U.S. patent application number 14/048331 was filed with the patent office on 2014-06-26 for determining position of medical device in branched anatomical structure.
This patent application is currently assigned to Intuitive Surgical Operations, Inc.. The applicant listed for this patent is Intuitive Surgical Operations, Inc.. Invention is credited to Giuseppe Maria Prisco, Tao Zhao.
Application Number | 20140180063 14/048331 |
Document ID | / |
Family ID | 50477813 |
Filed Date | 2014-06-26 |
United States Patent
Application |
20140180063 |
Kind Code |
A1 |
Zhao; Tao ; et al. |
June 26, 2014 |
DETERMINING POSITION OF MEDICAL DEVICE IN BRANCHED ANATOMICAL
STRUCTURE
Abstract
Information extracted from sequential images captured from the
perspective of a distal end of a medical device moving through an
anatomical structure are compared with corresponding information
extracted from a computer model of the anatomical structure. A most
likely match between the information extracted from the sequential
images and the corresponding information extracted from the
computer model is then determined using probabilities associated
with a set of potential matches so as to register the computer
model of the anatomical structure to the medical device and thereby
determine the lumen of the anatomical structure which the medical
device is currently in. Sensor information may be used to limit the
set of potential matches. Feature attributes associated with the
sequence of images and the set of potential matches may be
quantitatively compared as part of the determination of the most
likely match.
Inventors: |
Zhao; Tao; (Sunnyvale,
CA) ; Prisco; Giuseppe Maria; (Calci, IT) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Intuitive Surgical Operations, Inc. |
Sunnyvale |
CA |
US |
|
|
Assignee: |
Intuitive Surgical Operations,
Inc.
Sunnyvale
CA
|
Family ID: |
50477813 |
Appl. No.: |
14/048331 |
Filed: |
October 8, 2013 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
61713010 |
Oct 12, 2012 |
|
|
|
Current U.S.
Class: |
600/424 |
Current CPC
Class: |
A61B 1/0005 20130101;
G06T 2207/30004 20130101; A61B 1/2676 20130101; A61B 5/064
20130101; A61B 6/5247 20130101; A61B 5/066 20130101; A61B 34/20
20160201; A61B 2090/365 20160201; G06T 2207/10068 20130101; A61B
8/5261 20130101; G06T 2207/10016 20130101; A61B 6/12 20130101; A61B
2034/105 20160201; A61B 5/0059 20130101; G06T 7/75 20170101; A61B
8/0841 20130101; A61B 2034/2065 20160201 |
Class at
Publication: |
600/424 |
International
Class: |
A61B 5/06 20060101
A61B005/06 |
Claims
1. A medical system comprising: a memory storing information of a
computer model of a branched anatomical structure; and a processor
programmed to register the computer model to a medical device for
determining a position of the medical device in the branched
anatomical structure by determining a most likely match between
information which has been extracted from a sequence of images that
has been captured by an image capturing device from a perspective
of a distal end of the medical device while moving through a path
including a plurality of lumens in the branched anatomical
structure and corresponding information that has been extracted
from the computer model of the branched anatomical structure.
2. The medical system of claim 1, wherein the information extracted
from the sequence of images comprises information of feature points
appearing in the sequence of images.
3. The medical system of claim 1, wherein the information extracted
from the sequence of images comprises information of lumens
appearing in the sequence of images.
4. The medical system of claim 3, further comprising: extracting
blobs from the sequence of captured images so that each lumen
appearing in the sequence of images is indicated by one of the
extracted blobs and the information of lumens appearing in the
sequence of images comprises information of the extracted
blobs.
5. The medical system of claim 4, wherein the information of the
extracted blobs comprises information of feature attributes of the
extracted blobs.
6. The medical system of claim 4, wherein the processor is
programmed to identify the set of potential matches in the computer
model of the branched anatomical structure for each of the
plurality of tracklets by: determining information of feature
attributes of the extracted blobs; and comparing the feature
attributes with corresponding feature attributes extracted from the
computer model of the branched anatomical structure.
7. The medical system of claim 6, wherein the determining
information of feature attributes of the extracted blobs comprises
determining information of one or more of: the number, shapes,
sizes, locations, and orientations of the extracted blobs.
8. The medical system of claim 7 wherein the determining
information of the feature attributes of the extracted blobs
comprises: determining at least one of one or more ratios of
dimensions related to the extracted blobs and one or more angles
between lines extending through centroids of pairs of extracted
blobs, wherein each pair of extracted blobs indicate lumens that
branch out of a same node in the branched anatomical structure.
9. The medical system of claim 8, wherein the identifying of the
set of potential matches in the computer model of the branched
anatomical structure for each of the plurality of tracklets
comprises restricting the set of potential matches to nodes
corresponding to synthetic images that are extractable from the
computer model of the branched anatomical structure and have the
same number and relative positions of lumens appearing in the
synthetic images as in the tracklet, wherein the relative positions
indicate which lumens appear within other lumens.
10. The medical system of claim 9, wherein the identifying of the
set of potential matches in the computer model of the branched
anatomical structure for each of the plurality of tracklets
comprises restricting the set of potential matches to nodes
corresponding to synthetic images that are extractable from the
computer model of the branched anatomical structure and have
feature attributes that are within a threshold difference to
feature attributes of the extracted blobs of the tracklet.
11. The medical system of claim 10, wherein the identifying of the
set of potential matches in the computer model of the branched
anatomical structure for each of the plurality of tracklets
comprises generating quantitative comparisons of the feature
attributes of the extracted blobs of the tracklet to corresponding
feature attributes of the synthetic images of members of the set of
potential matches by determining a value indicating a difference
between one of the feature attributes in the extracted blobs and a
corresponding one of the feature attributes in the synthetic
images.
12. The medical system of claim 1, wherein the sequence of images
comprises a plurality of tracklets, wherein each of the plurality
of tracklets comprises a subset of the sequence of images which is
topologically equivalent, but topologically different than adjacent
ones of the plurality of tracklets, and wherein the processor is
programmed to determine the most likely match by: identifying a set
of potential matches in the computer model of the branched
anatomical structure for each of the plurality of tracklets;
determining a probability indicating a quality of match for each
member of the set of potential matches for each of the plurality of
tracklets; and determining a maximal combined probability using
information of the probabilities for the set of potential matches
for each of the plurality of tracklets.
13. The medical system of claim 12, wherein the processor is
further programmed to determine the most likely match by:
determining a set of transitional probabilities for each member of
the set of potential matches for each of the plurality of
tracklets, wherein each member of the set of transitional
probabilities indicates a quality of match for a transition from
the tracklet to a next in sequence tracklet of the plurality of
tracklets; and determining the maximal combined probability using
information of the probabilities and transitional probabilities for
the set of potential matches for each of the plurality of
tracklets.
14. The medical system of claim 13, wherein the processor is
programmed to determine the set of transitional probabilities for
each member of the set of potential matches for each of the
plurality of tracklets using static knowledge of connectivity in
the computer model of the branched anatomical structure.
15. The medical system of claim 13, wherein the processor is
programmed to determine the set of transitional probabilities for
each member of the set of potential matches for each of the
plurality of tracklets using dynamic knowledge of the movement of
the medical device through the path in the branched anatomical
structure.
16. The medical system of claim 13, wherein the processor is
programmed to identify the set of potential matches in the computer
model of the branched anatomical structure for each of the
plurality of tracklets by only including branches of the computer
model of the branched anatomical structure which are topologically
equivalent to the tracklet.
17. The medical system of claim 13, wherein a same set of lumens
appears in each tracklet and wherein adjacent tracklets have at
least one lumen in common.
18. The medical system of claim 17, wherein the processor is
programmed to identify each tracklet by extracting and tracking
blobs in the sequence of images until a set of the blobs changes,
wherein each of the blobs corresponds to a lumen within the
branched anatomical structure.
19. The medical system of claim 17, wherein the processor is
programmed to determine membership of extracted blobs for each
tracklet by ignoring false additions and subtractions from the set
of extracted blobs in the sequence of images.
20. The medical system of claim 19, wherein the processor is
programmed to analyze the set of sequential images to detect
falsely identified lumens in the extracted blobs which would result
in false additions to the set of extracted blobs and to detect a
false determination of the medical device having entered a lumen
which would result in a false subtraction from the set of extracted
blobs.
21. The medical system of claim 13, wherein the processor is
programmed to determine information of feature attributes of the
extracted blobs by filtering the determined information by using at
least one of: averaging the determined information and eliminating
outliers in the determined information.
22. The medical system of claim 13, wherein the computer model of
the branched anatomical structure is spatially registered to the
branched anatomical structure which the medical device is moving
through, the medical system further comprising: a position sensor
for determining a position of the distal end of the medical device
relative to the branched anatomical structure; wherein the
processor is programmed to identify the set of potential matches in
the computer model of the branched anatomical structure for each of
the plurality of tracklets by using position information from the
position sensor.
23. The medical system of claim 13, wherein the computer model of
the branched anatomical structure is spatially registered to the
branched anatomical structure which the medical device is moving
through, the medical system further comprising: an orientation
sensor for determining an orientation of the distal end of the
medical device relative to a fixed horizontal plane; wherein the
processor is programmed to identify the set of potential matches in
the computer model of the branched anatomical structure for each of
the plurality of tracklets by using orientation information from
the orientation sensor.
24. The medical system of claim 13, wherein the computer model of
the branched anatomical structure is spatially registered to the
branched anatomical structure which the medical device is moving
through, the medical system further comprising: a roll sensor for
determining a roll angle of the medical device about a longitudinal
axis of the medical device; wherein the processor is programmed to
identify the set of potential matches in the computer model of the
branched anatomical structure for each of the plurality of
tracklets by using roll information from the roll sensor.
25. The medical system of claim 13, wherein the computer model of
the branched anatomical structure is spatially registered to the
branched anatomical structure which the medical device is moving
through, the medical system further comprising: an insertion sensor
for determining an insertion distance of the distal end of the
medical device into the branched anatomical structure. wherein the
processor is programmed to identify the set of potential matches in
the computer model of the branched anatomical structure for each of
the plurality of tracklets by using insertion information from the
insertion sensor.
26. A method for determining a position of the medical device in
the branched anatomical structure, the method comprising:
determining a most likely match between information that has been
extracted from a sequence of images that have been captured from a
perspective of a distal end of a medical device as the medical
device moves through a path including a plurality of lumens in the
branched anatomical structure and corresponding information that
has been extracted from the computer model of the branched
anatomical structure.
27. The method of claim 26, wherein the sequence of images
comprises a plurality of tracklets, wherein each of the plurality
of tracklets comprises a subset of the sequence of images which is
topologically equivalent, but topologically different than adjacent
ones of the plurality of tracklets, and wherein the determining of
the most likely match comprises: identifying a set of potential
matches in the computer model of the branched anatomical structure
for each of the plurality of tracklets; determining a probability
indicating a quality of match for each member of the set of
potential matches for each of the plurality of tracklets; and
determining a maximal combined probability using information of the
probabilities for the set of potential matches for each of the
plurality of tracklets.
28. The method of claim 27, wherein the determining of the most
likely match further comprises: determining a set of transitional
probabilities for each member of the set of potential matches for
each of the plurality of tracklets, wherein each member of the set
of transitional probabilities indicates a quality of match for a
transition from the tracklet to a next in sequence tracklet of the
plurality of tracklets; and determining the maximal combined
probability using information of the probabilities and transitional
probabilities for the set of potential matches for each of the
plurality of tracklets.
29. The method of claim 28, further comprising: extracting blobs
from the sequence of captured images so that each lumen appearing
in the sequence of images is indicated by one of the extracted
blobs and the information of lumens appearing in the sequence of
images comprises information of the extracted blobs.
Description
FIELD OF THE INVENTION
[0001] The present invention generally relates to medical systems
and in particular, to a system and method for determining the
position of a medical device in a branched anatomical structure by
comparing information of images captured at a distal end of the
medical device as it moves through the branched anatomical
structure with corresponding information extracted from a computer
model of the branched anatomical structure.
BACKGROUND
[0002] Image guided surgery helps surgeons navigate medical devices
to targets in patients so that therapeutic and/or diagnostic
medical procedures may be performed on the targets. For guidance,
the position of a distal end of a medical device may be tracked and
its image displayed along with or superimposed on a computer model
of an anatomical structure associated with the target. The computer
model may be generated from pre-operative and/or intra-operative
patient anatomy scan data such as x-ray, ultrasound, fluoroscopy,
computed tomography (CT), magnetic resonance imaging (MRI), and
other imaging technologies. The medical device may be an endoscope,
catheter, or medical instrument that has a steerable tip and
flexible body capable of conforming to body passages or lumens
leading to the target in a branched anatomical structure of the
patient.
[0003] Proper registration of the medical device, anatomical
structure, and computer model of the anatomical structure with
respect to each other is desirable for accurate image guided
surgery. Therefore, registration of these items is typically
performed prior to performing a medical procedure on a patient.
However, registration errors may develop during the performance of
the medical procedure due to movement of the anatomical structure
and/or difficulties in tracking the medical device as it moves
through the anatomical structure. For continuously moving branched
anatomical structures in which flexible medical devices are
navigated to target areas, maintaining proper registration between
tracked positions of medical devices and the branched anatomical
structures is especially challenging.
[0004] U.S. 2005/0182319 describes tracking movement of anatomical
structures using dynamic referencing and/or gating techniques and
tracking movement of medical devices as they move inside anatomic
structures using Electromagnetic (EM) tracking devices. However,
due to inaccuracies at least partly attributed to the dynamic
referencing, EM tracking devices and anatomical motion, accurate
registration of medical devices moving through branched anatomical
structures such as the lungs or heart are prone to error when a
plurality of lumens in the branched anatomical structure reside
within uncertainty regions resulting from such inaccuracies.
[0005] Luo, Xiongbiao et al. "On Scale Invariant Features and
Sequential Monte Carlo Sampling for Bronchoscope Tracking," Medical
Imaging 2011: Visualization, Image-Guided Procedures, and Modeling,
edited by Kenneth H. Wong, David R. Homes III, Proc. Of SPIE Vol.
7964, 79640Q, describes a two-stage image-based method for
determining a maximum similarity between a current bronchoscope
camera frame and a generated virtual frame to determine a
bronchoscope pose. The first stage predicts the inter-frame motion
parameters between consecutive images of a bronchoscope video using
scale invariant feature transform (SIFT) features and epipolar
geometry analysis. The second stage recursively approximates the
posterior probability density of the current bronchoscope camera
pose in accordance with the estimated result of the first stage.
Since the second stage generates a set of random samples that are
defined as the camera motion parameters and the similarity between
virtual bronchoscopic and patient specific real bronchoscopic
images, the current camera motion parameters can be determined to
be equal to the pose of one sample that corresponds to the maximum
similarity inside the sample set. Although the method has been
shown to provide good accuracy, the complexity of its calculations
using Sequential Monte Carlo (SMC) methods is unsuitable for
real-time applications with a computational time around 3.0 seconds
per frame.
[0006] Soper, Timothy D. et al. "In Vivo Validation of a Hybrid
Tracking System for Navigation of an Ultrathin Bronchoscope within
Peripheral Airways," IEEE Transactions on Biomedical Engineering,
Vol. 57, No. 3, March 2010, pp. 736-745, describes a hybrid
approach in which both electromagnetic tracking (EMT) and
image-based tracking (IBT) are employed along with an error state
Kalman filter which adaptively estimates the localization error
between the two tracking inputs. When the error becomes large,
however, the system is incapable of self-correcting itself.
Therefore, operator intervention may become necessary once the
tracking diverges from the true path of the bronchoscope in an
anatomical structure.
[0007] U.S. Pat. No. 7,756,563 describes a method to perform camera
pose estimation in bronchoscopy by continuously aligning a live
camera view to a corresponding view which has been rendered from a
computer model. It assumes that the estimation from a previous
frame is accurate to seed the iterative optimization of the current
frame. The method, however, is not able to recover from any
tracking failure which is very likely to occur in real life.
OBJECTS AND SUMMARY
[0008] Accordingly, one object of one or more aspects of the
present invention is a medical system and method implemented
therein for determining a position of a medical device in a
branched anatomical structure as the medical device moves through
the branched anatomical structure.
[0009] Another object of one or more aspects of the present
invention is a medical system and method implemented therein that
are self-correcting for determining a position of a medical device
in a branched anatomical structure as the medical device moves
through the branched anatomical structure.
[0010] Another object of one or more aspects of the present
invention is a medical system and method implemented therein that
are computationally efficient and suitable for real-time
applications for determining a position of a medical device in a
branched anatomical structure as the medical device moves through
the branched anatomical structure.
[0011] Another object of one or more aspects of the present
invention is a medical system and method implemented therein that
provide accurate results for determining a position of a medical
device in a branched anatomical structure as the medical device
moves through the branched anatomical structure.
[0012] These and additional objects are accomplished by the various
aspects of the present invention, wherein briefly stated, one
aspect is medical system comprising: a memory storing information
of a computer model of a branched anatomical structure; and a
processor programmed to register the computer model to a medical
device for determining a position of the medical device in the
branched anatomical structure by determining a most likely match
between information which has been extracted from a sequence of
images that has been captured by an image capturing device from a
perspective of a distal end of the medical device as the medical
device moves through a plurality of lumens in the branched
anatomical structure and corresponding information extracted from
the computer model of the branched anatomical structure.
[0013] Another aspect is method for determining a position of the
medical device in the branched anatomical structure, the method
comprising: determining a most likely match between information
extracted from a sequence of images that have been captured from a
perspective of a distal end of a medical device as the medical
device moves through a plurality of lumens in the branched
anatomical structure and corresponding information extracted from
the computer model of the branched anatomical structure.
[0014] Additional objects, features and advantages of the various
aspects of the present invention will become apparent from the
following description which should be taken in conjunction with the
accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0015] FIG. 1 illustrates a medical system, utilizing aspects of
the present invention, which includes a hand-operated medical
device.
[0016] FIG. 2 illustrates an alternative medical system, utilizing
aspects of the present invention, which includes a teleoperated
medical device.
[0017] FIG. 3 illustrates a diagram of preoperative tasks conducted
prior to performing a medical procedure on a patient.
[0018] FIG. 4 illustrates a view of an auxiliary screen during
navigation of a medical device to a target area in a branched
anatomical structure.
[0019] FIG. 5 illustrates a flow diagram of a method for
registering a computer model of a branched anatomical structure to
a medical device for determining a position of the medical device
in the branched anatomical structure, utilizing aspects of the
present invention.
[0020] FIG. 6 illustrates an image captured by image capturing
device from a perspective of a distal end of a medical device as
the medical device moves through a branched anatomical
structure.
[0021] FIG. 7 illustrates blobs extracted from the captured image
of FIG. 6 along with line segments useful for defining feature
attributes for the blobs.
[0022] FIGS. 8-13 illustrate various examples of blob topologies
which may be extracted from images captured by image capturing
device from a perspective of a distal end of a medical device as
the medical device moves through a branched anatomical
structure.
[0023] FIG. 14 illustrates an example of a false negative blob
identification for an image captured by image capturing device from
a perspective of a distal end of a medical device as the medical
device moves through a branched anatomical structure.
[0024] FIG. 15 illustrates an example of a false positive blob
identification for an image captured by image capturing device from
a perspective of a distal end of a medical device as the medical
device moves through a branched anatomical structure.
[0025] FIGS. 16-17 illustrate examples of blob characteristics
which may be used to define feature attributes of blobs extracted
from images captured by image capturing device from a perspective
of a distal end of a medical device as the medical device moves
through a branched anatomical structure.
[0026] FIGS. 18-19 illustrate an example of a sequence of blobs
extracted from images captured by image capturing device from a
perspective of a distal end of a medical device as the medical
device moves towards a bifurcation in a branched anatomical
structure.
[0027] FIG. 20 illustrates blobs extracted from the captured image
of FIG. 6 along with line segments used for defining feature
attributes for the blobs.
[0028] FIG. 21 illustrates a captured image blob whose shape is a
feature attribute being used as a template to find matching blobs
from synthetic images generated from a computer model.
[0029] FIG. 22 illustrates a schematic of a medical device with
optional position, orientation, roll, and insertion sensors as the
medical device moves through a lumen of an anatomical
structure.
[0030] FIG. 23 illustrates a Cartesian coordinate system associated
with the optional position sensor of FIG. 22.
[0031] FIG. 24 illustrates an orientation angle associated with the
optional orientation sensor of FIGS. 22.
DETAILED DESCRIPTION
[0032] FIG. 1 illustrates, as an example, a medical system 100
including a steerable medical device 110, one or more sensors 131,
a sensor(s) processor 130, one or more signal communication cables
132 coupling the one or more sensors 131 to the sensor(s) processor
130, an image capturing element 141, an image processor 140, an
optical fiber or electrical cable 142 coupling the image capturing
element 141 to the image processor 140, a display processor 150, a
primary display screen 151, an auxiliary display screen 152, a main
processor 160, and memory 161. Although shown as separate units,
the sensor(s) processor 130, image processor 140, display processor
150, and main processor 160 may be implemented in a single
processor or their respective functions distributed among a
plurality of processors, wherein each of such processors may be
implemented as hardware, firmware, software or a combination
thereof. As used herein, the term processor is understood to
include interface logic and/or circuitry for translating and/or
communicating signals into and/or out of the processor as well as
conventional digital processing logic. The memory 161 may be any
memory device or data storage system as conventionally used in
computer systems. The primary and auxiliary display screens, 151
and 152, are preferably computer monitors capable of displaying
three-dimensional images to an operator of the system 100. However,
for cost or other considerations, either or both of the primary and
auxiliary display screens, 151 and 152, may be a standard computer
monitor capable of only displaying two-dimensional images.
[0033] The medical device 110 has a flexible body 114, a steerable
tip 112 at its distal end 111, and a hand-operable handle 116 at
its proximal end 115. Control cables (not shown) or other control
means typically extend from the handle 116 to the steerable tip 112
so that the tip 112 may be controllably bent or turned as shown for
example by dotted line versions of the bent tip 112. The medical
device 110 may be an endoscope, catheter or other medical
instrument having a flexible body and steerable tip.
[0034] The image capturing element 141 may be a stereoscopic or
monoscopic camera or other imaging device disposed at the distal
end 111 for capturing images that are transmitted to and processed
by the image processor 140 and/or display processor 150 and
displayed on the primary display screen 151, auxiliary display
screen 152, and/or other display means according to the various
aspects of the invention as described herein. Alternatively, the
image capturing element 141 may be a fiber-optic bundle that
couples to an imaging and processing system on the proximal end of
the medical device 110, such as a fiberscope. The image capturing
element 141 may also be single or multi-spectral that captures
image data in the visible or infrared/ultraviolet spectrum. Thus,
any image capturing element, device, or system referred to herein
may be any one or a combination of these and other imaging
technologies. One of a plurality of fiber optic cables (not shown)
may be coupled at its proximal end to a light source (not shown)
for illumination purposes at the distal end 111. Other of the
plurality of fiber optic cables (not shown) may be configured with
position and bend or shape sensors such as Fiber Bragg Gratings (or
other strain sensors such as those employing Rayleigh scattering)
distributed along the length of the medical device 110 so that
light passing through these fiber optic cables is processed by the
sensor(s) processor 130 to determine a current pose and shape of
the medical device 110.
[0035] FIG. 2 illustrates, as an example, an alternative embodiment
of the medical system 100 in which the handle 116 is replaced by an
electromechanical interface 170, controller 180, and input device
190 for teleoperating the medical device 110. The interface 170
includes actuators for actuating cables in the medical device 110
to steer its tip 112 as well as an actuator for moving the entire
medical device 110 forward and backward so that it may be inserted
into and retracted out of a patient through an entry port such as a
natural body orifice or a surgeon created minimally invasive
incision. In addition, the interface 170 may include an actuator
for rotating the medical device 110 about its central longitudinal
axis. The controller 180 is preferably implemented as hardware,
firmware or software (or a combination thereof) in the same one or
more computer processors as the processors 130, 140, 150, and 160,
or a different computer processor. The flexible body 114 may be
passively or actively bendable. The medical system can also be a
hybrid of the above two examples.
[0036] Examples of such steerable medical devices are described in
U.S. 2010/0249506 A1 entitled "Method and System for Assisting an
Operator in Endoscopic Navigation" and WO 2009/097461 A1entitled
"Apparatus and Methods for Automatically Controlling an Endoscope,
which are each incorporated herein by reference. Details on the
determination of the endoscope's position and bending using Fiber
Bragg Gratings may be found, for examples, in U.S. 2007/0156019 A1
entitled "Robotic Surgery System Including Position Sensors Using
Fiber Bragg Gratings", U.S. 2008/0212082 A1 entitled "Fiber Optic
Position and/or Shape Sensing Based on Rayleigh Scatter", U.S.
2008/0218770 A1 entitled "Robotic Surgical Instrument and Methods
using Bragg Fiber Sensors", and U.S. 2009/0324161 A1 entitled
"Fiber Optic Shape Sensor", which are each incorporated herein by
reference.
[0037] FIG. 3 illustrates, as an example, a diagram of preoperative
tasks that may be performed in planning a medical procedure to be
performed on a patient. In the following example, the branched
anatomical structure may move during a medical procedure as in the
periodic motion of the air and blood circulatory systems or in a
non-periodic motion such as a cough or other body spasm.
[0038] In block 301, a set of images of a patient is acquired using
an appropriate imaging technology from which a three-dimensional
(3-D) computer model of the branched anatomical structure may be
generated. Examples of such an imaging technology include, but are
not limited to, fluoroscopy, Magnetic Resonance Imaging,
thermography, tomography, ultrasound, Optical Coherence Tomography,
Thermal Imaging, Impedance Imaging, Laser Imaging, and nano-tube
X-ray imaging. If the branched anatomical structure is subject to
expansion/contraction cycles, such as the human lungs, it may be
advantageous to acquire the set of images at an extremum using a
triggering signal, such as from a respirator or motion
detector.
[0039] In block 302, a three-dimensional (3-D) computer model of
the branched anatomical structure is generated from the acquired
images of the anatomical structure. In block 303, one or more
targets may be identified in the branched anatomical structure. The
targets are locations or objects in or adjacent to the anatomical
structure where or upon which a medical procedure is to be
performed. For example, the target may be a tumor in or adjacent to
the anatomical structure. The target(s) may be identified by a
surgeon or radiologist in a conventional manner by analysis of the
acquired images of the anatomical structure or the generated 3-D
computer model information, whichever is more convenient and/or
reliable for such identification.
[0040] In block 304, a navigational path may be determined to and
through the anatomical structure for the working end of the medical
device 110 to travel to each target. In this case, the working end
is assumed to be the distal end 111 of the medical device 110. The
surgeon may determine a suitable navigational path to a target by
analyzing the acquired images of the anatomical structure or the
generated 3-D computer model so as to take into account any damage
to the patient that the medical device 110 may cause as it moves
towards the target as well as the shortest time and/or shortest
path. Alternatively, a computer program may be executed by a
processor to perform such analysis to determine the navigational
path using artificial intelligence techniques.
[0041] FIG. 4 illustrates, as an example, a view of the auxiliary
display screen 152 during navigation of the medical device 110 to a
target area in an anatomical structure. The view may be either a
2-D or 3-D view of a computer model 420 of the branched anatomical
structure and a computer model 410 of the medical device 110, which
is updated in real-time as the medical device 110 moves through the
anatomical structure. Also shown is an indication 421 of the
target. Thus, the auxiliary screen 152 assists the surgeon to steer
the medical device 110 through the anatomical structure to the
target.
[0042] In this example, the branched anatomical structure is a pair
of lungs having a plurality of natural body passages or lumens
including a trachea, bronchi, and bronchioles. The medical device
110 is a bronchoscope and its entry port into the patient is the
patient's mouth. Due to the nature of the lungs, the medical device
110 may be guided through a number of linked lumens or branches of
the bronchial tree. In doing so, the flexible body 114 of the
medical device 110 conforms to the passages through which it
travels. Although a pair of lungs is shown in the present example,
it is to be appreciated that the various aspects of the present
invention are also applicable and useful for other anatomical
structures such as the heart, brain, digestive system, circulatory
system, and urinary system, in addition to the respiratory
system.
[0043] In addition, or alternatively, to displaying computer models
of the branched anatomical structure and medical device on the
auxiliary display screen 152 as shown in FIG. 4, an image captured
by the image capturing element 141 may be shown side-by-side on the
primary display screen 151 with a synthetic image which is
generated from the 3-D computer model of the branched anatomical
structure from the perspective of the distal end 111 of the medical
device 110. In this case, an arrow may be displayed on the
synthetic image to indicate a direction to be taken towards the
target. For additional details, see, for example, U.S. application
Ser. No. 13/107,562 entitled "Medical system providing dynamic
registration of a model of an anatomical structure for image-guided
surgery," filed May 13, 2011, Attorney Docket No. ISRG03100/US,
which is incorporated herein by reference.
[0044] A number of pre-operative registration tasks are performed
in a conventional manner in preparation of performing a medical
procedure on a patient using the system 100. First, the medical
device 110 is localized to a fixed reference frame by, for example,
touching the distal end 111 of the medical device 110 to one or
more known and stationary points in the fixed reference frame.
Second, the patient may be registered to the fixed reference frame
by touching the distal end 111 of the medical device 110 to one or
more points on the patient, which points correspond to identifiable
points on the acquired images of the patient such as natural body
features or artificial markers. Third, the computer model of an
anatomical structure may be registered to the patient using
corresponding reference points on the patient and the computer
model such as natural body features or artificial markers. Thus,
the medical device 110, branched anatomical structure, and computer
model of the anatomical structure may be registered in this way to
each other and to the fixed reference frame.
[0045] During the performance of a medical procedure on the
patient, however, due in part to inherent inaccuracies in tracking
the position of the distal end of the medical device 110 as it
moves through the branched anatomical structure, registration
errors may occur between the medical device and the computer model
of the branched anatomical structure. The registration errors may
result from errors in the kinematics used for tracking the distal
end of the medical device 110, errors associated with the sensors
used for tracking the position of the distal end, and/or errors
caused by movement of the branched anatomical structure. As a
result of these and other possible errors, misregistration of the
branched anatomical structure to the medical device 110 may
develop. As a consequence, the navigation guidance assistance
provided by the system may be in error.
[0046] FIG. 5 illustrates, as an example, a method, which is
preferably implemented by the main processor 160, for registering a
computer model of a branched anatomical structure to a medical
device so that a position of the medical device in the branched
anatomical structure may be determined as the medical device moves
through the branched anatomical structure. In this case, the
position of the medical device in the branched anatomical structure
indicates the lumen of the branched anatomical structure that the
medical device is currently in. The method may be used as a
standalone registration technique or it may be combined with other
registration techniques. For example, the method may be used for
global registration and another technique used for local
registration.
[0047] In brief, the method identifies, tracks, and correlates
lumen information that is extracted from images captured from the
perspective of a distal end of the medical device as the medical
device moves through the branched anatomical structure with
corresponding lumen information extracted from the computer model
of the branched anatomical structure to enable registration of the
distal end of the medical device to the branched anatomical
structure. In particular, the method compares information of
topologies, feature attributes, behaviors, and/or relationships of
lumens seen in a sequence of captured images with corresponding
information of topologies, feature attributes, behaviors, and/or
relationships of lumens in the computer model of the branched
anatomical structure. A most likely match between a path of lumens
seen in the sequence of captured images and a path of lumens in the
computer model of the branched anatomical structure is then
determined to indicate which lumen of the branched anatomical
structure the medical device is currently in.
[0048] In a first part of the method, comprising blocks 501-504,
blobs are extracted, labeled and tracked in sequentially received
images until a change occurs in the set of blobs appearing in the
images. Each extracted blob is extracted so as to indicate a
currently entered or enterable lumen in the branched anatomical
structure. For example, lumens in the captured image presumably
appear as darker holes in the image due to the fact that the
further side of the lumen is less illuminated. Image processing
techniques can then be used to extract the salient dark regions in
the image. These extracted salient dark regions are thus referred
to herein as "blobs". Because the extracted blobs indicate lumens,
the terms blob and lumen may be used interchangeably herein. A
sequence of images that are topologically equivalent (e.g., a
sequence in which the same set of blobs or lumens appears) is
referred to herein as a "tracklet". Thus, the path that the medical
device moves through the branched anatomical structure is a
sequence of tracklets.
[0049] In block 501, the method receives an image that has been
captured, for example by the image capturing element 114, from the
perspective of the distal end of the medical device as the medical
device moves through the branched anatomical structure. An example
of such a received image is shown in FIG. 6 in which a bifurcation
in the lungs is shown and downstream bifurcations appear in each of
the lumens of the bifurcation. The received image may be a
monoscopic image (as shown) or a stereoscopic image providing depth
information.
[0050] In block 502, the method processes the received image to
extract blobs using any suitable one of a number of well known blob
detection algorithms such as Maximally Stable Extremal Regions
(MSER). An example of a processed image is shown in FIG. 7 in which
the coarse outlines of blobs extracted from the received image of
FIG. 6 appear. Also shown in FIG. 7 is a line segment for each
bifurcation which has one end connected to the centroid of one
lumen and the other end connected to the centroid of the other
lumen of the bifurcation.
[0051] To better define the blobs as closed curves that are
indicative of lumens in the branched anatomical structure,
additional image processing may be desirable using well known image
filtering and/or edge detection techniques before and/or during
blob extraction. Lens distortion correction may also be performed
to help the association of the image observation in the captured
images with the computer model of the branched anatomical
structure. FIGS. 8-20 illustrate various blob topologies that may
result from such processing.
[0052] For example, FIG. 8 illustrates the medical device moving
through a lumen 801 where no downstream node is visible at the
time. FIG. 9 illustrates an example of the medical device moving
through a lumen 901 where a downstream bifurcation comprising
lumens 902, 903 is visible. FIG. 10 illustrates an example of the
medical device moving through a lumen 1001 where a downstream
trifurcation comprising lumens 1002, 1003, 1004 is visible. FIG. 11
illustrates another example of the medical device moving through a
lumen 1101 where a downstream trifurcation comprising lumens 1102,
1103, 1104 is visible at a different orientation than the
trifurcation of FIG. 10. FIGS. 12 and 13 illustrate a sequence of
images. FIG. 12 illustrates an example of the medical device moving
through a lumen 1201 where a first downstream bifurcation
comprising lumens 1202, 1203 is visible and a second downstream
bifurcation comprising lumens 1204, 1205 is visible within the
lumen 1203. FIG. 13 illustrates an example of the medical device
having moved into lumen 1203 so that lumens 1201, 1202 are no
longer visible.
[0053] In block 503, each of the extracted blobs is tracked. If the
blob is being seen for the first time in the current sequence of
images being received by looping through blocks 501-504, then it is
labeled with a previously unassigned reference number so that the
method may keep track of the blob (or more particularly, the lumen
that the blob indicates). On the other hand, if the blob was seen
in a previously received image, then it has already been assigned a
label and it is tracked instead. Tracking in this context means
identifying the blob from one image to another. Generally, such
identification may be simply made by taking into account the
relative positions that blobs appear in the images.
[0054] As an example of blob tracking, FIGS. 18 and 19 illustrate
two sequential images in a tracklet. In FIG. 18, a bifurcation
comprising blobs/lumens 1802, 1803 is seen from a distance while
the medical device is moving through a lumen 1801 towards the
bifurcation. In FIG. 19, the same bifurcation is seen from a closer
distance as the medical device has moved through the lumen 1801
with its distal end being steered towards blob/lumen 1803. In this
example, blobs 1802, 1803 may be easily tracked between the images
shown in FIGS. 18 and 19, because their positions relative to each
other remain the same (i.e., blob 1802 continues to be the left
lumen of the bifurcation and blob 1803 continues to be the right
lumen of the bifurcation).
[0055] As another slightly more complicated example of blob
identification, previously described FIGS. 12 and 13 illustrate a
different pair of sequential images resulting in a transition
between adjacent tracklets. By comparison of FIGS. 12 and 13, it is
most probable that in FIG. 13, the medical device has moved into
one of the lumens 1202, 1203 since the total number of lumens
(i.e., blobs) has decreased. Further, it is most probable that the
medical device has moved into lumen 1203. This is because a
downstream bifurcation is visible in FIG. 13 and a downstream
bifurcation is only visible in lumen 1203 according to the earlier
received image shown in FIG. 12.
[0056] In block 504, the method makes a determination whether
topological equivalence is maintained in the current sequence of
images being received as the method loops through blocks 501-504.
If the determination in block 504 is a YES, then in block 505, the
method associates the received image with the current tracklet and
loops back to block 501 to add to the current sequence of images.
On the other hand, if the determination in block 504 is NO, then in
block 506, a new tracklet is defined starting with the most
recently received image. The current tracklet is then closed so
that it includes only the current sequence of images up to, but not
including the most recently received image (i.e., the image first
showing a change in the set of blobs by a labeled blob disappearing
or a new blob appearing).
[0057] As an example, a current tracklet may result in extracted
blobs as shown in FIG. 12. After the medical device moves into
lumen/blob 1203, a new tracklet would then result in extracted
blobs as shown in FIG. 13.
[0058] When the number of blobs decreases between a pair of
sequential images, a primary assumption is the medical device has
passed by a node (e.g., bifurcation or trifurcation) and entered
into a lumen branching out from the node, such as described in
reference to FIGS. 12 and 13. On the other hand, when the number of
blobs increases between a pair of sequential images, the primary
assumption is the medical device has moved forward in a lumen to a
point where a downstream node, such as a bifurcation or
trifurcation, is visible. These primary assumptions may be refuted
by knowledge of the progression of the tracked blobs in sequential
images and/or sensor data. For example, a disappearing blob
approaching towards and exiting from a boundary of the image in a
progression of sequentially captured images affirms the primary
assumption that the medical device has passed a node, whereas a
different behavior of a disappearing blob in the sequence of images
may refute the primary assumption. As another example, a sensor
capable of sensing the insertion and retraction of the medical
device into and out of the branched anatomical structure can refute
the primary assumption with a contrary indication.
[0059] One potential problem with such primary assumptions is the
determination of false negatives (i.e., an incorrect determination
that the number of blobs/lumens in the current image is less than
the number of blobs/lumens in the prior image) and false positives
(i.e., an incorrect determination that the number of blobs/lumens
in the current image is more than the number of blobs/lumens in the
prior image). Using FIG. 9 as a basis for an example of a false
negative as shown in FIG. 14, an obstruction 1404 blends in so well
with background tissue that it appears the blob 902 behind it has
disappeared. Again using FIG. 9 as a basis for an example of a
false positive as shown in FIG. 15, a foreign object 1504 appears
within the lumen 901.
[0060] Depending upon the type and cause of such false negatives
and positives, additional processing and/or logic may be performed
by the method to eliminate false determinations. For example, a
single downstream blob/lumen positioned off to one side as shown in
FIG. 14 seems unlikely, so the false negative should be ignored. As
another example, although three blobs/lumens indicating a
trifurcation may appear as shown in FIG. 15, a comparison of
feature attributes for the extracted blobs with corresponding
feature attributes for trifurcation blobs in synthetic images
generated from the computer model of the branched anatomical
structure may expose a false positive determination in the current
image. In addition to blob topologies and/or characteristics that
are inconsistent with those expected from the computer model of the
branched anatomical structure, a requirement for temporal
consistency between successive images may also expose false
negatives and positives. False negatives and positives resulting
from backing up the medical device and going forward again may also
be detected by analyzing successive images. As an example of
backing up and going forward again, if the right lumen of a
bifurcation gradually disappears in successive images to the right
of the field of view resulting in a single lumen being seen, then a
second lumen gradually appears in successive images from the right
of the field of view, this may indicate a situation where the
medical device has first gone into the left lumen of the
bifurcation, then backed out of the left lumen so that the right
lumen of the bifurcation is once again visible.
[0061] In the second part of the method, comprising blocks 507-510,
the current sequence of images (i.e., the current tracklet) is
processed and a most likely match with a path of lumens in the
computer model of the branched anatomical structure is determined
by taking into account the current tracklet and its adjacent
tracklets. There are a number of potential paths that the medical
device may take as it moves through lumens of the branched
anatomical structure. Each tracklet identified in the first part of
the method may correspond to a different lumen of the branched
anatomical structure. Thus, a sequence of tracklets which have been
identified by looping through blocks 501-510 as the medical device
moves through the branched anatomical structure provide a pictorial
history of images captured as the medical device moves along a path
in the branched anatomical structure. Each time a new tracklet is
identified, it is added to the sequence of tracklets and the
updated sequence of tracklets is matched against the potential
paths that the medical device may take as it moves through lumens
of the branched anatomical structure. It may happen that two or
more potential paths are close matches to the updated sequence. In
that case, although the closest match may be designated the most
likely match up to that point, when a next tracklet is identified
after looping back through blocks 501-504, the updated sequence of
tracklets may match more closely with one of the other potential
matches to indicate a "true" match versus the previously identified
"false" match. Thus, a "false match" error is self-corrected in the
method. In contrast, prior art methods generally require the
operator to recognize when a "false match" has occurred and to
correct the error, which may be very difficult and time consuming
for the operator.
[0062] In block 507, the method identifies potential matches of the
current tracklet with locations within the computer model of the
branched anatomical structure. Although the current tracklet is not
limited to being a tracklet which has ended by the medical device
passing through a node of the branched anatomical structure by
moving through one lumen leading up to the node and entering into
another lumen branching out from the node, this type of tracklet is
commonly encountered and will be used for describing the
method.
[0063] The potential matches represent a limited set of nodes
(e.g., bifurcations and trifurcations) from the universe of all
nodes in the branched anatomical structure. Numerous filtering
criteria may be used to limit the set of nodes (e.g., by
identifying potentially matching nodes or by ruling out certain
nodes). One such filtering criterion is topological equivalence
between the current tracklet and potential matches of synthetic
images which may be generated from the computer model of the
branched anatomical structure. The synthetic images represent views
within the three-dimensional computer model of the branched
anatomical structure that correspond to images captured at the
distal end of the medical device as it moves through the branched
anatomical structure. Although such synthetic images are described
as being used herein, it is to be appreciated that information
indicative of such synthetic images may be extracted instead from
the computer model. As an example of using topological equivalence
as a filtering criterion, if the tracklet indicates the node being
passed is a bifurcation, then only bifurcation nodes may be
included. Conversely, if the tracklet indicates the node being
passed is a trifurcation, then only trifurcation nodes may be
included. As another example, if the tracklet indicates that a
downstream node is visible through one of the lumens of an upstream
node, then only synthetic images indicating nodes satisfying such a
topology may be included in the set of potential matches.
Topological equivalence is preferably a threshold requirement for a
synthetic image to be considered as a potential match to the
tracklet.
[0064] Other filtering criteria for limiting the set of nodes to
generate potential matches with the current tracklet may use
available sensor information. As an example, FIG. 22 illustrates a
schematic diagram in which the medical device 110, as controlled by
interface 170, moves through an opening 2201 and along a curved
path 2202 in an anatomical structure represented by block 2200.
Four optional sensors 2221, 2222, 2223, 2224 provide information
that may be employed to further limit the set of potential node
matches. Signal communication cables or lines 132 transfer
information from the distal end sensors 2222, 2223 back to the
interface 170. As shown in FIG. 1, the sensor information is then
provided to a sensor(s) processor 130, which may include an
analog-to-digital (A/D) converter, so that the information may be
converted into a form suitable for processing by the main processor
160.
[0065] An insertion sensor 2221 may optionally be provided to
provide information on how much the medical device has been
inserted into the anatomical structure 2200. With this information,
only nodes within a threshold distance to an insertion depth from
the opening will be included in the set of potential matches. This
type of sensor is especially useful for detecting a direction
reversal situation, such as when a medical device has backed out of
a bifurcation lumen after previously entering the lumen. Typically,
such an insertion sensor 2221 may conventionally be provided in or
adjacent to the interface 170 to detect linear movement of the
medical device 110 into and out of the anatomical structure
2200.
[0066] A position sensor 2222 may optionally be provided at the
distal end 111 of the medical device 110 to provide position
information in three-dimensional space (such as point X.sub.K,
Y.sub.K, Z.sub.K illustrated in FIG. 23). With this information,
only nodes within a threshold distance (such as within an
uncertainty range of the sensor) from the sensed position will be
included in the set of potential matches.
[0067] An orientation sensor 2223 may optionally be provided at the
distal end 111 of the medical device 110 to provide orientation
information for the distal end. For example, as illustrated in FIG.
24, information of an angle .phi. indicating how much a line 2411
corresponding to the horizontal line of the received image (which
has been captured within lumen 2413) deviates from a reference line
2412 that is perpendicular to a gravity vector may be provided by
the orientation sensor 2223. With this information, orientations of
blobs in the received images may be adjusted to more accurately
reflect what they are expected to look like in synthetic images
generated from the computer model of the branched anatomical
structure. This orientation information is particularly useful to
prevent possible errors in determining which of the blobs is to the
left or to the right of the other (i.e., avoidance of left/right
reversal errors) since there will only be one way to associate a
two-lumen bifurcation.
[0068] A roll sensor 2224 may optionally be provided to provide
information of how much the medical device 110 has been rotated
about its longitudinal axis. This information may be useful to
estimate an orientation of the distal end when an orientation
sensor is unavailable. Typically, such a roll sensor 2224 may
conventionally be provided in or adjacent to the interface 170 to
detect rotation of the medical device 110 about its longitudinal
axis.
[0069] In other embodiments, additional filtering data can include
historical user inputs that can used to narrow the universe of node
possibilities. For example, navigation control inputs (e.g.,
steering commands at one or more nodes) can be used as basis for
identifying the most likely lumen(s) entered at a given node(s).
This directional indication(s) can supplement or be used in place
of blob tracking techniques, for example to enhance path
tracking/lumen identification. Alternatively, a user can provide
inputs specifically for node filtering purposes (e.g., specific
anatomic features such as lumens, bifurcations, and/or other tissue
structures can be manually "marked", labeled, or otherwise
identified by the user for subsequent incorporation into the node
filtering algorithm). This manual identification of anatomical
landmarks can significantly reduce the range of node
possibilities.
[0070] Quantitative comparisons may then be performed on the
remaining members of the set after filtering the set of potential
matches using a topological equivalence criterion and optionally,
any available sensor information and/or user inputs. One technique
for performing such quantitative comparisons uses feature
attributes determined from the current tracklet and corresponding
feature attributes extracted from the computer model of the
anatomical structure for nodes of the computer model. As an
example, feature attributes of the current tracklet should be
reasonably close to feature attributes of synthetic images
corresponding to the nodes of the computer model in order for the
nodes and their synthetic images to be included in the set of
potential matches. Since feature attributes may change as the
distal end of the medical device is steered towards and through one
of the lumens branching out from the node, feature attributes for
the current tracklet are preferably determined from images captured
prior to such distal end steering taking place.
[0071] To perform such quantitative comparisons between the
tracklet and the potential matches, the method determines feature
attributes for blobs in the current tracklet. The feature
attributes are used to distinguish different bifurcations from one
another and different trifurcations from one another in the
branched anatomical structure.
[0072] As an example, FIG. 16 illustrates a bifurcation having left
and right blobs/lumens 1602, 1603 as seen in blob/lumen 1601.
Measurements of the heights, "a" and "c", and widths, "b" and "d",
of the left and right lumens may be made according to a common
scale (such as pixels in the received image) and ratios calculated
to determine feature attributes, such as the ratio of heights,
"a/c", the ratio of widths, "b/d", aspect ratio for the left
blob/lumen, "b/a", and aspect ratio for the right blob/lumen,
"d/c". As may be appreciated, the aspect ratios in particular may
be very informative since they may indicate an angle at which the
lumen extends away from the node.
[0073] As another example, FIG. 17 illustrates a bifurcation having
left and right blobs/lumens 1702, 1703 as seen in blob/lumen 1701.
As a reference line, a horizontal line 1707 on a fixed horizontal
plane (i.e., a plane perpendicular to a gravity vector) is shown.
Another line 1706 is shown which extends through centroids 1704,
1705 of the blobs/lumens 1702, 1703. An angle, .theta., may then be
measured or otherwise determined between lines 1707, 1706 to serve
as a feature attribute of the bifurcation. The length of the line
segment connecting the centroids 1704, 1705 may also be used for
generating other feature attributes of the bifurcation. For
example, ratios may be calculated by dividing the heights and
widths of each blob by the length of the line segment. An angle,
.PSI., may also be measured or otherwise determined between height
lines 1712, 1713 of blobs 1702, 1703 to serve as another feature
attribute of the bifurcation. A similar angle may also be measured
or otherwise determined between width lines of the blobs to serve
as yet another feature attribute of the bifurcation.
[0074] As another example, FIG. 20 illustrates blobs extracted from
the received image shown in FIG. 6. The extracted blobs in this
case may result from additional image processing of the blobs shown
in FIG. 7. A foreground bifurcation includes blobs/lumens 2001,
2002. Looking into blob/lumen 2001, a downstream bifurcation
including blobs/lumens 2011, 2012 is seen. Likewise, looking into
blob/lumen 2002, another downstream bifurcation including
blobs/lumens 2021, 2022 is seen. A line segment 2003 connects the
centroids of blobs/lumens 2001, 2002, a line segment 2013 connects
the centroids of blobs/lumens 2011, 2012, and a line segment 2023
connects the centroids of blobs/lumens 2021, 2022. Several feature
attributes may be defined and quantified using the line segments
2003, 2013, 2023. For example, the length of each line segment may
used in a ratio with heights and/or widths of their respective
blobs to define feature attributes for the blobs. As another
example, angles between pairs of line segments may define other
feature attributes of the image. Note that measurements of these
feature attributes are orientation independent (i.e., result in the
same values even if the image capturing device is rotated so as to
change the orientation of the view). Thus, feature attributes of
this kind are advantageous over orientation dependent feature
attributes such as shown in FIG. 17 which require a fixed reference
line.
[0075] As still another example, FIG. 21 illustrates the use of an
extracted blob as a template to be compared with corresponding
blobs of synthetic images generated from the computer model of the
branched anatomical structure (e.g., shape of left blob/lumen of
bifurcation in captured image compared to shape of left lumen in
synthetic image of a bifurcation from the computer model of the
branched anatomic structure). In this way, blob templates may also
be used as feature attributes. As may be appreciated, many feature
attributes beyond those described herein may also be defined using
various lines and angles which are indicative of the relative
shapes, sizes, and orientations of the extracted blobs. Further,
rather than determining the feature attributes of extracted blobs
for only one received image, the feature attributes may be
determined for the tracklet (i.e., sequence of topologically
equivalent images) by averaging the determined information and/or
eliminating outliers in the determined information for the images
in the tracklet. In addition to feature attributes determined from
topological and geometrical features of extracted blobs as
described above, other feature attributes may be defined and used
in the methods described herein which are related to feature points
identified in captured images. As an example, a scale invariant
feature transform (SIFT) may be used to extract feature points from
the captured images. The feature attributes used to compare the
real image and the computer model can also be three-dimensional
features extracted from multiple image frames using Structure from
Motion (SfM) techniques. See, e.g., Richard Hartley and Andrew
Zisserman, Multiple View Geometry in Computer Vision, Cambridge
University Press, 2.sup.nd Edition, 2004. Also, see, e.g., C.
Tomasi and T. Kanade, "Factoring image sequences into shape and
motion," Proceedings of the IEEE Workshop on Visual Motion, pages
21-28, Princeton, N.J., October 1991.
[0076] In addition to using feature attributes to distinguish
bifurcations and trifurcations in a branched anatomical structure,
feature attributes determined for a sequence of images may also be
used for detecting false negatives and false positives as part of
performing block 504. As an example, FIGS. 18 and 19 illustrate a
sequence of images in which the image shown in FIG. 18 is received
prior to the image shown in FIG. 19. As the distal end of the
medical device is steered towards the right lumen 1803, the aspect
ratio of the left lumen is expected to get smaller while the aspect
ratio of the right lumen is expected to get larger. Other feature
attributes may also be defined to indicate the direction of
steering. For example, if the centroid of the blob/lumen 1802 is
"A", the centroid of the blob/lumen 1803 is "B" and the centroid of
the lumen 1801 through which the medical device is currently moving
through is "C", then as the distal end of the medical device is
steered towards the right lumen 1803, the distance between the
centroids of blobs/lumens 1801, 1803 is expected to get smaller.
Thus, by tracking the distances between the centroids, A and B, of
the blobs/lumens 1802, 1803, to the centroid, C, of the lumen 1801,
the direction that the distal end of the medical device is being
steered may be determined. This information may be used to avoid
false negatives as described above.
[0077] The quantitative comparisons performed by the method in
block 507 indicate how "close" the feature attributes of blobs in
the current tracklet match corresponding feature attributes of
blobs in synthetic images of the potential node matches. As an
example, a difference between the aspect ratio of the left blob of
a bifurcation in the current tracklet and the aspect ratio of the
left blob of a bifurcation in each of the synthetic images of the
potential matches may be calculated. As another example, as shown
in FIG. 21, differences (e.g., distances between opposing arrows at
Z.sub.m and Y.sub.n) between the shape of an extracted blob 2100
and a synthetic blob 2101, which has been generated from the
computer model of the branched anatomical structure, may be
determined and used for quantitative comparison purposes. The lower
the difference in this case, the better the match between the
tracklet and the node being represented in the synthetic image.
[0078] The quantitative comparisons are preferably converted to a
probability or confidence score for each potential match or
hypothesis. As an example, a likelihood of a j.sup.th hypothesis
(h.sub.i,j) for the i.sup.th tracklet (t.sub.i) may be expressed as
follows:
Pr(O.sub.i|t.sub.i=h.sub.i,j)=Pr(O.sub.i.sup.(I)|t.sub.i=h.sub.i,j)Pr(O.-
sub.i.sup.(S)|t.sub.i=h.sub.i,j) (1)
where O.sub.i represents all the observations associated with
t.sub.i; O.sub.i.sup.i.sup.(I) represents all image observations
associated with t.sub.i; and O.sub.i.sup.(S) represents all
available sensor observations associated with t.sub.i.
[0079] After determining the quantitative comparisons for each of
the members in the set of potential matches, the method stores the
results of the quantitative comparisons for the current tracklet in
a memory such as memory 161.
[0080] In block 508, the method identifies potential matches for
the transition between the current tracklet to the new tracklet.
Similar filtering techniques as described in block 507 may be
performed in block 508. For example, a synthetic image
corresponding to a view within a lumen may be generated from the
computer model of the branched anatomical structure for each lumen
in the branched anatomical structure. Captured images associated
with the current tracklet and the next tracklet may then be
compared with pairs of synthetic images corresponding to pairs of
connected lumens within the branched anatomical structure for
topological equivalence. Available sensor information may also be
used to eliminate pairs of connected lumens within the branched
anatomical structure from the set of potential matches for the
transition between the current and new tracklets. Quantitative
comparisons are then performed on the remaining members of the set
of potential matches and preferably converted to a transition
probability or confidence score for each potential match or
hypothesis.
[0081] After determining the quantitative comparisons for each of
the members in the set of potential transition matches, the method
stores the results of the quantitative comparisons for the
transitions from the current tracklet to the new tracklet in a
memory such as memory 161.
[0082] In block 509, the method then determines a most likely match
for the sequence of tracklets resulting from looping through blocks
501-510 by taking into account information of one or more adjacent
tracklets in the sequence of tracklets making up the path of the
medical device as it moves through the branched anatomical
structure. Information of prior tracklets has been stored in the
memory during the processing of blocks 507 and 508 for those
tracklets while looping through blocks 501-510.
[0083] The likelihood of a hypothetical path "T" for the current
tracklet is a combination of the likelihood of the individual
hypotheses for the current tracklet and the transition
probabilities of adjacent tracklets. The transition probability
that the medical device switches state from the j.sup.th hypothesis
at the i.sup.th tracklet to the k.sup.th hypothesis at the
(i+1).sup.th tracklet may be expressed as follows:
q.sub.i,j,k=Pr(t.sub.i+1=h.sub.i+1,k|t.sub.i=h.sub.i,j) (2)
which is a combination of static prior knowledge of connectivity in
the computer model of the branched anatomical structure and dynamic
knowledge of which lumen branching out of a node indicated by the
current tracklet was entered by the medical device. The static
prior knowledge of connectivity in the computer model of the
branched anatomical structure may be obtained, for example, by
performing a patient scan using a suitable imaging modality to
capture image slices from which an accurate three-dimensional
computer model of the branched anatomical structure may be
generated. The dynamic knowledge of which lumen branching out of a
node indicated by the current tracklet was entered by the medical
device may be obtained, for example, by programming the processor
with artificial reasoning which follows analytical reasoning as
described herein.
[0084] The most likely match for the current sequence of tracklets
is then based upon the most likely hypothetical path taken by the
medical device. As an example, the solution may be the path that
maximizes the combined probability expressed as follows:
T*=argmax.sub.TPr(O|T) (3)
[0085] One technique for efficiently solving the above discrete
path optimization problem is using Dynamic Programming (DP),
similar to the inference algorithm in the Hidden Markov Model
(HMM).
[0086] After completing processing of block 509, the method then
redefines the new tracklet as the current tracklet in block 510 and
loops back to block 501 to expand on and process the new current
tracklet which was previously detected in block 504. The method
then continues to loop through blocks 501-510 as the medical device
continues to move through the branched anatomical structure.
[0087] Although the various aspects of the present invention have
been described with respect to one or more embodiments, it will be
understood that the invention is entitled to full protection within
the full scope of the appended claims.
* * * * *