood to leak past the clamp to perfuse
the kidney, thereby giving the surgeon more time to perform the
surgery before permanent kidney tissue damage takes place. The
hyperspectral imager with "3 shot" illumination of the present
invention aids in visualizing and determining when the kidney is at
risk of permanent damage before damage occurs by providing the
surgeon real-time information regarding kidney cell and tissue
status.
[0331] Currently there are three illumination methods implemented
in the DLP HSI for hyperspectral imaging: Full Spectral Sweep.
Spectral Sweep, and the "3 Shot" method described in the present
invention. These illumination methods will be discussed in the
remaining sections; however, new illumination methods are easily
created by programming the spectral light engine. In fact, the
integrated DLP HSI system of the present invention has the ability
to change the illumination spectrum to any imaginable narrow or
broadband spectrum in the wavelength range of the OL 490. This
versatility means that the DLP HSI can be used for any number of
imaging applications.
[0332] The full spectral sweep method is now described. To utilize
the practical wavelength range of the system, the full spectral
sweep method sweeps bandpass illuminations having 10 nm bandwidths
from 450 nm to 650 nm in 4 nm increments. A hyperspectral image
cube acquired with this method consists of 51 slices and is
generally used to explicitly measure the absorbance spectroscopy of
an area of interest. Formally, the following series of
illuminations is programmed into the OL 490:
{L.sub.1,FullSweep=L.sub.OL490(.lamda..sub.1),L.sub.2,FullSweep=L.sub.OL-
490(.lamda..sub.2), . . .
L.sub.51,FullSweep=L.sub.OL490(.lamda..sub.51)} (10)
for,
.lamda..sub.n=450+4(n-1), n={1:51} (11)
where, L.sub.OL490(.lamda..sub.n) is the total radiant power
emitted from the OL 490 at a center wavelength, .lamda..sub.n, with
a 10 nm full width half maximum (FWHM) bandwidth.
[0333] After proportioning with the background, each slice in the
cube represents the absorbance of the bandpass illumination at
which the image is acquired. Thereby, each spatial pixel in the
processed cube represents a discrete absorbance spectrum of the
tissue or other material in the physical position corresponding to
that pixel. The wavelength limits of the full spectral sweep method
intentionally truncate the wavelength limits of the OL 490 since
the output intensity of the OL 490 is very low near the extremes of
its wavelength range.
[0334] This illumination method is only used to help calibrate the
system as a spectrophotometer, and no algorithms currently exist to
process its image cubes for visualization of tissue chemistry.
However, with the spectral information acquired using full spectral
sweep illumination, it is possible to search for spectral
signatures of chemical chromophores and develop new processing
algorithms.
[0335] The spectral sweep method is now described. To mimic the
illumination of LCTF-based hyperspectral imaging systems, the
spectral sweep method sweeps bandpass illuminations with 10 nm
bandwidths from 520 nm to 645 nm in 1 nm increments. A
hyperspectral image cube acquired with this method consists of 126
slices and explicitly measures the absorbance spectroscopy of an
area of interest. Formally, the following series of illuminations
is programmed into the OL 490:
{L.sub.1,Sweep=L.sub.OL490(.lamda..sub.1),L.sub.2,Sweep=L.sub.OL490(.lam-
da..sub.2), . . . L.sub.126,Sweep=L.sub.OL490(.lamda..sub.126}
(12)
for,
.lamda..sub.n=520+(n-1), n={1:126} (13)
where, L.sub.OL490(.lamda..sub.n) is again the total radiant power
emitted from the OL 490 at a center wavelength, .lamda..sub.n, with
a 10 nm FWHM bandwidth.
[0336] After proportioning out the background, each slice in the
cube represents the absorbance of the bandpass illumination at
which the image is acquired. Thereby, each spatial pixel in the
processed cube represents a discrete absorbance spectrum of the
tissue or other material in the physical position corresponding to
that pixel, identical to the full spectral sweep method, but with
better spectral resolution and narrower wavelength range. Each
pixel absorbance spectrum is then compared to reference spectra for
oxy-hemoglobin (HbO.sub.2) and deoxy-hemoglobin (Hb) in FIG. 50A by
multivariate least squares analysis to calculate the percent
HbO.sub.2 for that pixel. The output bitmap, after processing with
`Oxyz--Jet`, is a color-coded two-dimensional image, where more
intense pixels (red) signify that the pixel absorbance spectrum
more closely resembles the HbO.sub.2 reference and less intense
pixels (blue) signify that the pixel absorbance spectrum more
closely resembles the Hb reference.
[0337] Spectral sweep illumination is slow for two fundamental
reasons: acquisition of 126 frames takes 126 times longer than
acquisition of a single frame and processing of such large data
cubes takes considerable time. One suggested solution to this speed
issue has been selecting only a few narrow illumination bands to
significantly reduce the number of slices in the cube, thereby
reducing the acquisition time and processing time. (See Guo, B,
Gunn, S R, Damper, R I, Nelson, J D B, Band selection for
hyperspectral image classification using mutual information. 2006;
3:522-526; and, Du, Z, Jeong, M K, Kong, S G, Band selection of
hyperspectral images for automatic detection of poultry skin
tumors. 2007; 4:332-339). This solution has been shown to speed up
data acquisition and processing, but does not necessarily gather
the same spectral information as a spectral sweep.
[0338] The present invention "3 shot" method is now described. The
"3 shot" method of the present invention enables the gathering of
spectral information equivalent to the spectral sweep method with
fewer images being captured. The "3 shot" method sequences through
three complex broadband illumination spectra in the 527 nm to 638
nm wavelength range. A hyperspectral image cube acquired with this
method consists of three (3) slices and is used to calculate the
percent HbO.sub.2 for each spatial pixel without explicitly
measuring the absorbance spectra. Formally, the following
illuminations are programmed into the OL 490:
L 1 , 3 shot = n = 8 119 L OL 490 ( .lamda. n ) V 1 ( .lamda. n ) (
14 ) L 2 , 3 shot = n = 8 119 L OL 490 ( .lamda. n ) V 2 ( .lamda.
n ) ( 15 ) L 3 , 3 shot = n = 8 119 L OL 490 ( .lamda. n ) ( 16 )
##EQU00007##
[0339] For the wavelengths described by equation 8,
V.sub.1(.lamda..sub.n) represents the positive subtraction of
normalized reference spectra and for the wavelengths described by
equation 9, V.sub.2(.lamda..sub.n) represents the negative
subtraction of normalized reference spectra as shown in FIG. 50B.
The limits of summation are intentionally n=8 and n=119, because
the reference spectra have been filtered by a moving average filter
that trims seven discrete values from the beginning and end of each
spectrum.
[0340] FIG. 51 is a block diagram that shows the experimental
procedure used to capture images using the "3 shot" method of the
present invention and the MATLAB.RTM. algorithm used to process the
images captured using the "3 shot" method of the present
invention.
[0341] FIGS. 52A-52C are an illustration of the "3 shot"
illumination method of the present invention used for visualizing
blood oxygenation. FIG. 52A shows the normalized absorbance spectra
in the 520 nm-645 nm wavelength range for HbO.sub.2 and Hb
subtracted from each other, where the positive areas become the two
illumination spectra. FIG. 52B shows that the relative intensity of
each illumination spectrum is stretched from 0 to 100 to maximize
the overall light intensity and match the required OL-490 input
format. FIG. 52C shows the absorbers and scatterers that are not
HbO.sub.2 and Hb.
[0342] After proportioning out the background, each slice in the
cube represents the absorbance of the broadband illumination at
which the image is acquired. Absorbance spectra are not explicitly
measured, but a mathematic combination of the three slices results
in a single two-dimensional image where higher pixel values
indicate absorbance similar to HbO.sub.2 and lower pixel values
indicate absorbance similar to Hb. The data cube is forty two (42)
times smaller than the spectral sweep data cube, the processing
algorithm is considerably simpler, and the resulting visualization
of tissue oxygenation is the same. Thus, a color-coded bitmap image
that appears nearly identical to one generated by the spectral
sweep method can be generated by the "3 Shot" method much more
quickly.
[0343] Three complex broadband illumination spectra were derived
from the reference absorbance spectra of Hb and HbO.sub.2 for the
"3 Shot" method described in the present invention. Initially the
OL 490 was programmed to illuminate with these spectra by directly
copying the spectra into Gwectra files. The optical output measured
by a spectrometer when these spectra were explicitly programmed is
seen in FIG. 53A. Explicitly using the calculated spectra, the
measured optical output was significantly different than the
desired illumination. In the plot, the solid line (measured optical
output) is shifted horizontally from the dotted line (desired
optical output), and the measured first peak is not as intense as
the desired first peak.
[0344] In order to generate an illumination spectrum that more
closely matches the desired optical output, a new Gwectra file was
written. In the new file, the intensities at each wavelength were
shifted horizontally according to the center wavelength calibration
curves mentioned earlier and vertically according to the error
between the measured and desired intensities. Programming the OL
490 with the new Gwectra file, the optical output is seen in FIG.
53B. In the plot, the solid and dashed lines (measured optical
output and desired optical output, respectively) coincide much
better than in the previous plot.
[0345] Reprogramming the mirrors of the DMD in the OL 490 allows
for refinement of illumination spectra for the "3 Shot" method. The
same wavelength calibration and intensity adjustment was performed
for the other two complex broadband illuminations. For most of the
wavelength range, the actual optical output matches well with the
desired optical output. Due to the minimum bandpass limitation when
using the 350 .mu.m slit, it is fundamentally impossible to mimic
the two bumps desired between 620 nm and 640 nm. If lower overall
light intensity is adequate, the 150 .mu.m slit, which has a
smaller minimum bandpass limitation, can be used to further refine
the actual optical output.
[0346] Speed is a key requirement in the DLP hyperspectral imaging
system design. Camera selection is often based primarily on speed.
The purpose of the "3 Shot" method of the present invention is
speed. One of the main reasons hyperspectral imaging is not a
primetime medical imaging modality is its inherent lack of speed.
The DLP HSI prototype of the present invention is purposefully
versatile. The user can adjust exposure times (constant or
variable), binning of the CCD array, processing algorithms for
chemometric visualization, lens aperture, and OL 490 illumination
methods. To address the need for speed, it is important to
understand the role each of these parameters play in determining
the final frame rate of the system.
[0347] Several fundamental bottlenecks limit the maximum speed of
the DLP HSI. The minimum exposure time of the OL 490 is 80 .mu.s,
the minimum exposure time of the HQ2 is 210 .mu.s, the maximum
transfer rate of the camera to the computer is 20 MHz, and the
image processing time is significant.
[0348] Theoretically, increasing exposure time in each slice will
increase acquisition time for each image thereby increasing total
acquisition time for the hyperspectral cube. Practically, exposure
time is varied to achieve near 12,000 counts at the maximum
intensity when capturing a background cube, so exposure time is not
considered an independent variable. Rather, exposure time depends
on lens aperture, camera binning, gain, illumination intensity, and
focal distance. Illumination intensity is dependent on illumination
method and is fixed for either the spectral sweep illumination
method or the "3 shot" illumination method of the present
invention. Focal distance is variable and is adjusted depending on
the physical constraints of the clinical or surgical setup. In
surgery, the camera must be at least 1 m away from the subject and
should be unobtrusive to the surgeon. On the lab bench, the camera
is set to a focal distance of 45 cm. Camera gain can be varied
according to CoolSNAP HQ2 specifications while the lens aperture
can be varied from f/16 to f/1.4. The two parameters in concert
affect the exposure time primarily, with side effects of focus
range and detector noise. For the following speed tests, `high
gain` refers to a camera gain of 3 and a lens aperture of f/1.4,
while `low gain` refers to a camera gain of 1 and a lens aperture
of f/8. Camera binning affects both exposure time and spatial
resolution. For the following speed tests, bins of 2.times.2 and
4.times.4 are set. In the clinic, `low gain` and 4.times.4 binning
is typically used to ensure good quality imaging at reasonable
speeds.
[0349] A final parameter that does not affect exposure time but
still affects the speed of the system is the processing algorithm.
The most basic processing algorithm reads the hyperspectral image
cube into MATLAB and outputs one of the raw images as a Windows
bitmap file for display. For visualization of oxygenation, "Oxyz
Jet" and "3shot Jet-*," where * can be any pair of thresholds, are
the processing algorithms used. In each of the oxygenation
algorithms, the hyperspectral image cube and background cube in the
same parent directory are read into MATLAB for calculating
absorbance. Chemometric analyses transform the absorbance cube into
a chemically relevant image which is output as a bitmap for
display.
[0350] To measure the effect of each parameter on the overall speed
of the DLP HSI, the system was set to a focal distance of 45 cm
with the Spectralon target filling the detector field of view
(FOV), and a combination of parameters based on the
2.times.2.times.2.times.2 factorial analysis model presented above
are programmed into the GUI. Temporary data cubes are acquired to
set the exposure time that allows for a maximum response near
12,000 counts as seen in Table 4. Due to the illumination intensity
of the "3 Shot" illuminations the exposure time could not be
adequately lowered to keep the detector from saturating under `high
gain` conditions.
TABLE-US-00004 TABLE 4 Exposure time (.mu.s) required to get
maximum response near 12,000 counts for the given combination of
parameters. Spectral Sweep 3 Shot Raw Oxy Raw Oxy 2 .times. 2 Low
Gain 980 980 5750 5750 High Gain 184000 184000 saturates saturates
4 .times. 4 Low Gain 200 200 1250 1250 High Gain 40000 40000
saturates saturates
[0351] After the exposure times are determined, the system is set
to acquire continuously for a period of ten (10) minutes. The
number of output bitmaps created within that ten (10) minutes is
counted, and the time to output one processed image for the set
parameters is calculated by equation 17.
t = 600 # bitmaps ( 17 ) ##EQU00008##
[0352] Table 5 shows the results of all 12 runs. As expected from
casual observations, the fastest time to output one "3 Shot" image
visualizing oxygenation is near 1/3 of a second (0.33 seconds),
which means that under these conditions, the system is operating a
nearly 3 frames per second (fps). Acquiring Spectral Sweep images
with 2.times.2 binning causes an exception violation in the
software because the resulting image cube is too large for MATLAB
to process (696.times.520.times.126). Comparing Oxy to Raw, it
appears that the processing algorithm has an increased effect on
speed when binning is reduced, but no differential effect on speed
when gain is increased. From the spectral sweep results, it also
appears that the processing algorithm has a greater effect on speed
than does gain level. These observations may be helpful in
determining how to make image acquisition and processing most
efficient with the DLP HSI.
TABLE-US-00005 TABLE 5 Time (sec) to output one processed image for
the given combination of parameters. Spectral Sweep 3 shot Raw Oxy
Raw Oxy 2 .times. 2 Low Gain exception exception 0.48 0.71 High
Gain exception exception saturates saturates 4 .times. 4 Low Gain
14.63 23.08 0.32 0.39 High Gain 9.68 18.75 saturates saturates
[0353] To further illustrate the timing of each process in the
hyperspectral image acquisition sequence, the timing diagram in
FIG. 54 follows the flow of data from initializing acquisition to
visualization of a processed bitmap. The closed loop describes
acquisition of a single slice of a hyperspectral cube, thus the
time required to complete that loop must be multiplied by the
number of slices or illumination spectra in order to calculate
acquisition time. By examining the time log of the GUI during
acquisition, processing time for each algorithm is deduced. This
processing time is subtracted from the output time in Table 5 for
4.times.4 binning and low gain for "3 Shot" (high gain for spectral
sweep) to determine the total acquisition time. Dividing the total
acquisition time by the number of slices equals the time to
complete one cycle of the closed loop. For spectral sweep, the
acquisition time is 116 ms/slice, and for "3 Shot", the acquisition
time is 93 ms/slice. This difference, 23 ms/slice, is less than the
difference in exposure time for the two methods, 38.75 ms/slice,
indicating that initialization of the camera or light source plays
a large factor in acquisition time.
[0354] FIGS. 55A and 55B are diagrams illustrating the processing
algorithms for visualizing the images of blood oxygenation captured
using the spectral sweep and the "3 shot" method of the present
invention. FIG. 55A shows the spectral sweep comparing measured
spectrum to reference spectra via multivariate least squares
analysis to quantify relative concentration HbO.sub.2. FIG. 55B
shows the "3 shot" method which subtracts the image representing Hb
absorbance from the image representing HbO.sub.2 absorbance and
dividing the broadband absorbance to quantify relative
concentration of HbO.sub.2.
[0355] The methods are further described infra. To visualize the
ischemia induced by occluding blood flow to a finger and the
ensuing reactive hyperemia upon removal of the occlusion, a
subject's hand is imaged with the DLP HSI in spectral sweep and "3
Shot" modes (FIGS. 56A and 56B, respectively). The system is warmed
up, focused at a distance of 18 cm, and set to capture a background
cube. The subject's hand was placed palm up in the FOV of the
camera so that three fingers and part of the palm were imaged.
`Control` hyperspectral image cubes, i.e., five spectral sweep
cubes and five "3 Shot" cubes, were acquired of the non-occluded
fingers. Each spectral sweep cube was analyzed with `Oxyz Jet` and
each "3 Shot" cube was analyzed with `3 shot jet (mid)17`
processing algorithms, resulting in five output bitmap images
color-coded for percent HbO.sub.2 for each illumination method.
After acquiring `Control` images, a rubber band is wrapped three
times around the base of the subject's middle finger and five
`Occluded` hyperspectral image cubes for each method were acquired
and processed in the same way as the `Control` images.
[0356] To visualize the real-time progression of reactive hyperemia
after removal of the occlusion, "3 Shot" images were acquired and
processed continuously after acquiring the `Occluded` images. While
the system continued outputting "3 Shot" images color-coded for
percent HbO.sub.2, the rubber band was cut with a pair of scissors.
After a period of three minutes, the system was switched to
spectral sweep mode to collect five `Reperfusion` hyperspectral
image cubes.
[0357] The spectral sweep hyperspectral images were averaged for
each time point in the experiment (n=5, for `Control`, `Occluded`,
and `Reperfusion`). The "3 Shot" output bitmaps are averaged for
`Control` and `Occluded` time points (n=5), and the mean pixel
value for a constant 49 pixel area represents percent HbO.sub.2 at
each time point. For the real-time progression of reactive
hyperemia, the mean pixel value for the 49 pixel area was plotted
versus time for several seconds prior to cutting the rubber band
and 180 seconds immediately thereafter.
[0358] The results are now discussed infra. Visual inspection of
the processed spectral sweep images (See FIG. 57A) showed that the
DLP hyperspectral imaging system can differentiate between
oxygenated and deoxygenated tissue. In FIG. 57A, all three fingers
were colored shades of red, orange, and yellow, which corresponded
to a percent HbO.sub.2 between 60 and 80%, as indicated on the
colorbar at the right. As expected under `Control` conditions,
there was no difference in the surface oxygenation amongst the
three fingers. In FIG. 57B, the middle finger was colored
blue-green, which corresponds to a percent HbO.sub.2 between 40 and
50%, while the other two fingers appeared the same as they did in
the `Control` image. As expected under `Occluded` conditions, the
rubber band tourniquet inhibits blood flow to the finger and
effectively de-oxygenates the tissue. In FIG. 57C, the middle
finger was colored red-orange, corresponding to a percent HbO.sub.2
higher than 75%. As expected under `Reperfusion` conditions, the
DLP HSI visualizes an overshoot in tissue oxygenation related to
reactive hyperemia most likely caused by vascular autoregulation of
the previously occluded finger.
[0359] To quantify this overshoot related to reactive hyperemia,
inspection of the 3 Shot output images was necessary. In FIG. 58A,
the processed "3 Shot" output image under `Occluded` conditions
matched its counterpart spectral sweep image (FIG. 57B). The
non-occluded fingers appeared redder than those in the spectral
sweep image, indicating a higher percent HbO.sub.2 than measured by
the spectral sweep method. The two methods were calibrated further
to determine the truer method of measuring absolute oxygenation.
Because of this discrepancy in absolute measures, all results must
be considered in relative terms rather than in absolute oxygen
percentages.
[0360] Inspecting the "3 Shot" images immediately after the
occlusion is removed and several minutes later (FIGS. 58B and 58C,
respectively), it appeared that there is an immediate overshoot of
percent HbO.sub.2 in the previously occluded middle finger and a
subsequent return to normal tissue oxygenation levels. This
overshoot is quantified by calculating the mean of the pixel values
in the black sample area of each image divided by the average
`Control` values for the same pixel area. With this calculation,
the transient response of reactive hyperemia in the middle finger
was examined. In FIG. 59, the average percent HbO.sub.2 relative to
control for the black sample area is plotted for every "3 Shot"
output image acquired between occlusion and reperfusion. While
occluded, the surface oxygenation was less than 70% of the control
oxygenation, but in the first ten (10) seconds after removing the
tourniquet the surface oxygenation raises to 112% of control.
[0361] Variability in consecutive images was apparent when viewing
the output images in real-time and is also apparent in the wide
vertical spread of the data points in FIG. 59. This image to image
variability is not entirely understood, but may be an artifact of
the OL 490 output variability. Even with the variability, the
overall trend of oxygenation appeared to mimic an under-damped 2nd
order system response. There is an initial overshoot, then
undershoot after about 1.5 minutes, with percent HbO.sub.2
eventually returning to the control level after longer than 3
minutes.
[0362] The DLP HSI can be used to visualize ischemia and reactive
hyperemia, as evidenced by the proof-of-principle finger occlusion
test. Processed Spectral Sweep output images showed high contrast
between oxygenated and de-oxygenated tissue and indicated gross
physiological changes. Processed "3 Shot" output images showed the
same contrast between oxygenated and de-oxygenated tissue, but the
two illumination and processing methods do not give identical
absolute measurements of percent HbO.sub.2.
[0363] Images acquired and processed by the "3 Shot" method were
generated at about 3 frames per second, while spectral sweep images
are generated at about 3 frames per minute. Both are useful methods
for visualizing the spatial distribution of surface tissue
oxygenation, but the "3 Shot" method is preferable when visualizing
short duration physiological changes. Analyzing several minutes of
continuous "3 Shot" outputs shows reactive hyperemia of the surface
tissue in a previously occluded middle finger.
[0364] The following descriptions detail particular examples of
procedures, conditions, etc. which are applicable to utilizing the
present invention "3 Shot" methods.
[0365] Partial nephrectromy is a surgical procedure in which a
section of a patient's kidney is removed, usually to eliminate a
tumor. Sparing the remainder of the kidney often requires vascular
occlusion in order to temporarily interrupt renal blood and prevent
hemorrhage. Two methods of renal vascular occlusion: artery-only
occlusion (AO) and artery and vein occlusion (AV), have
demonstrated differential effects on renal tissue viability during
2 to 24 hour long periods of ischemia. (See Gong, E M, Zorn, K C,
Orvieto. M A, Lucioni, A, Msezane, L P, Shalhav, A L, Artery-only
occlusion may provide superior renal preservation during
laparoscopic partial nephrectomy. Urology 2008; 72:843-846). To
understand the effects of competing clamping methods during shorter
periods of ischemia, the DLP HSI was used to image the renal tissue
oxygenation during partial nephrectomies in pigs and then humans
for periods of ischemia less than one hour long.
[0366] The following describes the results of the study involving
pigs. Following approval by the UTSW Institutional Animal Care and
Use Committee (IACUC), four female Yorkshire pigs, weighing between
60-80 kg, underwent AO or AV occlusion of each kidney. At the
beginning of the study, one kidney was used to perfect the
investigational technique and, therefore, was not included in the
analysis. In order to verify the images obtained by the prototype
DLP HSI, hyperspectral image cubes are acquired by sweeping 126
contiguous bandpasses, identical to the spectral sweep illumination
method, with an already characterized LCTF-based hyperspectral
imaging system. The primary difference between the two systems is
the method of filtering light. In the LCTF system broadband light
is reflected from the tissue sample and then filtered by a liquid
crystal tunable filter (LCTF) before being detected by a CoolSNAP
HQ2 array detector. In the DLP system's spectral sweep method,
narrowband light from the spectral light engine is reflected from
the tissue sample and directly detected by a CoolSNAP HQ2 array
detector. There is no LCTF equivalent to the DLP "3 Shot" method as
a LCTF is incapable of providing complex spectra as provided by the
DLP.
[0367] After anesthetizing each subject according to protocol, the
surgeon approached the right kidney through an open midline
incision and pulled the peritoneum away from the visible surface of
the kidney. While the surgeon exposed the kidney, background cubes
for both illumination methods are acquired in a dark corner of the
surgical suite. The DLP HSI tripod was then abutted to the surgical
bedside (See FIG. 60) and the friction head was adjusted so that
the exposed kidney is fully visible in the camera's FOV. The wheels
of the tripod dolly were locked in place so that all images are
acquired from the same angle and distance. `Control` hyperspectral
image cubes are acquired and processed by spectral sweep and "3
Shot" methods as in the finger occlusion test described above. The
surgeon then dissected the hilem and clamped either the artery only
(AO) or the artery and the vein (AV) with a curved Satinsky
clamp.
[0368] The renal vasculature remained occluded for one hour before
removing the clamp. Spectral sweep `Occluded` image cubes are
acquired and processed in approximately 5 minute intervals
throughout the occlusion. "3 Shot" image cubes were acquired and
processed continuously during application of the clamp, several
times during the hour of occlusion, and continuously during removal
of the clamp. After one minute of reperfusion, spectral sweep
`Reperfusion` image cubes were acquired at 5 minute intervals for
30 minutes. Following ischemia and reperfusion of the right kidney,
the identical procedure was carried out on the left kidney using
the opposite clamping technique. During the procedure, the
subject's rectal temperature was maintained between 37.degree. C.
and 39.degree. C. and its serum oxygenation, as measured by a pulse
oximeter on the animal's ear, was kept between 98% and 100%.
[0369] Percent HbO.sub.2 was measured over a uniform 81 pixel area
chosen from the center of each kidney at each time point during the
study. An average of the pixels for each region was calculated
along with a standard deviation for each image cube. Images from
each 10 minute time frame during ischemia were grouped as a single
dataset in order to determine differences between AV and AO
clamping conditions at each of six time periods. A regression model
was fit to the individual levels using SAS statistical software
(SAS Institute. Cary, N.C.) in order to determine whether there is
a statistically significant difference between clamping methods
with regard to kidney tissue ischemia over a period of one
hour.
[0370] Real-time progression of ischemia and reactive hyperemia was
measured in a 225 pixel area for all of the "3 Shot" output bitmaps
acquired during tightening and removal of the Satinsky clamp in one
pig subject. The mean pixel value in each image's sample area was
plotted as a function of the time the output image was originally
processed. This plot showed the transient behavior of the initial
decline of percent HbO.sub.2 after occlusion and the extent of
reactive hyperemia after the occlusion is removed. It is unknown
whether these immediate transient effects will be different between
AV and AO clamping.
[0371] In the porcine kidney studies, "3 Shot" images appear
similar to spectral sweep images (See FIG. 61). Comparing FIGS. 61C
and 61D to FIGS. 61E and 61F, the relative tissue oxygenation is
the same for both illumination methods. It is obvious that between
`Control` and `Occluded` states there is a drop in surface
oxygenation for the renal unit, but the surrounding muscle and
adipose tissue remains highly oxygenated. The major difference
between spectral sweep and "3 Shot" processed images is the
threshold vector by which the final image is mapped to percent
HbO.sub.2. For the spectral sweep method, the spectral information
stored in the 126 slice data cube allows a wider range of values,
so the background is truly blue, representing virtually no
oxygenation. For the "3 Shot" method, the range of pixel values in
a processed image is much less than that for a spectral sweep
image, so a tighter threshold must be applied to visualize the same
contrast. This tighter threshold causes the mapped percent
HbO.sub.2 values to be slightly different than those mapped in the
spectral sweep images, so the detectable range and resolution of
percent HbO.sub.2 may be less with the "3 Shot" method than with
the spectral sweep method.
[0372] The overall trend of pixel intensities in "3 Shot" output
images is identical to the trend of spectral sweep images and
matches the expected oxygenation trend. Therefore, "3 Shot" output
images are sufficient for mapping the spatial distribution of
tissue oxygenation over time. For the first three pig subjects,
continuous acquisition of "3 Shot" images during application and
removal of the Satinsky clamp was not performed because the primary
concern was capturing quality images to test hour long effects of
clamping. Before imaging the fourth pig subject, data from the
other subjects indicated that a majority of tissue chemistry
changes occur during the first minute after clamping and during the
first few minutes after removing the clamp. So, for the fourth pig
subject, continuous "3 Shot" images are acquired during both
critical time periods for each kidney.
[0373] A 225 pixel area from the center of the kidney is averaged
in each image and divided by the average of the same area for the
first 20 control images to result in the percent HbO.sub.2 relative
to control. In FIG. 62, this value is plotted for the kidney
clamped AO (n=247) and for the kidney clamped AV (n=217) for 12
seconds prior to tightening of the clamp and about one minute
following application of the clamp. For the first 10 seconds of
occlusion both methods of clamping result in a linear decline of
oxygenation. The transient response then becomes non-linear, and
flattens after 30 seconds in both cases. There is a zero-order time
delay between the two responses, with the AO decline lagging behind
the AV decline. This lag may be due to the flow of blood through
the renal vein, which would not occur under AV clamping. There is
also a steady-state difference between the oxygenation levels, with
AO being higher than AV.
[0374] The "3 Shot" outputs acquired during the removal of the
Satinsky clamp are analyzed using the same procedure for the AO
clamped kidney (n=1540) and the AV clamped kidney (n=269). Images
are continuously acquired until it visually appears that there is
no longer any fluctuation in kidney oxygenation levels. From the
number of images in each case, it is apparent that visual
fluctuations cease more quickly in the AV case than in the AO case.
Another interesting observation is that the AO case has an
overshoot of oxygenation associated with reactive hyperemia and
then declines to a steady state lower than its control value, and
lower than the AV steady state. This may be a potential detriment
to clamping the artery only instead of the artery and the vein.
Perhaps, there is a longer term oscillation that is not captured
due to lack of images in the AV case.
[0375] An unanticipated advantage of near real-time "3 Shot" images
showing the surgeon the spatial distribution of tissue oxygenation
is illustrated when clamping fails to cause the entire kidney to
become ischemic. In pig 3, the surgeon clamped the renal
vasculature AO as in previous pig subjects. Raw views of the kidney
seem to imply the clamping is successful (See FIG. 63A), but as the
surgeon watches the real-time "3 Shot" output display in the GUI,
he notices that the upper pole is still highly oxygenated but the
lower pole has become ischemic (See FIG. 63B). This prompts a
re-examination of the renal vasculature, in which the surgeon finds
a second artery supplying oxygenated blood to the renal unit.
Apparently, in this case, the clamped artery supplies blood to the
lower pole, and an unclamped second artery supplies blood to the
upper pole. Upon clamping the second artery, the upper pole also
becomes ischemic, but never reaches the same level of ischemia as
the initially occluded lower pole (See FIG. 63C). Therefore, the
DLP HSI can also be used to determine the precise tissue areas that
are supplied by an artery, which helps surgeons to know which
arteries to occlude during partial nephrectomies.
[0376] Hyperspectral imaging with the DLP HSI is useful for
comparing competing renal vasculature clamping methods with regard
to kidney tissue oxygenation during nephron-sparing surgery.
Analyzing spectral sweep images indicates that the percent
HbO.sub.2 of the kidney is higher when the artery only (AO) is
clamped than when the artery and vein (AV) is clamped between 10
and 40 minutes. Images from other time periods indicate no
difference in percent HbO.sub.2 between AO and AV clamping. The
spectral sweep image cubes captured by the prototype system
visualize the same spatial oxygenation map as other hyperspectral
imaging systems, even though the spectrum measured at each pixel is
not always consistent between systems.
[0377] Generating "3 Shot" output images at 3 frames per second
allows the surgeon to visualize real-time physiologic changes in
tissue oxygenation. This is useful in determining whether or not
blood flow is adequately inhibited from the entire kidney.
Analyzing "3 Shot" images captured during the process of tightening
and releasing the clamp shows the transient response of renal
tissue at the onset of ischemia and during reactive hyperemia.
Though only one subject was sufficiently imaged during this time
period, it appears that AV clamping may be advantageous to AO
clamping when immediate ischemia or no hyperfusion is desired.
[0378] The following discussion is related to similar studies
performed on human subjects. Animal studies are necessary to ensure
the safety and utility of prototype medical devices, but the end
use of the DLP HSI is for human imaging, so its utility in live
human surgeries must also be tested. An Institutional Review Board
(IRB) approved protocol at UTSW permits the prototype hyperspectral
imaging system to image open cavity partial nephrectomies in human
subjects.
[0379] The following methods were used in the human studies. The
surgical procedure for humans is different than that for the
porcine study, because human subject survival is of the utmost
importance. The surgeon makes an incision in the side of the
subject to expose the kidney. `Control` hyperspectral data cubes
are acquired with the DLP HSI in spectral sweep and "3 Shot" modes.
Then, the surgeon clamps the renal vasculature with a predetermined
method (AO or AV) to occlude blood flow to the kidney. `Occluded`
hyperspectral data cubes are acquired at non-critical points in the
surgery, when the surgeon allows all other room lights to be turned
off. After the kidney is sufficiently bereft of blood to prevent
hemorrhage, the surgeon removes the tumor and stitches the exposed
renal tissue closed. One more `Occluded` data cube is acquired
before the clamp was removed. The surgeon removed the clamp, and
`Reperfusion` hyperspectral data cubes are acquired prior to
closing the abdominal cavity.
[0380] To date, three human subjects have been imaged. The
following results section is focused on the first of these human
subjects. The subject is an otherwise healthy male, about 60 years
old, with a tumor on his right kidney.
[0381] In human surgery, the room lights and operating lights
cannot be switched off for minutes at a time as is possible during
animal surgeries. Therefore, only one spectral sweep data cube was
captured as a control to verify the image results of the "3 Shot"
illumination method. Subsequent hyperspectral images were captured
with the "3 Shot" method so that the room and operating lights only
remain off for several seconds.
[0382] View finder pictures of the kidney show the gross anatomy
before and after the tumor is removed, and "3 Shot" image outputs
show the relative oxygenation of the kidney throughout the surgery.
In FIG. 64A, the blue sample area highlights a uniform region of
the kidney tissue. The sample area is used to trace the trend of
oxygenation throughout the surgery.
[0383] Hyperspectral images are acquired at five different time
points during the surgery. One of the images is unreliable due to
light pollution from the headlamp worn by the surgeon, and the
other four images are shown in FIGS. 64B-64E. In the `Control`
image, FIG. 64B, the kidney is highly oxygenated and the spatial
distribution of tissue oxygenation is fairly uniform. In the first
`Occluded` image, FIG. 64C, most of the kidney exhibits lower
oxygenation levels. After the tumor is removed and the clamp is
still tightened, FIG. 64D, the kidney shows decreased oxygenation
levels in all areas. When the clamp is finally removed. FIG. 64E,
the kidney oxygenation levels return to near those measured in the
`Control` image.
[0384] In FIG. 65, the average and standard deviation of the 81
pixel values in each sample area are plotted to show the trend of
oxygenation throughout the surgery. The overall trend implies that
the kidney becomes most ischemic in image three and then recovers
to a level similar to control.
[0385] The DLP HSI can successfully map the spatial distribution of
tissue oxygenation during a human partial nephrectomy. Light
pollution from the operating room lights and the surgeons head lamp
interferes with hyperspectral imaging, so fewer images can be
captured in human cases than in animal cases. However, the "3 Shot"
output images acquired during the human study are useful for the
surgeon to visualize which areas of the kidney are higher and lower
oxygenated. Analysis of many more subjects is needed in order to
compare AO versus AV clamping in human nephron-sparing
surgeries.
[0386] Nerve damage may be correlated to the loss of blood flow in
the surrounding tissue bed. To diagnose nerve damage in the lower
limb of patients, physicians probe the leg with a mechanical or
electrical prick and ask the patient if they can feel it. If the
patient cannot feel the prick, that area is diagnosed as
neuropathic. This process is time consuming, and the spatial
resolution of the diagnosed neuropathy is a function of the number
of discrete point measurements made by the physician. With the DLP
hyperspectral imaging system it may be possible to map the spatial
distribution of neuropathy along the entire lower limb by mapping
its tissue oxygenation.
[0387] The following discussion is related to measurements of lower
limb neuropathy. A clinical study is devised to test the utility of
hyperspectral imaging for monitoring wound care in amputees and for
studying their neuropathy. Amputees and patients exhibiting lower
limb neuropathy are voluntarily enrolled in the clinical study at a
weekly wound care clinic in the Dallas VA. After signing the
consent forms, each subject is seated in a room reserved for
hyperspectral imaging, the overhead lights are turned off, and the
DLP HSI is used to capture spectral sweep and "3 Shot" images along
the entire limb of interest.
[0388] To date, five human subjects have been enrolled in the
study. The following results section is focused on the first of
these human subjects. The subject is a male over 50 years old who
has previously been diagnosed with neuropathy in the lower half of
his left leg, as measured by the prick test.
[0389] Spectral sweep and "3 Shot" hyperspectral image cubes are
acquired from the subject's toes to his knee in the lower limb's
frontal plane. The output images color-coded for percent HbO.sub.2
with the "3 Shot" method shows similar relative oxygenation levels
to the color-coded outputs of the spectral sweep method. Overlaying
the spectral sweep images on a picture of the subject's leg creates
a surface tissue oxygenation map for the entire leg (See FIG.
66).
[0390] Due to the curvature of the ankle, the detected absorbance
images captured between the foot and the shin contain an unreal
intensity gradient and are not shown. Those images unsuccessfully
predicted the spatial map of tissue oxygenation, emphasizing an
important constraint of the prototype: the topography of the tissue
being imaged. Tissue planes further from the camera and light
source optics may appear less oxygenated than closer tissue planes,
when their real oxygenation is the same.
[0391] It is difficult to see any difference in oxygenation from
the bottom half of the leg to the top half of the leg in the images
alone. A sample area of pixel values along the midline of the leg
is plotted versus the axial distance from the toes (See FIG. 67) to
quantify this difference. This plot indicates that there is an
increase in surface tissue oxygenation progressing from the toes to
the knee. However, this increase does not seem to match the results
of the prick test. If surface tissue oxygenation mimics underlying
neuropathy, there should be a jump in oxygenation between the
region of neuropathy and the normal region, but the hyperspectral
images do not show this jump. This emphasizes another important
constraint of the prototype: the depth of the tissue being imaged.
The DLP HSI relies on measuring the absorbance of light in the
visible wavelength range, and visible light does not penetrate deep
into the tissue, so the measured percent HbO.sub.2 is only
indicative of epithelial tissue oxygenation. Epithelial tissue is
highly regulated and its oxygenation levels do not correspond to
neuropathy like deeper tissue oxygenation does. Development of a
prototype that operates in the near infrared wavelength range may
be useful for visualizing deeper tissue oxygenation.
[0392] The DLP HSI is useful for mapping the spatial distribution
of tissue oxygenation for areas that are larger than the field of
view of the camera. Several image cubes are captured in overlapping
regions, and the resulting output images are overlain to create a
large spatial map.
[0393] When acquiring images with the DLP HSI, it is important to
know the topography and depth of the tissue of interest. The tissue
area should be a uniform distance from the system's detector and
light source illumination optics to eliminate unreal effects of
distance on measured percent HbO.sub.2. The tissue of interest
should also be on the surface of the subject (e.g., epithelium),
since visible light does not penetrate past the surface layers of
tissue. In the case presented, epithelial oxygenation does not seem
to correlate exactly with neuropathy, but a larger sample size is
needed to confirm these results.
[0394] The following discussion is related to measurements taken
during brain surgery. Neurosurgeons investigate brain behavior by
dissecting the scalp and viewing the cerebral cortex with surgical
microscopes. Current imaging only visualizes the physical anatomy
of the cortex, or by introducing dyes as contrast agents some
functionality can be visualized. Point measures of cerebral
HbO.sub.2 content are possible with near infrared spectroscopic
techniques, but they do not show a spatial image of oxygenation.
(See Quaresima, V, Sacco, S, Totaro, R, Ferrari, M, Noninvasive
measurement of cerebral hemoglobin oxygen saturation using two near
infrared spectroscopy approaches. 2000; 5:201-205.) By coupling the
DLP hyperspectral imaging system to a surgical microscope,
neurosurgeons may be able to visualize the spatial profile of
cerebral tissue chemistry.
[0395] The CoolSNAP HQ2 is directly mounted to the C-mount of a
Zeiss surgical microscope at the UTSW animal laboratory, and the OL
490 liquid light guide is coupled to the microscope's ring
illumination optics. After setting the focal distance of the
microscopic system, a background cube is captured with the DLP HSI
and the surgeon dissects a small cross-section of an anesthetized
rabbit's scalp. Imaging through the optics of the microscope, the
DLP HSI is set to continuously acquire "3 Shot" images of the
exposed cortical tissue. Several seconds of control images are
captured while an external pump supplies oxygen to the subject.
Continuous "3 Shot" images are acquired while the external oxygen
supply for the subject is cut off for a period of 20 minutes and
turned back on.
[0396] Watching the spatial distribution of tissue oxygenation, and
drawing from his own experience, the surgeon notices one area of
the exposed cortex is damaged at the beginning of the imaging
session. In FIG. 68A, the central region is the cortical tissue and
the three red regions around the circumference are scalp and other
tissues. The black sample box surrounds an area of healthy brain
tissue and the blue sample box surrounds an area of damaged brain
tissue. In the control image, the damaged tissue appears ischemic
compared to the normal tissue's level of oxygenation. There are no
noticeable changes in the damaged tissue oxygen levels shown in the
hyperspectral images during the period of external oxygen cutoff.
However, after the oxygen is cut off for several minutes, the
normal brain tissue begins to show ischemia as indicated by the
lower percent HbO.sub.2 in the black sample area of FIG. 68B.
[0397] Initial brain imaging with the DLP HSI coupled to a Zeiss
surgical microscope suggests neurosurgery is another viable
application for this novel medical imaging platform. The digital
visualizations of oxygenation generated by hyperspectral imaging
correspond with what the surgeon expects to see. Even though
conclusive medical results cannot be derived from the initial
imaging session, the results indicate that DLP HSI can enable
surgeons to visualize cerebral tissue chemistry in near
real-time.
[0398] In view of the foregoing description of the present
invention, it should be appreciated that a variety of procedures
and diagnostic techniques may be improved by incorporation of the
present invention. The following discussion describes some of such
procedures and diagnostics techniques. This discussion is not
intended to limit the scope of the invention as set forth in the
appended set of claims.
[0399] There are a variety of applications of hyperspectral imaging
of the retina. For example, measuring oxygenation of the retina as
may be effected by diabetic retinopathy, central retinal vein
occlusion, branch retinal vein occlusion, central retinal artery
occlusion, branch retinal artery occlusion, sickle cell
retinopathy, retinopathy of prematurity and retinal vascular
inflammatory diseases. Diabetic retinopathy is the most common
cause of new cases of blindness in adults 20-74 years of age.
Retinal vascular occlusions, sickle cell retinopathy and
retinopathy of prematurity can also lead to irreversible severe
visual loss. These conditions often lead to retinal ischemia and
subsequent retinal neovascularization with the associated blinding
complications of retinal detachment, vitreous hemorrhage and
neovascular glaucoma. Laser photocoagulation of the peripheral
retina can markedly reduce the risk of these complications.
Targeting the areas of retina that are suffering from ischemia
improves the clinical outcomes while reducing peripheral visual
field loss. Furthermore, these conditions can also lead to macular
edema (i.e., swelling of the central retina) and macular ischemia.
Distinguishing between these two causes of visual loss helps to
target patients that would benefit from light macular laser therapy
versus pharmacologic interventions. Traditionally, fluorescein
angiography (FA) has been used to evaluate for both peripheral
retinal ischemia and macular ischemia. However, FA is painful, has
risks associated with intravenous injections, leads to
discoloration of the skin for twenty four (24) hours, can lead to
allergic reactions of all degrees of severity, often leads to
nausea and vomiting, and may be impossible to perform in patients
with poor vein access. Furthermore, the technique evaluates
vascular flow and permeability rather than retinal oxygenation. It
is believed that hyperspectral imaging of the retina measuring
oxyhemoglobin and deoxyhemoglobin may allow for earlier
identification of areas of ischemia before permanent loss of
capillary perfusion has occurred. It also would allow for a
non-invasive determination of retinal oxygenation. This would speed
up and simplify the process of obtaining this clinically important
piece of information, while reducing the risk of complications and
eliminating the associated pain and discomfort.
[0400] The present invention hyperspectral imaging system may also
be used for measuring optic nerve oxygenation. Such information may
be useful in evaluating such conditions as glaucoma, ischemic optic
neuropathy and optic neuritis. It is believed that assessment of
optic nerve head oxygenation may help diagnose patients with
increased risk for glaucoma or ischemic optic neuropathy. It may
also help differentiate between patients with ischemic optic
neuropathy versus optic neuritis.
[0401] The present invention may also be used for measuring macular
pigments thereby providing information regarding age-related
macular degeneration and juxtafoveal telangiectasis. Age-related
macular degeneration (AMD) is the most common cause of irreversible
blindness in patients over sixty (60) years of age. Despite
advances in therapeutics, clinical outcomes are still suboptimal.
There is no effective therapy for atrophic AMD. Macular pigments
including lutein and zeaxanthin are pigments that are present in
the retina in the macular region. They absorb particularly harmful
wavelengths of light (i.e., blue light) and protect the retina from
phototoxicity. These pigments have been shown to decrease with age
and more rapidly in patients with AMD and juxtafoveal
telangiectasis. It is believed that the present invention provides
an easy modality to measure the levels of macular pigments which in
turn may help to identify patients that would benefit from dietary
supplements with lutein before they develop clinical disease.
[0402] The present invention may also be used for measuring
pigments in the retinal photoreceptors and retinal pigment
epithelium thereby providing information related to retinitis
pigmentosa and other hereditary retinal degenerations, e.g.,
age-related macular degeneration. Hereditary retinal degenerations
and age-related macular degeneration lead to abnormalities in the
photoreceptors and underlying retinal pigment epithelium (RPE). The
health of the photoreceptors and the retinal pigment epithelium can
be correlated to the levels of pigments (i.e., rhodopsin,
lipofuscin and melanin). It is believed that the present invention
will permit measuring these pigments accurately thereby allowing
better monitoring of the clinical course in these patients.
Moreover, it is believed that the present invention will also allow
measurement of the response to prophylactic interventions. Finally,
it is believed that the present invention may aid in the early
diagnosis of these conditions or subclassification into clinically
relevant subgroups due to its ability to obtain the foregoing
measurements.
[0403] The present invention may also be used for the diagnosis of
autoimmune retinitis, infectious retinitis and infiltrative
neoplastic conditions. There are a limited number of physical
changes that the retina can undergo in response to injury. The
clinical presentation of retinal autoimmune infiltration (e.g.,
Wegener's granulomatosis, sarcoidosis, etc.), infectious retinitis
(e.g., cytomegalovirus, toxoplasmosis and herpetic retinitis as in
acute retinal necrosis or progressive outer retinal necrosis), and
neoplastic infiltration (e.g. lymphoma and leukemia) can sometimes
be indistinguishable from one another. In those cases, surgical
intervention to obtain vitreous and/or retinal tissue for
diagnostic purposes is often necessary. There are significant risks
associated with this procedure. However, lack of therapy would
often lead to blindness or loss of the eye. Furthermore, an
erroneous diagnosis leading to an erroneous therapeutic
intervention can precipitate acute exacerbation of the disease
(e.g., treating a viral infection with steroids) or delayed
diagnosis (e.g., anti-inflammatory therapy for a neoplastic
condition) resulting in loss of vision, loss of the eye or
progression of a life-threatening disease. It is believed that the
present invention hyperspectral imaging system and methods can look
for spectral signatures of pathogens thereby helping to ascertain a
diagnosis in a non-invasive and rapid way.
[0404] The present invention may also be used for non-invasive
evaluation of disease biomarkers. It has been said that the eye is
a window to the body. The eye provides the clinician with an
unobstructed view of vessels. This has for years allowed medical
professionals to diagnose systemic conditions like diabetic and
hypertensive retinopathy. It is believed that hyperspectral imaging
will allow medical professionals to tap deeper into this invaluable
resource. Not only can the spectrum of substances in the visible
light spectrum be identified like oxyhemoglobin and
deoxyhemoglobin. The spectral signatures of a wide range of
biomarkers both in the visible and invisible light spectrum can
also be identified. This will allow medical professionals to
diagnose and non-invasively monitor a wide range of conditions,
including but limited to depth of anesthesia during surgical cases,
systemic intoxications, drug overdose, pancreatitis, liver disease,
prostate and other cancers with known biomarkers, metabolic
diseases, infectious diseases with pathogens that produce specific
spectral signatures, etc. Furthermore, it is believed that it will
allow medical professionals to discover new spectral signatures
that correlate with common and uncommon diseases and that could
serve as new biomarkers in the diagnosis and management of those
conditions.
[0405] Moreover, the present invention may also be used to monitor
the clinical course of battlefield blast injuries based on retinal
vasculature diameter changes. It should be appreciated that the
foregoing description of ophthalmological applications are but some
of the procedures and diagnostic methods that can be improved by
the incorporation of the present invention hyperspectral imaging
system and methods. For example, as discussed supra, real-time
diagnostic measurements of the kidneys provides surgeons with never
before available information to use during surgery.
[0406] Surgery is the mainstay treatment for patients diagnosed
with kidney cancer. As such, removal of the tumor by either partial
or total nephrectomy will inherently decrease renal function and
risk the development or progression of chronic kidney disease.
Chronic kidney disease (CKD) is prevalent in 11% of the adult
population, including 17% of patients older than 60 years. It is
associated with an increased risk of hospitalization,
cardiovascular event, and mortality from any cause. In fact,
patients with moderate kidney disease (CKD, stage 3--GFR<60
ml/min/1.73 m.sup.2) have a 40% increased risk of cardiovascular
events and 20% increased risk of death. As such, for patients who
have been diagnosed with kidney cancer, most of which are over 60
years old and require surgery, preserving as much renal function as
possible is the standard of care and may prevent cardiovascular
morbidity and mortality.
[0407] As approximately 80% of renal tumors identified are small
(i.e., Stage T1 which are typically <7 cm), kidney sparing
surgery (e.g., partial nephrectomy) is the preferred technique to
maximally preserve functioning kidney tissue. However, to
comfortably and safely perform this procedure, the urologic surgeon
routinely clamps the renal artery to stop blood flow to the kidney
(i.e., ischemia). This improves visualization and minimizes blood
loss. However, the resulting ischemia risks kidney injury so that
surgeons often will cover the kidney with ice slush for 7-10
minutes prior to tumor excision in an effort to decrease kidney
metabolism which is highly oxygen dependent (i.e., aerobic
metabolism). Nevertheless, warm ischemia (i.e., no ice) of more
than 20 minutes, and cold ischemia of more than 30 minutes can
result in irreversible damage in the remaining kidney. Therefore,
despite the good intent of preserving as much kidney as possible
during cancer surgery, the patient and surgeon still risk decreased
kidney function secondary to the surgical technique.
[0408] Thus, if partial nephrectomy can be performed with only
partial occlusion of the renal artery such that the kidney still
gets oxygen the kidney is less likely to suffer from irreversible
kidney damage. Unfortunately, there is no guideline as to what
degree should the artery be occluded and, more importantly, what is
the minimum oxygen necessary to protect the kidney. This is
critical information since incomplete artery occlusion could
increase blood loss and diminish visualization during surgery.
Although the ability to monitor tissue oxygen levels in real-time
is now available using tissue probes that are often used in
neurosurgery, the present invention hyperspectral imaging system
and methods provides a viable and effective alternative to these
probes. Additionally, the present invention, unlike the probes,
provides such information in a non-invasive manner. As described
above in view of the porcine and human studies, using the present
invention which includes advanced camera technology and novel
software, light reflected off the surface of an imaged tissue or
organ can be used to measure oxygen levels. This is an improvement
over existing tissue oxygen monitors which must use a needle probe
and can only measure oxygen at one location. Incorporation of
oxygenation and blood flow monitoring technology in partial
nephrectomy can warn surgeons when oxygen levels are dangerously
low and perhaps guide them in how long and to what degree the renal
artery is partially or completely clamped. Recent laboratory data
gathered using an animal model has demonstrated that 50% occlusion
of the renal artery maintained kidney oxygen levels longer and at
higher values during partial nephrectomy. Though blood loss was
increased slightly, it did not increase operative time or surgery
difficulty. Most importantly, when compared to conventional partial
nephrectomy, the cases where tissue oxygen levels were higher (e.g.
50% artery occlusion) had less immediate and sustained kidney
dysfunction after surgery. In view of the foregoing, it is believed
and shown in part above that the present invention hyperspectral
imaging system and methods can be used to monitor the vascular
blood flow and monitor tissue oxygen in clinical use and determine
if kidney function can be protected and improved when performing
partial nephrectomy with real-time monitoring of kidney oxygen
levels.
[0409] The present invention will thus ensure that the recovery of
kidney function after partial nephrectomy will be improved by
real-time monitoring and minimization of the ischemic insult
routinely incurred. This present invention has the potential to
alter the way kidney surgery for cancer is performed. The
preservation of renal function during and after partial nephrectomy
would be incrementally improved benefiting many patients with
compromised kidney function.
[0410] The present invention may also be used to monitor, analyze
and diagnose blast brain injury (BBI) and traumatic brain injuries
(TBI). Overpressure brain injury is one of the most common injuries
experienced by soldiers. Soldiers are exposed to various levels of
blast pressure ranging from relatively low pressure that occurs
from firing of weapons systems to very high pressures that occur
from explosions of IEDs and mortar rounds. Damaging pressure waves
are magnified by reflection from adjacent structures or the ground
resulting in exposure of human tissues to extremely high pressure.
Pressure can also be transmitted to the brain via the thorax when
body armor is struck by high velocity projectiles.
[0411] Blast pressures can cause neural injury by several
mechanisms. Pressure waves can cause substantial pulmonary injury
with alveolar disruption. High pressure air injected into the blood
can cause air emboli resulting in brain ischemia. Brain tissues can
be injured directly from pressure waves passing through the cranial
vault. High pressure can also be transmitted to the brain through
the blood when high pressure occurs in the thoracic cavity. Lastly,
brain function may be depressed by circulating metabolites and
cytokines that are released into the blood stream by non-neural
tissues that are injured by the blast wave.
[0412] Monitoring the severity and progression of blast-induced
brain injury is critical to ensuring optimal care for wounded
soldiers. The present invention can be used in the battlefield to
objectively assess the degree of these injuries to facilitate
appropriate triage of soldiers exposed to explosions. The present
invention can scan the retina to detect appropriate biomarkers
associated with the brain injury and determine the extent of
injury.
[0413] The present invention may take one of several embodiments
for performing the foregoing analysis. A direct fundoscope will
enable imaging of the retina to examine for retinal hemorrhage.
This system may be augmented to wirelessly connect to trained
personnel in any location in the world who can interpret these
images. Alternatively, hyperspectral measurements of retinal
oxygenation may be taken with the present invention. It is known
that retinal hyperemia is associated with blast injury and
therefore overall and regional differences in retinal oxygenation
can be measured. Moreover, measurement of air emboli may be taken
with the present invention. Air emboli have been implicated as a
cause for blast injury associated brain injury. Using the present
invention, air bubbles in the retinal arterial blood supply may be
identified and quantified. Further, it is believed that
hyperspectral images of corneal surface blood vessels provide
another means for analyzing traumatic brain injury.
[0414] Lastly, the present invention may be used to measure blood
chemistry. Blast injuries cause widespread tissue damage resulting
in elevated serum levels of leukotrienes C4, D4 and E4 and of
6-keto-PGF alpha and TxB2, the stable products of prostacyclin
(PGI2) and thromboxane A2 (TxA2). These substances are elevated to
a much greater extent in blast injury than blunt head trauma
suggesting that these and other chemicals may serve as specific
markers for the presence, extent and progress of blast injury to
the brain. The eye can serve as a window to the microcirculation.
Using visible spectroscopy, chemicals in the blood will be detected
and quantified in the retinal microcirculation. It is believed that
the chemical environment of the retinal microcirculation correlates
to the magnitude of brain dysfunction experienced after blast
injury. This complex chemical environment will be manifested as
complex spectra that will be unique for the blood chemical
environment associated with various stages of blast injury. It is
further believed that the statistical classification of these
spectra will enable the diagnosis and characterization of various
stages of blast injury.
[0415] Further diagnosis may be performed by a near infrared
version of the present invention. Such a system will enable
chemometric assessment of tissues as deep as 1 centimeter and
provide more informative spectral information than exists in the
visible light range. This system will enable imaging of the retina
and may provide information about blood flow changes in the face
that occur as a result of blast injury. In a further embodiment,
the present invention may include an infrared imaging system
operating in the thermal range (5-15 micron) that will facilitate
tissue spectroscopy in the wavelength range where molecular
vibrations attenuate infrared energy. Attenuation of black body
radiation by molecular vibration will enable development of
scanning devices that will detect blood and tissue chemical
composition with a high degree of discrimination for various
disease states. It is believed that minor changes in blood
chemistry associated with various disease states will result in
highly unique spectral signatures in this region of the
electromagnetic spectrum.
[0416] The foregoing measurements may be obtained by combining a
conventional fundus camera and the present invention hyperspectral
imaging illumination system.
[0417] A further embodiment of the present invention provides a
convenient means for threat detection. The present invention
hyperspectral imaging system provides for the detection of
pathogens and poisons on surfaces of equipment, assembly lines,
etc. For example, the present invention may be used to screen a
food supply for toxic agents, e.g., salmonella (which has a known
strain-specific IR spectra), anthrax, botulina, small pox, E. coli,
shigella. etc. Further the present invention may be used to detect
various chemicals, such as, organophosphates pesticides, ricin
toxin, sarin, soman, tabun or VX gas.
[0418] A novel hyperspectral imaging system has been developed for
real-time, non-invasive, in vivo imaging of tissue chemistry in
surgical and clinical applications.
[0419] Using a spectral light engine to illuminate with simple
narrow bandpasses or complex broadband spectra and a CCD FPA to
detect the reflectance image from a tissue sample, the DLP
hyperspectral imaging system remotely visualizes tissue oxygenation
in live human and animal subjects. Illuminating with many narrow
bandpasses in the spectral sweep mode results in large data cubes
with full spectral information for each pixel in the image, but
requires 20 to 30 seconds to generate a single chemically-encoded
image. Spectral sweep data cubes acquired with the DLP HSI match
those captured by LCTF-based hyperspectral imagers, which, along
with color tile calibration results, validates the DLP HSI
prototype as an imaging spectrophotometer. Illuminating with only
three complex broadband illuminations in the present invention "3
Shot" mode results in small data cubes with adequate spectral
information to generate chemically-encoded images identical to the
spectral sweep images at a much faster rate, i.e., 3 processed
images per second. Typically, the wavelengths to be used depend on
the tissue type due to the penetration of tissue depth. For
example, in visible spectra, HbO.sub.2, Hb, HbCO.sub.2 can be
measured. In near-infrared (NIR) spectra, HbO.sub.2, Hb, water and
lipids can be measured. As shown in FIGS. 69A through 69C, the
present invention can provide chemically encoded images. FIG. 69A
shows an image of a human hand which has experienced a burn. FIG.
69B shows a hyperspectral image of the hand from FIG. 69A
displaying chemically encoded data for relative contribution of
oxyhemoglobin in the hand. Contrarily. FIG. 69C shows a
hyperspectral image of the hand from FIG. 69A displaying chemically
encoded data for relative contribution of water in the hand. Thus,
it can be seen from these figures that the present invention is
capable of displaying a variety of chemical encoding data.
[0420] A hyperspectral imaging system integrating DLP technology to
illuminate the area of interest with band pass of light ranging
over a spectral range from the visible to the near-infrared that is
interfaced with a digital focal plane array for acquisitioning and
storing a series of digital spectroscopic images (a hyperspectral
data cube) that is analyzed using chemometric methods providing
chemical information within the area of interest aiding the
clinician in detecting, diagnosing and monitoring disease.
Clinicians could also monitor patients with this technology for
determining the best pharmacological therapy and maximize favorable
patient outcomes. The present invention utilizes DLP technology
with a focal plane array and chemometric deconvolution for
visualizing the biochemical levels within a human non-invasively
for clinical and surgical applications. The present invention can
also be illuminated with spectrum of light.
[0421] As shown in FIG. 70A, a single column 411 of mirrors may be
positioned in an "on" position while the remaining columns are
positioned in an "off" position. The resulting spectrum is shown in
FIG. 70B. Similarly, as shown in FIG. 71A, a single column 416 of
mirrors may be positioned in an "on" position while the remaining
columns are positioned in an "off" position. The resulting spectrum
is shown in FIG. 71B. It is apparent to one of ordinary skill in
the art that the selection of individual columns of mirrors can
control the output spectrum of the present invention, and by
varying the selected column, such spectrum can be shifted within
the range of the present invention, i.e., the difference of the
positions of the spectrum in FIGS. 70B and 71B. Furthermore, unlike
all other known devices, the present invention can also provide
complex spectra for illuminating a target. For example, as shown in
FIG. 72A, various columns of mirrors 411, 412, 413, 414, 416, 417
and 418 can be positioned in an "on" position while column 415 is
positioned in an "off" position thereby resulting in a complex
spectrum. Moreover, a different number of mirrors within a
particular column can be positioned in an "on" position thereby
providing a different intensity of a particular wavelength of
light. Thus, the complex spectrum created by the mirror positioned
depicted in FIG. 72A is shown in FIG. 72B.
[0422] The DLP.RTM. hyperspectral imaging and chemometric
deconvolution can be applied to all products that use light, for
example, the fields of clinical endoscopy, clinical chemistry,
microscopy, surgical microscopy, drug discovery, microarray
scanners and microplate readers.
[0423] It should be appreciated that software included in some
embodiments of the present invention can perform a variety of
functions. For example, the software can provide a user interface
through which a user can control all aspects of illumination,
measurement and analysis. Thus, the software can provide full
control of all hardware devices. Moreover, the software can perform
mathematical functions based on internally programmed or embedded
functions or can call math functions from outside software, e.g.,
MatLab.
[0424] It should also be appreciated that the present invention can
be used to compare hyperspectral images of a target of interest,
e.g., a patient's kidney, with images of know targets to evaluate
the condition of the target of interest. For example, a
hyperspectral image of a kidney can be obtained with the present
invention and then compared against a library of kidney images
including but limited to normal kidneys, ischemic kidneys, cancer
containing kidneys, etc. Such comparison allows a medical
professional to evaluate the condition/state of a target by
comparison to an image library of similar targets in various
conditions. Thus, a protocol for repairing/healing the target can
be easily determined based on the closest match from the image
library.
[0425] It should be appreciated that the range of usable
wavelengths for the present invention is limited when using
currently available digital micromirror devices, i.e., the physical
size of each respective mirror element and the materials and
coatings used on the front surface windows which protect the DMD
limit the range of wavelengths. For example, the size of the
mirrors limits what wavelengths will properly reflect from their
surfaces, while the window materials and coatings limit what
wavelengths of light can pass therethrough. The present invention,
as claimed, is not limited by such DMD characteristics and
therefore as new DMDs become available, they will be directly
applicable to the present invention.
[0426] It is contemplated that any embodiment discussed in this
specification can be implemented with respect to any method, kit,
reagent, or composition of the invention, and vice versa.
Furthermore, compositions of the invention can be used to achieve
methods of the invention.
[0427] It will be understood that particular embodiments described
herein are shown by way of illustration and not as limitations of
the invention. The principal features of this invention can be
employed in various embodiments without departing from the scope of
the invention. Those skilled in the art will recognize, or be able
to ascertain using no more than routine experimentation, numerous
equivalents to the specific procedures described herein. Such
equivalents are considered to be within the scope of this invention
and are covered by the claims.
[0428] All publications and patent applications mentioned in the
specification are indicative of the level of skill of those skilled
in the art to which this invention pertains. All publications and
patent applications are herein incorporated by reference to the
same extent as if each individual publication or patent application
was specifically and individually indicated to be incorporated by
reference.
[0429] The use of the word "a" or "an" when used in conjunction
with the term "comprising" in the claims and/or the specification
may mean "one," but it is also consistent with the meaning of "one
or more," "at least one," and "one or more than one." The use of
the term "or" in the claims is used to mean "and/or" unless
explicitly indicated to refer to alternatives only or the
alternatives are mutually exclusive, although the disclosure
supports a definition that refers to only alternatives and
"and/or." Throughout this application, the term "about" is used to
indicate that a value includes the inherent variation of error for
the device, the method being employed to determine the value, or
the variation that exists among the study subjects.
[0430] As used in this specification and claim(s), the words
"comprising" (and any form of comprising, such as "comprise" and
"comprises"). "having" (and any form of having, such as "have" and
"has"), "including" (and any form of including, such as "includes"
and "include") or "containing" (and any form of containing, such as
"contains" and "contain") are inclusive or open-ended and do not
exclude additional, unrecited elements or method steps.
[0431] The term "or combinations thereof" as used herein refers to
all permutations and combinations of the listed items preceding the
term. For example, "A, B, C, or combinations thereof" is intended
to include at least one of: A, B, C, AB, AC, BC, or ABC, and if
order is important in a particular context, also BA, CA, CB, CBA,
BCA, ACB, BAC, or CAB. Continuing with this example, expressly
included are combinations that contain repeats of one or more item
or term, such as BB, AAA, MB, BBC, AAABCCCC, CBBAAA. CABABB, and so
forth. The skilled artisan will understand that typically there is
no limit on the number of items or terms in any combination, unless
otherwise apparent from the context.
[0432] All of the compositions and/or methods disclosed and claimed
herein can be made and executed without undue experimentation in
light of the present disclosure. While the compositions and methods
of this invention have been described in terms of preferred
embodiments, it will be apparent to those of skill in the art that
variations may be applied to the compositions and/or methods and in
the steps or in the sequence of steps of the method described
herein without departing from the concept, spirit and scope of the
invention. All such similar substitutes and modifications apparent
to those skilled in the art are deemed to be within the spirit,
scope and concept of the invention as defined by the appended
claims.
[0433] Thus, it is seen that the objects of the present invention
are efficiently obtained, although modifications and changes to the
invention should be readily apparent to those having ordinary skill
in the art, which modifications are intended to be within the
spirit and scope of the invention as claimed. It also is understood
that the foregoing description is illustrative of the present
invention and should not be considered as limiting. Therefore,
other embodiments of the present invention are possible without
departing from the spirit and scope of the present invention.
* * * * *