U.S. patent application number 14/181025 was filed with the patent office on 2014-06-12 for iterative reconstruction.
This patent application is currently assigned to General Electric Company. The applicant listed for this patent is General Electric Company, Purdue Research Foundation, University of Notre Dame du Lac. Invention is credited to Charles Addison Bouman, JR., Bruno Kristiaan Bernard De Man, Jiang Hsieh, Ken David Sauer, Jean-Baptiste Thibault, Zhou Yu, Kai Zeng.
Application Number | 20140161340 14/181025 |
Document ID | / |
Family ID | 43898495 |
Filed Date | 2014-06-12 |
United States Patent
Application |
20140161340 |
Kind Code |
A1 |
Zeng; Kai ; et al. |
June 12, 2014 |
ITERATIVE RECONSTRUCTION
Abstract
An improved iterative reconstruction method to reconstruct a
first image includes generating an imaging beam, receiving said
imaging beam on a detector array, generating projection data based
on said imaging beams received by said detector array, providing
said projection data to an image reconstructor, enlarging one of a
plurality of voxels and a plurality of detectors of the provided
projection data, reconstructing portions of the first image with
the plurality of enlarged voxels or detectors, and iteratively
reconstructing the portions of the first image to create a
reconstructed image.
Inventors: |
Zeng; Kai; (Rexford, NY)
; Bouman, JR.; Charles Addison; (West Lafayette, IN)
; De Man; Bruno Kristiaan Bernard; (Clifton Park, NY)
; Hsieh; Jiang; (Brookfield, WI) ; Sauer; Ken
David; (South Bend, IN) ; Thibault;
Jean-Baptiste; (Milwaukee, WI) ; Yu; Zhou;
(Palatine, IL) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
General Electric Company
Purdue Research Foundation
University of Notre Dame du Lac |
Schenectady
West Lafayette
Notre Dame |
NY
IN
IN |
US
US
US |
|
|
Assignee: |
General Electric Company
Schenectady
NY
Purdue Research Foundation
West Lafayette
IN
University of Notre Dame du Lac
Notre Dame
IN
|
Family ID: |
43898495 |
Appl. No.: |
14/181025 |
Filed: |
February 14, 2014 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
12607309 |
Oct 28, 2009 |
8655033 |
|
|
14181025 |
|
|
|
|
Current U.S.
Class: |
382/131 |
Current CPC
Class: |
G06T 2211/424 20130101;
G06T 11/008 20130101; A61B 6/5258 20130101; G06T 11/006
20130101 |
Class at
Publication: |
382/131 |
International
Class: |
A61B 6/00 20060101
A61B006/00; G06T 11/00 20060101 G06T011/00 |
Claims
1. An improved iterative reconstruction method to reconstruct a
first image, the method comprising: generating an imaging beam;
receiving said imaging beam on a detector array; generating
projection data based on said imaging beams received by said
detector array; providing said projection data to an image
reconstructor; enlarging one of a plurality of voxels and a
plurality of detectors of the provided projection data, wherein
enlarging the plurality of voxels comprises modeling the plurality
of voxels with an overlap between neighboring voxels, and wherein
enlarging the plurality of detectors comprises modeling the
plurality of detectors with an overlap between neighboring
detectors; and iteratively reconstructing portions of the first
image with the plurality of enlarged voxels or detectors to create
a reconstructed image.
2. The method of claim 1, further comprising outputting the
iteratively reconstructed first image; and wherein enlarging the
plurality of detectors comprises modeling the plurality of
detectors with a detector aperture larger than a physical aperture
of the plurality of detectors.
3. The method of claim 1, further comprising applying a band
suppression model to the reconstructed image.
4. The method of claim 1, further comprising generating trapezoidal
footprints for the plurality of enlarged voxels or detectors.
5. The method of claim 4, wherein the trapezoidal footprints are
generated such that overlap of the plurality of enlarged voxels or
detectors is linear and minimizes peaks within a forward projection
model of the plurality of enlarged voxels or detectors.
6. The method of claim 1, further comprising implementing
non-uniform voxel sizes for the plurality of voxels.
7. The method of claim 6, wherein the non-uniform voxels include
smaller voxels near high-frequency regions of the projection
data.
8. The method of claim 6, wherein the non-uniform voxels include
smaller voxels at edge regions of an object represented within the
projection data.
9. The method of claim 1, further comprising wobbling a focal spot
of the imaging beam.
10. The method of claim 1, further comprising sinogram
pre-processing of the provided projection data, wherein the
pre-processed projection data is used for voxel or detector model
enlargement.
11. The method of claim 1, further comprising modeling the
plurality of detectors with a footprint geometrically dissimilar to
an actual physical shape of the plurality of detectors.
12. The method of claim 1, wherein enlarging the plurality of
voxels comprises modeling the plurality of voxels with a voxel size
larger than a physical distance between neighboring voxels; and
wherein enlarging the plurality of detectors comprises modeling the
plurality of detectors with a detector size larger than a physical
distance between neighboring detectors.
13. An imaging system comprising: a source constructed to project
an imaging beam toward an object; a detector array positioned to
receive said imaging beam and generate projection data; a
translatable table configured for disposal of said object thereon;
and an image reconstructor electrically coupled to said detector
array, said image reconstructor having a processor programmed to:
generate an enlarged voxel/detector model to represent a plurality
of voxels and a plurality of detectors of the projection data,
wherein the enlarged voxel/detector model defines an overlap
between at least one of adjacent voxels and adjacent detectors; and
iteratively reconstruct portions of the first image using the
enlarged voxel/detector model to create a reconstructed image.
14. The imaging system of claim 13, wherein the processor is
further programmed to apply a band suppression model to the
reconstructed image.
15. The imaging system of claim 13, wherein the processor is
further programmed to generate trapezoidal footprints for the
plurality of enlarged voxels or detectors; and wherein the
trapezoidal footprints are generated such that overlap of the
plurality of enlarged voxels or detectors is linear and minimizes
peaks within a forward projection model of the plurality of
enlarged voxels or detectors.
16. The imaging system of claim 13, wherein the processor is
further programmed to implement non-uniform voxels for the
plurality of voxels; and wherein the non-uniform voxels include
smaller voxels near high-frequency regions of the projection data
or at edge regions of an object represented within the projection
data.
17. A method for iteratively reconstructing an image comprising:
accessing projection data acquired from a detector array having a
plurality of detector elements; providing the projection data to an
image reconstructor; modeling a first plurality of voxels of the
projection data with a first voxel size; modeling a second
plurality of voxels of the projection data with a second voxel
size; and iteratively reconstructing the first plurality of voxels
and the second plurality of voxels to generate a reconstructed
image using the image reconstructor; and wherein the first voxel
size is smaller than the second voxel size.
18. The method for iteratively reconstructing an image of claim 17
further comprising selecting the first and second pluralities of
voxels as a function of at least one of location and anatomical
content of the projection data.
19. The method for iteratively reconstructing an image of claim 17
further comprising: selecting the first plurality of voxels from
one of a high frequency region of the projection data and an edge
region of the projection data; and selecting the second plurality
of voxels from a region outside the one of the high frequency
region and the edge region.
20. The method for iteratively reconstructing an image of claim 17
further comprising modeling at least one of the first plurality of
voxels and the second plurality of voxels with a larger voxel size
than an actual inter-voxel distance.
Description
CROSS REFERENCE TO RELATED APPLICATION
[0001] The present application is a continuation of and claims
priority to U.S. Ser. No. 12/607,309, filed Oct. 28, 2009, the
disclosure of which is incorporated herein by reference.
BACKGROUND OF THE INVENTION
[0002] The subject matter disclosed herein relates to medical
imaging, and more particularly, to improved iterative
reconstruction methodologies in medical imaging.
[0003] A computed tomography (CT) imaging system typically includes
an imaging beam source (e.g., x-ray source or other suitable
source) that projects fan- or cone-shaped imaging beams through an
object being imaged, such as a patient, to an array of radiation
detectors. The beam is collimated to lie within an X-Y plane, or to
cover a set of such planes generally referred to as the "imaging
planes." Intensity of radiation from the beam received at the
detector array depends on attenuation of the imaging beam by the
object. Attenuation measurements from each detector are acquired
separately to produce a transmission profile.
[0004] The imaging beam source and the detector array are rotated
within a gantry and around the object to be imaged so that a
projection angle at which the imaging beam intersects the object
constantly changes. A group of imaging beam attenuation
measurements (such as integral projection data from the detector
array at one gantry angle) is referred to as a "view". A "scan" of
the object comprises a set of views made at varying projection
angles, during one or more revolutions of the imaging beam source
and detector array.
[0005] In an axial scan, the projection data is processed to
construct an image that corresponds to one or more two-dimensional
slices or other patterns taken through the object. To form these
slices or patterns, iterative reconstruction of a full field of
view may be performed to increase image quality. Iterative
reconstruction refers to a method that forms an image by repeatedly
adjusting an existing estimate according to the quality of a match
between measured data and simulated measurements from a current
estimate of the image. The quality of the image estimate may also
be affected by consideration of the characteristics of the image
alone, such as its smoothness and/or satisfaction of a
pre-established model. Multiple iterations are performed to create
a resulting reconstructed image that approximately matches the
acquired projection data. A full set of reconstructed images is
referred to as a 3-D reconstruction, because the set is formed into
a three dimensional representation of the object with each image
pixel or picture element corresponding to a single voxel or volume
element in the 3-D reconstruction.
[0006] Traditionally, direct analytical algorithms, such as the
Filtered Back-Projection (FBP) algorithm, have been used to
reconstruct images from CT data. Iterative techniques, such as the
Maximum A Posteriori Iterative Coordinate Descent (MAP-ICD)
algorithm, have also been recently considered for reconstruction of
volumetric CT data to provide means to improve general image
quality over conventional techniques. It has been demonstrated that
reduced noise, enhanced resolution, better low contrast
performance, and reduced artifacts, can all be achieved with
iterative reconstruction of clinical images.
[0007] However, Iterative reconstruction (IR) is not yet available
on commercial scanners, which typically use the analytical FBP
algorithm or its variants. To enable clinical use, current IR may
need to better compete with the spatial resolution properties and
artifact level of FBP.
BRIEF DESCRIPTION OF THE INVENTION
[0008] According to an example embodiment, an improved iterative
reconstruction method to reconstruct a first image includes
generating an imaging beam, receiving said imaging beam on a
detector array, generating projection data based on said imaging
beams received by said detector array, providing said projection
data to an image reconstructor, enlarging one of a plurality of
voxels and a plurality of detectors of the provided projection
data, reconstructing portions of the first image with the plurality
of enlarged voxels or detectors, and iteratively reconstructing the
portions of the first image to create a reconstructed image.
[0009] According to another example embodiment, an imaging system
includes a source generating an imaging beam, a detector array
receiving said imaging beam and generating projection data, a
translatable table configured for disposal of an object thereon and
operable to translate in relation to said source and said detector
array, said source and said detector array rotating about said
translating table to helically scan said object, and an image
reconstructor electrically coupled to said detector array, said
image reconstructor having a processor responsive to computer
executable instructions that, when executed on the processor,
direct the processor to perform an improved iterative
reconstruction method to reconstruct a first image in response to
said projection data. The method includes enlarging one of a
plurality of voxels and a plurality of detectors of the provided
projection data, reconstructing portions of the first image with
the plurality of enlarged voxels or detectors, and iteratively
reconstructing the portions of the first image to create a
reconstructed image.
[0010] These and other advantages and features will become more
apparent from the following description taken in conjunction with
the drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0011] The subject matter, which is regarded as the invention, is
particularly pointed out and distinctly claimed in the claims at
the conclusion of the specification. The foregoing and other
features, and advantages of the invention are apparent from the
following detailed description taken in conjunction with the
accompanying drawings in which:
[0012] FIG. 1 is a pictorial view of an imaging system utilizing a
method of reconstructing an image in accordance with an embodiment
of the present invention;
[0013] FIG. 2 is a block diagrammatic view of the imaging system in
accordance with an embodiment of the present invention
[0014] FIG. 3 is a diagram of a simplified ideal imaging system
with an infinitely small imaging beam focal spot, according to an
example embodiment;
[0015] FIG. 4 is a diagram of a simplified real imaging system with
a finite size imaging beam focal spot, according to an example
embodiment;
[0016] FIG. 5 is a diagram of a simplified imaging system with an
enlarged voxel model, according to an example embodiment;
[0017] FIG. 6 is a diagram of a simplified imaging system with an
enlarged detector model, according to an example embodiment;
[0018] FIG. 7 illustrates a configuration and response of a voxel
and detector of an imaging system;
[0019] FIG. 8 illustrates a configuration and response of an
enlarged rectangular voxel and detector of an imaging system,
according to an example embodiment;
[0020] FIG. 9 illustrates a configuration and response of an
enlarged trapezoidal voxel and detector of an imaging system,
according to an example embodiment;
[0021] FIG. 10 is a graph of a forward projection of differing
voxel and detector footprints, according to an example
embodiment;
[0022] FIG. 11. is a diagram representing aliasing of a measured
projection in an imaging system;
[0023] FIG. 12 is a diagram representing aliasing of a measured
projection in an imaging system;
[0024] FIG. 13 illustrates a reconstructed image utilizing methods
of improved iterative reconstruction, according to an example
embodiment;
[0025] FIG. 14 illustrates a reconstructed image utilizing methods
of improved iterative reconstruction, according to an example
embodiment;
[0026] FIG. 15 illustrates a point spread function of the response
of an imaging system employing enlarged voxel model iterative
reconstruction, according to an example embodiment;
[0027] FIG. 16 illustrates a point spread function of the response
of an imaging system employing enlarged voxel model iterative
reconstruction, according to an example embodiment;
[0028] FIG. 17 illustrates a filtered point spread function of the
response of an imaging system employing enlarged voxel model
iterative reconstruction, according to an example embodiment;
[0029] FIG. 18 illustrates a filtered point spread function of the
response of an imaging system employing enlarged voxel model
iterative reconstruction, according to an example embodiment;
[0030] FIG. 19 illustrates a band-suppression filter model,
according to an example embodiment;
[0031] FIG. 20 illustrates an unfiltered modulation transfer
function of the response of an imaging system employing enlarged
voxel model iterative reconstruction, according to an example
embodiment;
[0032] FIG. 21 illustrates a filtered modulation transfer function
of the response of an imaging system employing enlarged voxel model
iterative reconstruction, according to an example embodiment
[0033] FIG. 22 is a flowchart of a method of iterative
reconstruction, according to an example embodiment; and
[0034] FIG. 23 is a flowchart of a method of improved iterative
reconstruction, according to an example embodiment.
[0035] The detailed description explains embodiments of the
invention, together with advantages and features, by way of example
with reference to the drawings.
DETAILED DESCRIPTION OF THE INVENTION
[0036] Detailed illustrative embodiments are disclosed herein.
However, specific functional details disclosed herein are merely
representative for purposes of describing example embodiments.
Example embodiments may, however, be embodied in many alternate
forms and should not be construed as limited to only the
embodiments set forth herein.
[0037] Accordingly, while example embodiments are capable of
various modifications and alternative forms, embodiments thereof
are shown by way of example in the drawings and will herein be
described in detail. It should be understood, however, that there
is no intent to limit example embodiments to the particular forms
disclosed, but to the contrary, example embodiments are to cover
all modifications, equivalents, and alternatives falling within the
scope of example embodiments.
[0038] It will be understood that, although the terms first,
second, etc. may be used herein to describe various steps or
calculations, these steps or calculations should not be limited by
these terms. These terms are only used to distinguish one step or
calculation from another. For example, a first calculation could be
termed a second calculation, and, similarly, a second step could be
termed a first step, without departing from the scope of this
disclosure. As used herein, the term "and/or" and the "/" symbol
includes any and all combinations of one or more of the associated
listed items.
[0039] As used herein, the singular forms "a", "an" and "the" are
intended to include the plural forms as well, unless the context
clearly indicates otherwise. It will be further understood that the
terms "comprises", "comprising,", "includes" and/or "including",
when used herein, specify the presence of stated features,
integers, steps, operations, elements, and/or components, but do
not preclude the presence or addition of one or more other
features, integers, steps, operations, elements, components, and/or
groups thereof. Therefore, the terminology used herein is for the
purpose of describing particular embodiments only and is not
intended to be limiting of example embodiments.
[0040] It should also be noted that in some alternative
implementations, the functions/acts noted may occur out of the
order noted in the figures. For example, two figures shown in
succession may in fact be executed substantially concurrently or
may sometimes be executed in the reverse order, depending upon the
functionality/acts involved.
[0041] It should also be noted that where the definition of terms
departs from the commonly used meaning of the term, applicant
intends to utilize the definitions provided below, unless
specifically indicated.
[0042] As used herein, the term "voxel" refers to a unit element
representing a volumetric pixel achieved through imaging
information acquired through use of an imaging system, but can also
refer to other basis functions, such as blobs.
[0043] As used herein, the term "image space" refers to a set of
vectors arranged in an array for use with a method of the present
invention. The array may be of any number of dimensions, such as
two-dimensional, three-dimensional, four-dimensional, for example.
An example of an image space that may be used in a method described
herein is a set of all possible images representable on a lattice
of a given dimension. A single element (vector) of the set of the
image space may be viewed on a visual display apparatus to allow a
user to gain information of the interior of a scanned object.
[0044] As used herein, the term "forward model" refers to a
description of the transformation from the image space of a scanned
object to the projection space for a scanned object, as modeled
after the operation of the CT imaging system. The operation of the
forward model on an image vector is referred to as "forward
projection."
[0045] As used herein, the term "computed tomography model" refers
to a mathematical description of the relation between a vector in
the image space and a vector in the projection space. A computed
tomography model includes a forward model and a cost function
chosen to evaluate the closeness of a match between a projection
vector and a forward projection of an image vector by a forward
model.
[0046] As used herein, the term "projection space" refers to a set
of vectors of integral imaging beam attenuation values. The vectors
that make up a projection space may comprise data from an imaging
system. Also, the vectors that make up a projection space may be
forward projections of vectors from an image space. It is
understood that the projections may also be represented as signal
intensities, in which case the forward model typically also
includes an exponentiation commonly referred to as Beer's law.
[0047] As used herein, the term "visual display device" refers to
any type of device such as a CRT monitor, LCD screen, projected
image, etc. used to visually inspect multidimensional vectors.
[0048] As used herein, the term "multi-slice computed tomography
imaging system" refers to an imaging system in which a detector
array contains multiple rows of detectors, each row occupying a
different position along the axis of the system about which the
gantry rotates.
[0049] As used herein, the term "filtered back projection" refers
to a technique of reconstructing images from projection data by
processing data in the projection space, then forming the value of
each element in the image space as a linear combination of values
from processed data, those values taken from projection space
points to which the given image element contributes in forward
projection.
[0050] As used herein, the term "high quality reconstruction image"
refers to an image space vector whose accuracy as a representation
of the object under study is comparable to those produced by
currently available commercial CT imaging systems and known in the
art.
[0051] While example embodiments of the present invention are
described with respect to apparatus and methods of reconstructing
an image using iterative techniques for an imaging system (such as
a multi-slice CT imaging system), the following apparatuses and
methodologies are capable of being adapted for various purposes
including, but not limited to the following applications: magnetic
resonance imaging (MRI) systems, CT systems, radiotherapy systems,
x-ray imaging systems, ultrasound systems, nuclear imaging systems,
positron emission tomography (PET) systems, magnetic resonance
spectroscopy systems, and other applications known in the art, such
as but not limited to applications outside medical imaging such as
nondestructive testing, geological and astronomical imaging, and in
general a large class of inverse problems.
[0052] Example embodiments of the present invention may provide
benefits including improved image quality of iteratively
reconstructed images, especially in terms of spatial resolution,
over/undershoots and aliasing artifacts. As is described more fully
below, example embodiments provide methodologies which may include
a combination of one or more of increased voxel/detector size,
trapezoidal response kernel (anti-symmetric functions for general
cases), band-suppression post-processing technique,
neighborhood-dependent non-constant voxel model (NDNC), adaptive
regularization for iterative reconstruction, general sinogram
preprocessing, and iterative reconstruction for focal spot wobbling
geometry.
[0053] Referring now to FIG. 1, a pictorial view of an imaging
system 10 utilizing a method of reconstructing an image of medical
patient 12 in accordance with an embodiment of the present
invention is shown. The imaging system 10 includes a gantry 14 that
has a rotating inner portion 16 containing an imaging beam source
18 and a detector array 20. Imaging beam source 18 projects an
imaging beam toward detector array 20. Source 18 and detector array
20 rotate about an operably translatable table 22. Table 22 is
translated along the z-axis between source 18 and detector 20 to
perform a helical scan. The beam, after passing through medical
patient 12, within a patient bore 24, is detected at detector array
20 to generate projection data that is used to create a CT
image.
[0054] Referring now to FIG. 2, a diagrammatic view of the imaging
system 10 in accordance with an exemplary embodiment. Source 18 and
detector array 20 rotate about a center axis 30. Beam 32 is
received by multiple detector elements 34 in multiple detector
rows. Each detector element 34 generates an electrical signal
corresponding to the intensity of an impinging imaging beam. As
beam 32 passes through patient 12, beam 32 is attenuated. Rotation
of the center portion of the gantry and the operation of source 18
are governed by a control mechanism 36. Control mechanism 36
includes an imaging beam controller 38 that provides power and
timing signals to imaging beam source 18 and a gantry motor
controller 40 that controls the rotational speed and position of
the center portion of the gantry. A data acquisition system (DAS)
42 samples analog data from detector elements 34 and converts the
analog data to digital signals for subsequent processing. An image
reconstructor 44 receives sampled and digitized imaging beam data
from DAS 42 and performs image reconstruction. A main controller 46
stores the image in a mass storage device 48.
[0055] Main controller 46 also receives commands and scanning
parameters from an operator via an operator console 50. A display
52 allows the operator to observe the reconstructed image and other
data from main controller 46. The operator supplied commands and
parameters are used by main controller 46 in operation of DAS 42,
imaging beam controller 38, and gantry motor controller 40. In
addition, main controller 46 operates a table motor controller 54,
which translates table 22 to position patient 12 in gantry 14.
[0056] Imaging beam controller 38, gantry motor controller 40,
image reconstructor 44, main controller 46, and table motor
controller 54 are preferably based on micro processors device
capable of accepting data and instructions, executing the
instructions to process the data, and presenting the results, such
as a computer having a central processing unit, memory
(nonvolatile, random-access, and/or read-only), and associated
input and output buses. Imaging beam controller 38, gantry motor
controller 40, image reconstructor 44, main controller 46, and
table motor controller 54 may be a portion of a central control
unit or may each be a discrete, stand-alone components as shown.
Therefore, the imaging beam controller 38, gantry motor controller
40, image reconstructor 44, and main controller 46 can be a
microprocessor, microcomputer, a minicomputer, an optical computer,
a board computer, a complex instruction set computer, an ASIC
(application specific integrated circuit), a reduced instruction
set computer, an analog computer, a digital computer, a molecular
computer, a quantum computer, a cellular computer, a
superconducting computer, a supercomputer, a solid-state computer,
a single-board computer, a buffered computer, a computer network, a
desktop computer, a laptop computer, a scientific computer or a
hybrid of any of the foregoing.
[0057] Imaging beam controller 38, gantry motor controller 40,
image reconstructor 44, and main controller 46 also include
operation control methods embodied in application code, such as
that shown in FIG. 21 for example. These methods are embodied in
computer instructions written to be executed by a processor,
typically in the form of software. The software can be encoded in
any language, including, but not limited to, assembly language,
VHDL (Verilog Hardware Description Language), VHSIC HDL (Very High
Speed IC Hardware Description Language), Fortran (formula
translation), Pascal, PL/I. C, C++, Visual C++, C#, Java, ALGOL
(algorithmic language), BASIC (beginners all-purpose symbolic
instruction code), visual BASIC and any combination or derivative
of at least one of the foregoing. Additionally, an operator can use
an existing software application such as a spreadsheet or database
and correlate various cells with the variables enumerated in the
algorithms. Furthermore, the software can be independent of other
software or dependent upon other software, such as in the form of
integrated software.
[0058] FIG. 3 is a diagram of a simplified imaging system,
according to an example embodiment. As illustrated imaging system
300 includes a point source 301 and a detector array 303. The point
source 301 may be a point source of radiation or an imaging beam
configured to penetrate a body forming voxel 302 for imaging.
Radiation emitted from the point source 301 passes through the body
becoming attenuated, and is incident upon the detector array 303.
The incident radiation may form footprint 304, representing the
projection of voxel 302. It is noted that a true point source is an
ideal case, one for which a final reconstructed image may be
compared upon implementation of iterative methodologies as
described herein. In a real case, there may be influences which
cause disturbances on a footprint.
[0059] FIG. 4 is a diagram of a simplified imaging system,
according to an example embodiment. As illustrated imaging system
400 includes a finite source 401 and a detector array 403. The
finite source 401 may be a finite source of radiation or an imaging
beam configured to penetrate a body forming voxel 402 for imaging.
Radiation emitted from the finite source 401 passes through the
body becoming attenuated, and is incident upon the detector array
403. However, as the source 401 is a finite source, a plurality of
incoherent radiation paths may exist from the finite source 401 and
the detector array 403. Thus, the incident radiation may form a
footprint 404 which includes cross-talk, convolution, and/or
blurring due to the size of the finite source 401. This cross-talk,
convolution, and/or blurring may reduce the resolution of voxel
402. Further, conventional models of finite beam width typically
include summing rays corresponding to multiple sample points on the
focal spot, multiple sample points on the detector, and multiple
small rotation increments within one view.
[0060] However, according to example embodiments of the present
invention, techniques for increasing or boosting resolution are
provided which may emulate the convolution (blurring) effects of
finite focal spot in real imaging system with enlarged
voxel/detector models. It is noted that although the terms
cross-talk and blurring are used as examples of interference,
example embodiments are applicable to any form of detector
interference. For example, "rotation" blur, finite x-ray/imaging
beam focal spot size, gaps between detector cells, detector
misalignments, and other interference may be reduced with example
embodiments. Since any deviation of the model from reality incurs a
penalty in terms of spatial resolution for the reconstructed
images, improving model accuracy by means of enlarged
voxels/detectors improves spatial resolution recovery.
[0061] FIG. 5 is a diagram of a simplified imaging system,
according to an example embodiment. The illustrated imaging system
500 includes a source 501 and a detector array 503. The source 501
may be a source of radiation or an imaging beam configured to
penetrate a body creating voxel 520 for imaging. Radiation emitted
from the source 501 passes through the body becoming attenuated,
and is incident upon the detector array 503. As further
illustrated, portion 520 of voxel 502 represents a typical imaging
body voxel as described in FIGS. 3-4. However, the size of voxel
502 is greater than conventional voxels 302 and 402. Thus, the
incident radiation may form footprint 505 which models blurring or
convolution from finite size focal spot. It is noted that the voxel
size is assumed to be larger than the voxel spacing due to the
conventional size of voxels (i.e., there is some overlap between
neighboring voxels). The inclusion of this model may mitigate the
blurring effect.
[0062] Alternatively or in combination with increased voxel sizes,
detector sizes may also be increased to emulate the blurring effect
in the forward projection. It is noted that as used herein,
increasing a detector size means the detector aperture modeled in a
reconstruction algorithm is larger than its true physical
aperture.
[0063] FIG. 6 is a diagram of a simplified imaging system,
according to an example embodiment. As illustrated imaging system
600 includes a source 601 and a detector array 603. The source 601
may be a source of radiation or an imaging beam configured to
penetrate a body creating voxel 602 for imaging. Radiation emitted
from the source 601 passes through the body becoming attenuated,
and is incident upon the detector array 603. As further
illustrated, detector array 603 includes enlarged detectors
(631-632) which are relatively larger than conventional detectors
as described in FIGS. 3-5. Thus, as the size of the detectors
within detector array 603 are larger. Thus, the incident radiation
may form a footprint which includes the effects of finite source
blurring, but also achieves a higher resolution thereby.
Furthermore, when modeling detector blur (e.g., due to cross-talk,
finite focal spot size, etc.) the detector cell size is
assumed/modeled to be larger than the detector cell spacing (i.e.,
there is some overlap between neighboring detector cells).
[0064] Hereinafter configurations and responses for rectangular and
trapezoidal voxel and detector models are described with reference
to FIGS. 7-9.
[0065] FIG. 7 illustrates a configuration and response of a voxel
and detector of an imaging system based upon conventional voxel and
detector sizes. As illustrated, the voxel/detector configuration
701 has a footprint 702 that results in a relatively smooth
response pattern 703-704. More clearly, as voxel and detector sizes
are of conventional sizes, and are rectangular when modeled, a
smooth rectangular response 704 is achieved. However, example
embodiments provide increased size for voxels and/or detectors,
which alters a response based on a configured footprint for the
voxels and detectors.
[0066] FIG. 8 illustrates a configuration and response of a voxel
and detector of an imaging system, according to an example
embodiment. As illustrated, the voxel/detector configuration 801
results in an overlapping rectangular footprint 802 which may
introduce peaks in the response 804. For example, for a
distance-driven projector and/or back-projector technique, two
voxel boundaries may be mapped onto an axis or onto the detector to
generate a rectangular footprint of said voxels. Rectangular voxel
profiles are the conventional model for 3D voxels. From this
rectangular footprint, the overlap kernel computes a weight for the
corresponding voxel-detector pairs. For example, as the
voxels/detectors are of increased size compared to conventional
arrangements, but the physical spacing is the same, there is
overlap which increases the response at regular intervals
representing the overlap of each voxel/detector. Thus, in order to
reduce the peaks in effort to smooth the response, a trapezoidal
footprint configuration may be used for the detectors/voxels.
[0067] FIG. 9 illustrates a configuration and response of a voxel
and detector of an imaging system, according to an example
embodiment. As illustrated, the voxel/detector configuration 901
results in an overlapping trapezoidal footprint 902 which may
reduce peaks in the response 904. For example, for a
distance-driven projector and/or back-projector technique, two
voxel boundaries may be mapped onto an axis or onto the detector to
generate a trapezoidal footprint of said voxels. From this
trapezoidal footprint, the overlap kernel computes a weight for the
corresponding voxel-detector pairs. For example, as the
voxels/detectors are of increased size compared to conventional
arrangements, but the physical spacing is the same, there is
overlap. However, the overlap is on the voxel/detector boundaries
which are angular, thus, decreasing or eliminating the regular
peaks shown in FIG. 8.
[0068] As described above, in contrast to FIG. 8, a trapezoidal
footprint is used for the voxels as this more accurately represents
the re-projection of a rectangular voxel onto the detector. Whereas
such a model does incur some additional arithmetic cost, the
overall computation can remain low, if a conventional kernel is
used in an inner loop of an iterative reconstruction algorithm
(typically the z-direction). This improved voxel profile can be
applied in x, y or z or a combination thereof. Alternatively, the
response of the detector cells may also be modeled by a trapezoidal
profile in a similar way.
[0069] The combination of using a trapezoidal voxel footprint for
voxels/detectors and a distance driven (or other) projection and
back-projection technique may provide increased resolution. For
example, the back-projection technique may be a distance-driven
technique as described in U.S. Pat. No. 7,227,982 and U.S. Pat. No.
6,724,856, both of which are hereby incorporated by reference in
their entirety. The distance-driven technique has two important
aspects. A first aspect includes the mapping of the voxel and
detector boundaries onto a common axis. A second aspect is the use
of the overlap kernel for computation of the projection and
backprojection coefficients. Thus, methodologies included herein
may include an adaptation of the overlap kernel, which may still be
used in combination with mapping boundaries onto a common axis.
However, this technique is not limited to a distance-driven
approach as described below, and instead may be applied to other
forward & backward operations, such as pixel driven, ray driven
and other techniques.
[0070] FIG. 10 is a graph of a forward projection of differing
voxel and detector footprints, according to an example embodiment.
As illustrated in graph 1000, the forward projection of rectangular
configurations of voxels/detectors results in regular peaks
depicted as disturbances 1021 and 1022 in the rectangular forward
model curve 1002. For example, the rectangular voxel/detector model
is represented by the model 1023 which includes references to
overlaps associated with the larger rectangular
voxels/detectors.
[0071] In contrast, the forward projection of trapezoidal
configurations of voxels/detectors mitigates any peaks found in
rectangular models as shown in the disturbance-free trapezoidal
forward model curve 1001. For example, the trapezoidal
voxel/detector model is represented by the model 1013 which
includes references to overlaps associated with the larger
trapezoidal voxels/detectors.
[0072] FIG. 11 is a diagram representing aliasing of a measured
projection in an imaging system. As illustrated, the forward
projection of an exemplary enlarged voxel/detector model 1101
includes smoother oblique edges when compared to the conventional
uniform voxel/detector model 1102, thereby resulting in an ideal
projection 1103 (see image voxels 1104) which is also smooth.
[0073] FIG. 12 is a diagram representing aliasing of a measured
projection in an imaging system. As illustrated, the forward
projection of non-uniform voxel model 1201 includes smoother
oblique edges when compared to the conventional uniform
voxel/detector model 1202, thereby resulting in an ideal projection
1203 (see image voxels 1204) which is also smooth.
[0074] FIGS. 13 and 14 illustrate reconstructed images utilizing
methods of improved iterative reconstruction, according to an
example embodiment. As shown, image 1300, which was reconstructed
using a conventional iterative reconstruction technique, includes
lower resolution as outlined in areas 1304 and 1302 of the image
1300. However, the image 1410 which was reconstructed using an
exemplary iterative reconstruction technique with enlarged
voxels/detectors includes better resolution as outlined in areas
1411 and 1412 of the image 1410. Thus, it is readily apparent that
the benefits of iterative reconstruction techniques as explained
herein result in better image clarity and resolution compared to
conventional iterative reconstruction techniques. Additionally, the
aliasing artifacts are also reduced, as shown through the
differences between FIGS. 13 and 14. More clearly, image 1300
includes a plurality of cross pattern line/streaks, which are
suppressed in image 1410.
[0075] However, example embodiments of the present invention are
not limited to enlarged and/or trapezoidal models for
voxels/detectors. Hereinafter, additional techniques which may be
combined with the discussed embodiments and provide further
benefits over conventional techniques. For example, FIGS. 15-18
provide point spread function illustrations both before and after
implementation of a band suppression filter configured to further
increase image quality of iteratively reconstructed images.
[0076] FIGS. 15 and 16 illustrate a point spread function of the
response of an imaging system, according to an example embodiment.
As illustrated the point spread function is represented in both
pictorial 1501 and graphical 1602 forms. The function depicts
disturbances 1620 and 1621 which result from peaks in a modulated
transfer function of the conventional iterative model. The
modulated transfer function (MTF) is represented in FIGS. 20-21. As
will become readily apparent, if a band suppression filter is
applied, the disturbances caused by MTF peaks are reduced or
mitigated.
[0077] FIGS. 17 and 18 illustrate a filtered point spread function
of the response of an imaging system, according to an example
embodiment. As illustrated the point spread function is represented
in both pictorial 1701 and graphical 1802 forms. The function
depicts reduced disturbances 1820 and 1821 which result from a
reduction in the peaks of a modulated transfer function of the
exemplary iterative model. The modulated transfer function (MTF) is
represented in FIGS. 20-21. As is apparent, if a band suppression
filter is applied, the disturbances caused by MTF peaks are reduced
or mitigated.
[0078] FIG. 19 illustrates an example of a band-suppression filter
model as applied to FIGS. 17-18, according to an example
embodiment. As illustrated, the band-suppression characteristics of
the filter 1900 reduce the disturbances illustrated in FIGS. 15-16.
The band-suppression filter 1900 may be implemented by taking a
Fourier transform (FFT) (in the image or projection domain),
attenuating the frequencies that correspond to the
over-/under-shoot artifacts resulting from
enlarged/voxels/detectors, and taking the inverse Fourier transform
(IFFT).
[0079] FIGS. 20-21 illustrate filtered and unfiltered modulation
transfer functions of the response of an imaging system, according
to an example embodiment. The unfiltered MTF of graph 2010 includes
a peak 2001 which is analog to the disturbances pointed out in
FIGS. 15-16. However, as band suppression filter 1900 is applied,
the post-filtration graph 2120 depicts a reduction or mitigation of
the peak as outlined in area 2102 of the graph 2120.
[0080] Hereinafter, a more detailed description of methodologies of
iterative reconstruction are provided. As noted above, traditional
reconstruction approaches, including FBP-type approaches, typically
assume point voxels, a point source, and point detectors.
Interpolation is performed to project or backproject the values
corresponding to these points. Iterative reconstruction approaches
often model the source as a point, the detectors as points, and the
voxels as points. However, some methods have been published that
take into account the finite extent of source, voxels and
detectors.
[0081] The voxel size may be chosen to be equal to the spacing
between two voxels, and the detector size may be chosen to be equal
to the spacing between two cells, or less, to model the active
area. In contrast, according to example embodiments, methods
including enlarged voxel and/or the detector models are provided.
For example, the models used in defining the iterative
reconstruction algorithm includes a voxel/detector size that is
larger than the inter-voxel/inter-detector distance. Thus, example
embodiments may overcome the drawbacks of the traditional point
source, voxel, and detector models.
[0082] FIG. 22 is a flowchart of a method of iterative
reconstruction, according to an example embodiment. The method 2200
includes an exemplary optimization iterative reconstruction method.
The method 2200 begins by acquiring projection data received by a
detector array (see FIGS. 1-6) at block 2201. The projection data
is then processed in iterative reconstruction process at block 2202
before the high quality reconstruction image is output at block
2203.
[0083] The iterative reconstruction process of block 2202 may
include any proposed enhancement outlined herein. According to at
least one example embodiment, the iterative reconstruction process
includes voxel/detector boosting through enlarging voxels,
detectors, or any combination thereof. For example, FIGS. 5-6
illustrate enlarged voxel/detector implementations. Furthermore,
the iterative reconstruction may include, in addition to enlarged
voxel/detector models, an iterative reconstruction forward model as
described above with reference to FIG. 9. Additionally, cost
function minimization in iterative image reconstruction may also be
included, and is discussed more fully in co-pending U.S. patent
application Ser. No. 12/199,833 entitled "METHOD AND SYSTEM FOR
IMAGE RECONSTRUCTION" which is hereby incorporated by reference in
its entirety.
[0084] However, example embodiments are not limited to iterative
reconstruction methods comprising only enlarged voxel/detector
sizes. For example, a plurality of different artifact reduction
methodologies may be used in combination therewith. Possible
additional artifact reduction techniques are described more fully
with reference to FIG. 23.
[0085] FIG. 23 is a flowchart of a method of improved iterative
reconstruction, according to an example embodiment. The method 2300
includes acquiring projection data at block 2301. The projection
data may be acquired from an imaging system as described in FIGS.
1-6.
[0086] Upon acquisition, the method 2300 includes processing the
projection data in iterative reconstruction process at block 2302
before the high quality reconstruction image is output at block
2303.
[0087] The iterative reconstruction process may include a plurality
of additional sub-processes configured to enhance the output image
and/or reduce artifacts. It is noted however, that although block
2302 is illustrated including all of the below-described
enhancements, any one or more enhancement may be omitted in any
particular implementation. Therefore, example embodiments should
not be limited to the particular combination illustrated, but
rather should be defined by the appended claims.
[0088] According to at least one example embodiment, the iterative
reconstruction includes enlarging voxels/detectors at block 2321.
The voxel/detector enlargement may be similar to the enlargement
described above with reference to FIGS. 3-6.
[0089] According to at least one example embodiment, the iterative
reconstruction includes implementing a trapezoidal footprint at
block 2322 for one or both of the voxels and detectors. For
example, trapezoidal footprint implementation is described more
fully above with reference to FIGS. 9-10.
[0090] According to at least one example embodiment, the iterative
reconstruction includes implementing a neighborhood-dependent
non-constant voxel model (NDNC) at block 2324.
[0091] For example, a NDNC model can be used to reduce the aliasing
artifacts in high-resolution images caused by jagged edges of
simply enlarged voxel models, which may be assumed to be uniform
across their entire volume. The NDNC voxel technique can be
implemented by computing a slope, which depends on the neighboring
voxels, and results in a corresponding slope in the projected voxel
footprint. The slope model can be extended to a non-linearly
varying voxel model, such as a higher order model. Unlike blobs
(e.g., "blob-like" voxels) these models model the variation within
the original voxel footprint based on information from neighborhood
voxels. The neighborhood of the prior voxel can be adapted to
operate on the edge values of the sloped voxel, for example the
prior will prefer the right-most value of a given voxel to be close
to the left-most value its right neighbor. Alternatively, the voxel
is sub-divided into 2 sub-voxels (left-right or up-down) or 4
sub-voxels (4 quadrants) (generally N sub-voxels). The density of
each of the sub-voxels is calculated by interpolation with
neighborhood voxels on the fly. In a forward projection portion of
the iterative reconstruction, the effective voxel footprint is
calculated by the sum of all sub-voxels. This may avoid storage of
a larger image matrix, as would be the case when reconstructing
using smaller voxels. Additionally, as there may be increased
overlap in the computation for the N different sub-voxels, the
total arithmetic cost may be much less than N times longer.
Moreover, those sub-voxels may be limited to some regions, for
example strong gradients regions, which will further decrease the
computational overhead. The slope and sub-voxel models may be
applied in x, y, or z axis or any combination thereof.
[0092] According to at least one example embodiment, the iterative
reconstruction may include implementation of a non-uniform voxel
size model at block 2325.
[0093] In contrast to NDNC, which retains voxel size, another
technique to achieve modeling around high frequency regions is to
use a smaller voxel size near those high frequency regions, with
larger voxel sizes in other regions, thereby including non-uniform
voxel sizes across different regions. However, the non-uniform
voxel size model may require more memory to store more voxels, and
a more complex implementation keeping track of the multi-resolution
image description (i.e., the voxel size changes as a function of
location and anatomical content). Conventional methods utilize
smaller voxels for an entire image space. However, the high
frequency components discussed above may be limited to edges or
high-contrast regions. Therefore, enhanced results may be achieved
by applying smaller voxels for only edge regions or high-frequency
regions. Thus, the proposed non-uniform voxel size model
implemented at edge-regions can keep the additional computation
cost at a relatively low level.
[0094] According to at least one example embodiment, the iterative
reconstruction may include adaptive regularization of the iterative
reconstruction at block 2326.
[0095] For example, in a case of quadratic priors within the
projection data, spatial resolution may suffer because of the
smoothing effect of the prior. Typically, edge-preserving priors
are used to achieve a better trade-off between noise reduction and
spatial resolution. Example embodiments may be configured to adjust
the prior parameters (any of them, but the prior strength and
weights in particular) based on a gradient map or edge map. For
example, a quadratic prior can be used with reduced strength near
edges in order to preserve spatial resolution. In addition example
embodiments may change the prior weights directionally based on the
direction of the gradient to achieve more isotropic resolution.
[0096] According to at least one example embodiment, the iterative
reconstruction may include general sinogram pre-processing at block
2327.
[0097] For example, any technique used for filtered back-projection
in terms of filter or preprocessing on sinogram may be adapted to
iterative reconstruction. For example, in high-resolution mode
(e.g., focal spot deflection mode), the projection data is
interleaved and processed by a filter to enhance resolution
(similarly to FBP but without the ideal ramp filter), then the
preprocessed data are used in iterative reconstruction. With the
aid of a pre-processing filter, a resolution boost may be achieved
by iterative reconstruction. Beside those filtration operations,
better physics modeling of the imaging system may be included into
the pre-processing, such as deconvolution of the imaging beam
source blurring effect, modeling of the electronic noise in the
sinogram data, beam hardening correction, scatter correction, and
other suitable operations. Hence, a more accurate sinogram
representation can be achieved with more accurate physics model,
leading to improved image quality of the output image.
[0098] According to at least one example embodiment, the iterative
reconstruction may include iterative reconstruction for general
focal spot wobbling geometry.
[0099] Focal spot wobbling at the imaging beam source level
increases the sampling rate of the CT scanner and therefore helps
achieve higher spatial resolution. Unlike the "interleaved"
technique discussed above used by filtered back-projection, the
wobbled views may be kept apart and the true wobbled focal spot
positions may be incorporated in the IR forward model.
Incorporating a correct forward model can enhance resolution,
minimize noise and avoid artifacts. This may be particularly useful
if the focal spot positions are sub-optimal for interlacing.
Further, more complex physical processes, such as the finite time
it takes for the focal spot to move, focal spot size modulation, mA
modulation, kVp modulation, various imaging beam filters, bowtie,
may also be modeled if necessary.
[0100] Additionally, according to at least one example embodiment,
the iterative reconstruction includes implementing a
band-suppression filter at block 2323 after iterative
reconstruction. For example, the band suppression filter may be
somewhat similar to the band suppression filter 1900 of FIG. 19.
Finally, the method 2300 includes outputting the reconstructed
image at block 2303.
[0101] As described above, improvements to iterative reconstruction
in imaging systems are provided. The improvements may be based on
increased or enlarged voxel/detector sizes in acquired projection
data, and may be used with a combination of other resolution
enhancing techniques as outlined above.
[0102] Example embodiments of the present invention improve the
spatial resolution of iterative reconstruction, reduce or eliminate
the aliasing artifacts in output images particularly in low
contrast structures, and reduce or eliminate over- and under-shoot
artifacts that may result from artificially boosting
resolution.
[0103] Thus example embodiments of the present invention provide
enhancements to iterative reconstruction that improve
reconstruction in line with results from conventional
reconstruction techniques including FBP with finely tuned
filtration kernels that help achieve high spatial resolution.
Additionally, the proposed resolution boosting techniques induce
limited impact on the reconstruction performance (i.e., speed). For
example, voxel/detector enlargement and other enhancements are
faster than techniques that directly model the physical blurring
process. Additionally, the reduction in aliasing artifacts improves
iteratively reconstructed images, especially for low contrast
structures. Without using the proposed techniques, streak artifacts
tend to appear especially when the regularization is reduced to
achieve higher resolution (see FIG. 10). Furthermore, as compared
with blob modeling methods, which also try to mimic artificially
larger voxel profiles in the forward model, example embodiments are
simpler (e.g., only a linear profile is used), resulting in better
reconstruction performance. For example, the NDNC voxel model
addition, which only requires more computation around edge pixels,
may reach the same artifacts level as images reconstructed with
two-times smaller pixels. Furthermore, example embodiments reduce
over- and under-shoot artifacts which improves images around edges,
such as skin and bone. Additionally, the modified voxel/detector
system model works efficiently in currently available framework,
which makes it inexpensive to integrate it into new iterative
reconstruction methods.
[0104] An embodiment of the invention may be embodied in the form
of computer-implemented processes and apparatuses for practicing
those processes. The present invention may also be embodied in the
form of a computer program product having computer program code
containing instructions embodied in tangible media, such as floppy
diskettes, CD-ROMs, hard drives, USB (universal serial bus) drives,
or any other computer readable storage medium, such as random
access memory (RAM), read only memory (ROM), or erasable
programmable read only memory (EPROM), for example, wherein, when
the computer program code is loaded into and executed by a
computer, the computer becomes an apparatus for practicing the
invention. The present invention may also be embodied in the form
of computer program code, for example, whether stored in a storage
medium, loaded into and/or executed by a computer, or transmitted
over some transmission medium, such as over electrical wiring or
cabling, through fiber optics, or via electromagnetic radiation,
wherein when the computer program code is loaded into and executed
by a computer, the computer becomes an apparatus for practicing the
invention. When implemented on a general-purpose microprocessor,
the computer program code segments configure the microprocessor to
create specific logic circuits. A technical effect of the
executable instructions is to reconstruct two dimensional
projection data into three dimensional image data that may be used
by a clinician for diagnostic purposes.
[0105] While the invention has been described with reference to
exemplary embodiments, it will be understood by those skilled in
the art that various changes may be made and equivalents may be
substituted for elements thereof without departing from the scope
of the invention. In addition, many modifications may be made to
adapt a particular situation or material to the teachings of the
invention without departing from the essential scope thereof.
Therefore, it is intended that the invention not be limited to the
particular embodiment disclosed as the best or only mode
contemplated for carrying out this invention, but that the
invention will include all embodiments falling within the scope of
the appended claims. Also, in the drawings and the description,
there have been disclosed exemplary embodiments of the invention
and, although specific terms may have been employed, they are
unless otherwise stated used in a generic and descriptive sense
only and not for purposes of limitation, the scope of the invention
therefore not being so limited. Moreover, the use of the terms
first, second, etc. do not denote any order or importance, but
rather the terms first, second, etc. are used to distinguish one
element from another. Furthermore, the use of the terms a, an, etc.
do not denote a limitation of quantity, but rather denote the
presence of at least one of the referenced item.
* * * * *