U.S. patent application number 16/183772 was filed with the patent office on 2019-03-14 for systems and methods for quantifying regional fissure features.
The applicant listed for this patent is Vida Diagnostics, Inc.. Invention is credited to Philippe Raffy, Youbing Yin.
Application Number | 20190076052 16/183772 |
Document ID | / |
Family ID | 56098342 |
Filed Date | 2019-03-14 |
![](/patent/app/20190076052/US20190076052A1-20190314-D00000.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00001.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00002.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00003.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00004.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00005.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00006.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00007.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00008.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00009.png)
![](/patent/app/20190076052/US20190076052A1-20190314-D00010.png)
View All Diagrams
United States Patent
Application |
20190076052 |
Kind Code |
A1 |
Yin; Youbing ; et
al. |
March 14, 2019 |
SYSTEMS AND METHODS FOR QUANTIFYING REGIONAL FISSURE FEATURES
Abstract
Analysis of pulmonary scans representative of a patient's
pulmonary structure can be used to classify a patient into one or
more of a plurality of populations. The patient's scan can be
mapped to a reference domain and analyzed to determine one or more
fissure features associated with a plurality of regions in the
reference domain. Comparison of the determined fissure features
with a plurality of fissure atlases, each associated with different
population, can be performed to classify the patient into one or
more of the populations. Data from different fissure atlases can be
compared to determine regions in the fissure atlases that
distinguish one population from another. Such distinguishing
regions can improve the ability to classify the patient while
reducing errors based on false classifications.
Inventors: |
Yin; Youbing; (Coralville,
IA) ; Raffy; Philippe; (Edina, MN) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Vida Diagnostics, Inc. |
Coralville |
IA |
US |
|
|
Family ID: |
56098342 |
Appl. No.: |
16/183772 |
Filed: |
November 8, 2018 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
15148767 |
May 6, 2016 |
10165964 |
|
|
16183772 |
|
|
|
|
62159098 |
May 8, 2015 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
A61B 6/037 20130101;
G06T 2207/30061 20130101; G06F 19/321 20130101; G16H 30/40
20180101; G06T 7/0014 20130101; G16H 50/80 20180101; A61B 6/5217
20130101; A61B 5/08 20130101; G06K 9/6267 20130101; G16H 30/20
20180101; G06K 2209/05 20130101; A61B 5/004 20130101; G06T
2207/10072 20130101; A61B 6/50 20130101; A61B 5/055 20130101; A61B
6/032 20130101 |
International
Class: |
A61B 5/08 20060101
A61B005/08; A61B 6/03 20060101 A61B006/03; A61B 5/055 20060101
A61B005/055; G06K 9/62 20060101 G06K009/62; A61B 6/00 20060101
A61B006/00; G06T 7/00 20060101 G06T007/00 |
Claims
1. A method for analyzing a patient based on a volumetric pulmonary
scan comprising: receiving a first volumetric pulmonary scan
representative of the patient's pulmonary structure; mapping the
first volumetric pulmonary scan to a reference domain; determining
one or more fissure features associated with a plurality of regions
in the reference domain and within the volumetric pulmonary scan;
comparing, in each of the plurality of regions, the determined one
or more fissure features to a plurality of fissure atlases, each
atlas comprising statistical data associated with the one or more
fissure features for a different population based on regional
analysis of volumetric pulmonary scans of the population;
classifying the patient into one or more of a plurality of
populations based on the comparison of the determined one or more
fissure features.
2. The method of claim 1, further comprising: after classifying the
patient into a population, adding the registered first volumetric
pulmonary scan to the fissure atlas associated with that
population; and updating the statistical data in the fissure atlas
to include the first volumetric pulmonary scan.
3. The method of claim 2, wherein updating the statistical data in
the fissure atlas to include the first volumetric pulmonary scan
data comprises weighting the contribution from the first volumetric
pulmonary scan data in the fissure atlas at each of a plurality of
regions.
4. The method of claim 1, wherein the plurality of populations
comprises responders to a particular therapy and non-responders to
the particular therapy.
5. The method of claim 1, wherein the plurality of populations
comprises patients with collateral ventilation and patients without
collateral ventilation.
6. The method of claim 1, wherein the first volumetric pulmonary
scan comprises a CT scan, an MRI scan, or a PET scan.
7. The method of claim 1, further comprising: identifying, for a
given fissure feature, distinguishing regions in the volumetric
pulmonary scans wherein the fissure feature is substantially
different among different populations; and wherein comparing the
determined one or more fissure features to a plurality of fissure
atlases comprises, for the given fissure feature, comparing the
fissure feature in the first volumetric pulmonary scan data to the
plurality of fissure atlases in the identified distinguishing
regions.
8. The method of claim 7, wherein a set of distinguishing regions
are identified corresponding to a pair of fissure atlases and a
single fissure feature; and wherein for a given fissure feature and
pair of populations, the distinguishing regions comprise regions in
which the difference between the fissure feature in the pair of
fissure atlases is considered significant.
9. The method of claim 8, further comprising: in each of a
plurality of regions, statistically analyzing the fissure feature
in a first fissure atlas and the fissure feature in a second
fissure atlas and determining the regions in which the difference
between the fissure feature in the two atlases is significant in
order to determine distinguishing regions corresponding to the
first and second fissure atlases and the first fissure feature.
10. The method of claim 1, wherein the one or more fissure features
comprises a feature selected from the group consisting of: fissure
integrity, fissure curvature, airway related measurements,
deformation field, local vascularity related measurements, and
local parenchymal related features.
11. A system for assessing a volumetric pulmonary scan of a patient
comprising: a database comprising a plurality of fissure atlases,
each fissure atlas comprising statistical data regarding one or
more fissure features in a plurality of volumetric pulmonary
regions for a given population; and a processor configured to:
statistically analyze the one or more fissure features in the
plurality of volumetric pulmonary regions to determine, for each
fissure feature, one or more distinguishing regions; receive a
first set of volumetric pulmonary scan data representative of the
volumetric structure of a patient's lungs; determine the one or
more fissure features in the first set of volumetric pulmonary scan
data at a plurality of regions; compare the determined one or more
fissure features to the plurality of fissure atlases in the
database; and classify the first set of volumetric pulmonary scan
data into one or more of the populations based on the
comparison.
12. The system of claim 11, wherein the processor is further
configured to: after classifying the patient into a population, add
the registered first volumetric pulmonary scan to the fissure atlas
associated with that population; and update the statistical data in
the fissure atlas to include the first volumetric pulmonary
scan.
13. The system of claim 11, wherein the processor is further
configured to: analyze the one or more fissure features in each of
a plurality of volumetric pulmonary regions in each of the
plurality of fissure atlases; determine, for each fissure feature,
one or more distinguishing regions in the plurality of regions, the
one or more distinguishing regions being regions in which the
fissure feature differs significantly between at least two fissure
atlases; and associate the determined distinguishing regions with
the corresponding fissure feature and fissure atlases.
14. The system of claim 13, wherein classifying the first set of
volumetric pulmonary scan data into one or more of the populations
based on the comparison comprises comparing determined fissure
features from the first set of volumetric pulmonary scan data to
corresponding fissure atlases in the determined distinguishing
regions.
15. The system of claim 11, wherein the statistical data in each of
the fissure atlases is associated with regions in a reference
domain, and wherein the processor is further configured to map the
first set of volumetric pulmonary scan data to the reference
domain.
16. The system of claim 11, wherein the processor further
configured to identify regions in the reference domain, and wherein
the identified regions comprise lobes, sub-lobes, and/or
custom-defined regions.
17. A method comprising: acquiring a first plurality of
three-dimensional pulmonary models, each of the models being
representative of the pulmonary structure of a patient belonging to
a first population; registering each of the three-dimensional
models in the first plurality of three-dimensional models to a
reference domain to create a first atlas; statistically analyzing
the first plurality of registered three-dimensional models in the
first atlas to determine a first fissure feature at each of a
plurality of regions in the first atlas; acquiring a second
plurality of three-dimensional pulmonary modes, each of the models
being representative of the pulmonary structure of a patient belong
to a second population, the second population being different from
the first; registering each of the three-dimensional models in the
second plurality of three-dimensional models to the reference
domain to create a second atlas; statistically analyzing the second
plurality of registered three-dimensional models in the second
atlas to determine the first fissure feature at each of a plurality
of regions in the second atlas; for each of a plurality of regions
in the first and second atlases, comparing the first fissure
feature of the first atlas to the first fissure feature of the
second atlas to determine regions in which the first fissure
feature differs significantly between the first and second
populations and considering such regions to be distinguishing
regions associated with the first population, the second
population, and the first fissure feature; receiving a diagnostic
three-dimensional pulmonary model of a first patient's lungs;
registering the received diagnostic three-dimensional pulmonary
model to the reference domain; analyzing the first fissure feature
in the distinguishing regions in the diagnostic three-dimensional
pulmonary model; and predicting if the first patient is in the
first population or the second population based on the analyzed
first fissure feature in the distinguishing regions in the
diagnostic three-dimensional pulmonary model.
18. The method of claim 17, wherein: the first population comprises
patients who responded positively to a particular therapy; the
second population comprises patients who did not respond positively
to the particular therapy; and the comparing the first fissure
feature of the first atlas to the first fissure feature of the
second atlas for each of the plurality of regions comprises
determining a correlation between the first fissure feature in each
region and the effectiveness of the particular therapy.
19. The method of claim 18, wherein determining a correlation
between the first fissure feature in each region and the
effectiveness of the particular therapy comprises determining in
which of the plurality of regions the difference in the first
fissure feature between the first population and the second
population is greatest.
20. The method of claim 18, wherein the first fissure feature
comprises a feature selected from the group consisting of: fissure
integrity, fissure curvature, airway related measurements,
deformation field, local vascularity related measurements, and
local parenchymal related features.
Description
CROSS REFERENCES
[0001] This application is a continuation application of U.S.
application Ser. No. 15/148,767, filed May 6, 2016, which claims
the benefit of U.S. Provisional Application No. 62/159,098, filed
May 8, 2015, the content of which is hereby incorporated by
reference in its entirety.
FIELD OF THE INVENTION
[0002] Embodiments of the invention generally relate to
visualization and characterization of pulmonary lobar fissures and
regional fissure features.
BACKGROUND OF THE INVENTION
[0003] Severe emphysema is a debilitating disease that limits
quality of life of patients and represents an end state of Chronic
Obstructive Pulmonary Disease (COPD). It is believed that 3.5
million people in the US have the severe emphysematous form of
COPD, and it is increasing in both prevalence and mortality.
Current treatment methods for severe emphysema include lung volume
reduction (LVR) surgery, which is highly invasive, and can be risky
and uncomfortable for the patient. New treatment methods for
treating emphysema include bronchoscopy guided LVR (BLVR) devices
such as one-way valves that aim to close off ventilation to the
diseased regions of the lung, but maintain ventilation to healthier
lung. Bronchoscopy-guided techniques have the promise to be less
invasive, less costly and more highly accurate treatments for
patients with severe disease and improve the quality of life of
severe emphysema patients.
[0004] Emphysema can present itself in various disease forms (i.e.,
phenotypes). Predicting the right treatment for these patients at
the appropriate time in the disease process likely depends on the
phenotype of the disease. Imaging techniques provide an in-vivo
mechanism to objectively quantify and characterize disease
phenotypes and can be used in the patient selection process for the
various procedural options. Quantitative imaging biomarkers can be
used to effectively phenotype disease and therefore predict those
patients most likely to respond to the targeted treatment options.
By triaging patients to the appropriate therapy, there exists a
greater promise for a positive impact on patient outcome, reduced
healthcare costs, and replacing more invasive procedures like LVR
surgery in treating patients with severe emphysema.
[0005] Fissures are important anatomical structures within lungs.
It is believed that fissures have an effect on regional lung
mechanics. For example, adjacent lobes can slide against each other
at fissure interfaces, which provide a means to reduce lung
parenchymal distortion. In addition, intact fissures play an
important role in reducing collateral ventilation among lobes and
the spread of diseases. Recently, fissure integrity has emerged as
a strong biomarker to predict the response to interventional
emphysema therapies including localized lung volume reduction. In
short, if the fissure of the lung is intact, an obstructive device
like a valve will more likely produce a seal leading to the
atelectasis (i.e., collapse) of the diseased lung sub-region.
Without an intact fissure, there is a possibility of collateral
ventilation and the likelihood of success of the procedure may be
reduced. However, accurately detecting and characterizing fissures
in diseased lungs is difficult.
[0006] Methods of detecting fissures include fitting the existing
portions of the fissures to a lobar atlas (as described in E. M.
van Rikxoort et al., "A method for the automatic quantification of
the completeness of pulmonary fissures: evaluation in a database of
subjects with severe emphysema.," European radiology, (2011): 0-7,
for example) or by an extrapolation of the existing portion of the
fissure to the absent portion (as described in J. Pu et al.,
"Computerized assessment of pulmonary fissure integrity using high
resolution CT.," Medical Physics, 37(9), (2010): 4661-4672, for
example). However, neither of these approaches makes full use of
the anatomic information available in the image data.
SUMMARY
[0007] Aspects of the disclosure are generally directed toward
systems and methods for quantifying and/or analyzing fissure
features from pulmonary scans of a patient, and analyzing the
patient based thereon. In some examples, a volumetric pulmonary
scan representative of a patient's pulmonary structure can be
mapped to a reference domain for comparison to other data sets
associated with the reference domain. Methods can include
determining one or more fissure features associated with a
plurality of regions in the reference domain. In various exemplary
embodiments, fissure features can include fissure integrity,
fissure curvature, airway related measurements, deformation field,
and/or local vascularity related measurements.
[0008] Exemplary methods can include comparing the determined one
or more fissure features to a plurality of fissure atlases. Each
fissure atlas can include statistical data associated with the one
or more fissure features for a different population. Such
statistical data can be based on regional analysis of volumetric
pulmonary scans pf the population. Based on the comparison of the
determined one of more fissure features to the plurality of fissure
atlases, the patient can be classified into one or more of a
plurality of populations.
[0009] Exemplary populations can include patients that respond
positively to a given therapy, patients who do not respond
positively to a given therapy, patients who have a diagnostic
symptom, and patients who do not have a diagnostic symptom. In some
examples, region-by-region comparison of a fissure feature between
the patient's volumetric pulmonary scan to one or more fissure
atlases can provide the necessarily information for classifying the
patient.
[0010] In some examples, a method can include identifying, for a
given fissure feature, distinguishing regions in the volumetric
pulmonary scans of a plurality of fissure atlases wherein the
fissure feature is substantially different among different
populations. Such regions can be identified based on comparisons of
like regions in different fissure atlases corresponding to
different populations. For example, in some embodiments, a region
in which a fissure feature is most different between a pair of
populations can be considered a distinguishing region with respect
to the fissure feature and the pair of populations.
[0011] In some embodiments, classifying the patient into one or
more of a plurality of populations comprises comparing fissure
features of the patient's volumetric pulmonary scan data to like
fissure features of fissure atlases in identified distinguishing
regions associated with the population(s) and fissure feature. The
patient can be classified into the population corresponding to the
fissure atlas that best matches the patient's scan data in the
distinguishing regions. In some such embodiments, scan data from
regions not considered distinguishing regions is omitted from the
analysis, minimizing a false classification of the patient based on
data potentially unrelated to the classification population.
[0012] Systems according to embodiments described herein can
include a database comprising a plurality of fissure atlases and a
processor configured to classify a first set of volumetric
pulmonary scan data into one or more populations.
BRIEF DESCRIPTION OF THE DRAWINGS
[0013] The following drawings are illustrative of particular
embodiments of the invention and therefore do not limit the scope
of the invention. The drawings are not necessarily to scale (unless
so stated) and are intended for use with the explanations in the
following detailed description. Embodiments of the invention will
hereinafter be described in conjunction with the appended drawings,
wherein like numerals denote like elements.
[0014] FIG. 1 shows a flowchart of a fissure characterization and
visualization method associated with certain embodiments of the
invention.
[0015] FIG. 2 shows a CT scan in a sagittal view in which the
fissures have been enhanced.
[0016] FIG. 3 shows three-dimensional models of surface rendering
of fissures of an emphysema patient in accordance with certain
embodiments of the invention.
[0017] FIG. 4A shows an example of a sagittal CT image of the right
lung without fissures identified and highlighted in accordance with
certain embodiments of the invention.
[0018] FIG. 4B shows an example of a sagittal CT image of the right
lung with fissures identified and highlighted in accordance with
certain embodiments of the invention.
[0019] FIG. 5 shows an example of a screen shot including
highlighting of the fissures in various two-dimensional CT images
and a corresponding three-dimensional volume rendering in
accordance with certain embodiments of the invention.
[0020] FIG. 6 shows an example of three-dimensional models of a
fissure (a)-(c) and of the sublobes surrounding the fissure
(d).
[0021] FIG. 7 shows an example of visualization of the spatial
relationship between fissures and regions of emphysema in a
three-dimensional model in accordance with certain embodiments of
the invention.
[0022] FIG. 8 shows an example of visualization of the spatial
relationship between fissures and tumors in a three-dimensional
model in accordance with certain embodiments of the invention.
[0023] FIG. 9 illustrates a fissure editing tool in use to revise
the identification of a portion of a fissure as complete or
incomplete on a CT image.
[0024] FIG. 10 illustrates a fissure editing tool in use to revise
the identification of a portion of a fissure as complete or
incomplete on a three-dimensional model of the fissure.
[0025] FIG. 11 shows a series of exemplary 3D fissure surface
models from different patients.
[0026] FIG. 12 shows a series of fissure atlas entries registered
to a single reference model.
[0027] FIG. 13 is a visual representation of the relationship
between the fissure integrity likelihood and response to a
treatment according to some embodiments.
[0028] FIG. 14 is a visual representation of the relationship
between the fissure integrity likelihood and response to a
treatment including confidence weights based on valve issues.
[0029] FIG. 15 is a process flow diagram illustrating an exemplary
method of generating a final fissure atlas.
[0030] FIG. 16 is an illustration highlighting the absolute
differences between the fissure probability maps of LVR responders
and LVR non-responders from FIG. 13.
[0031] FIG. 17 is a diagram illustrating an exemplary process for
determining differentiating features among populations.
[0032] FIG. 18 is a process flow diagram demonstrating the
classification of a patient according to some embodiments.
DETAILED DESCRIPTION
[0033] Aspects of the invention describe a process to automate,
display, interact with and characterize the fissures of the lung in
multiple dimensions. When the human lung is imaged in vivo with an
imaging acquisition device, like CT, that image can be
reconstructed and evaluated to depict normal and diseased states.
Because of the various subclasses of disease and the various
depictions (phenotypes) of a disease entity, evaluation of lobular
regions of the lung and the fissures separating them are important
to accurately characterize disease and predict response to BLVR
therapy.
[0034] This disclosure includes methods to provide visualization of
the fissures in two and three dimensions, define the fissure
boundaries, characterize their morphologic characteristics which
may be used for identifying a disease phenotype, and visualize
regions of intact and missing fissures, and observe the difference
between normal and diseased lung in an instantaneous and automated
way to enable clinical decision making.
[0035] The left and right lungs are each divided into a plurality
of lobes by deep clefts, which are the interlobar fissures,
referred to herein simply as fissures. The outer surface of the
lungs is lined by pleura, including an inner layer which is the
visceral pleura which dips into the fissures to surround the lobes.
The fissures therefore are the join between the lobes of the lung
and are defined by the outermost surface of the lobes and the
visceral pleura at the locations where the lobes abut each other.
Therefore, although the fissure itself is actually an interface
between abutting lobes, it is the very thin layer of the lobar
interfaces that can be detected on a volumetric image and is
interpreted as being the fissure. The right lung normally includes
three lobes (the upper, middle, and lower lobes) which are divided
by two fissures, known as the oblique and the horizontal fissure.
The left lung normally includes two lobes (the upper and lower
lobes) with one fissure, the oblique fissure, between them.
[0036] The edges of the lobes and the pleura that lines the lobes
define the fissures and separate the lobes such that the
ventilation of each lobe separates from that of adjacent abutting
lobes. In addition, the pleura normally form a smooth surface,
allowing abutting lobes to slide relative to each other during
inhalation and exhalation. However, in certain disease conditions,
the pleura may become thickened or adherent. In addition, abutting
lobes may adhere to each other and the pleura and lung margins that
normally define the fissure may be lost. In such locations, the
fissure is described as "incomplete," "missing," or "absent" and
air can flow between the lobes. Various embodiments described
herein identify the fissures using volumetric radiological images
such as CT and present them visually in 2D images or in 3D models
for a user such as a clinician. In some embodiments, the absent
portions of the fissures are also identified and can also be
visualized, as by showing the "absent" portions in a color which is
distinct from the existing fissures, in a location in which they
would normally be present in a complete fissure.
[0037] Various embodiments may be performed by a lung visualization
system, which may include a processor, such as a processor in a
computer, and may also include a visual display such as a monitor
or screen. The system may also include instructions included in
software (computer readable media), stored in memory of the system,
and operable on the processor. The software may include
instructions for the processor to perform the various steps and
methods described herein, including instructions to receive patient
data including volumetric imaging data, analyze the data to
characterize the fissures, and display images including
three-dimensional images of the fissures resulting from the
analysis of the imaging data on the visual display. The software
may be incorporated into 3D pulmonary imaging software.
[0038] It should also be understood that the three-dimensional
images or models described herein are not truly created in three
dimensions, because they exist on a flat two-dimensional visual
display. Rather, the three-dimensional images described herein use
perspective and shading, with the closest portions depicted in the
foreground and more distant portions in the background, along with
the ability of the user to rotate the images in some cases and/or
to see multiple views, to show the entire volumetric volume on the
visual display. In contrast, each image in the series of the
multi-dimensional volumetric images provided by CT and MRI scans,
for example, is a two-dimensional planar image that depicts the
tissue present in a single plane or slice. These images are
typically acquired in three orthogonal planes, which are referred
to as the three orthogonal views and are typically identified as
being axial, coronal and sagittal views.
[0039] Embodiments of the invention allow the clinician to interact
with the three-dimensional model of the lungs and the
two-dimensional volumetric images associated with and used to
generate the model. For example, the three-dimensional model and
the associated two-dimensional volumetric images may be presented
in a graphical user interface on a visual display. The user may
interact with the graphical user interface, such as by selecting a
button, icon, and/or one or more locations on the images or the
model or elsewhere using a mouse, stylus, keypad, touchscreen or
other type of interface known to those of skill in the art. The
creation of the three-dimensional model may be performed by the
system including a processor with software (computer readable
media) to perform this function as well as software to permit a
user to interact with the graphical user interface, to calculate
and display desired data and images, and to perform the other
functions described herein. The system may further include the
visual display on which the graphical user interface is displayed.
The three-dimensional model and two-dimensional volumetric images
may be provided to a user (such as a clinician or researcher) as a
graphical user interface on a visual display, which may be a
computer screen, on which the images and data may be manipulated by
the user.
[0040] Examples of the embodiments may be implemented using a
combination of hardware, firmware, and/or software. For example, in
many cases some or all of the functionality provided by examples
may be implemented in executable software instructions capable of
being carried on a programmable computer processor. Likewise, some
examples of the invention include a computer-readable storage
device on which such executable software instructions are stored.
In certain examples, the system processor itself may contain
instructions to perform one or more tasks. System processing
capabilities are not limited to any specific configuration and
those skilled in the art will appreciate that the teachings
provided herein may be implemented in a number of different
manners.
[0041] FIG. 1 shows a flowchart of a fissure characterization and
visualization method which may be carried out using software as
part of a pulmonary imaging system, for example. At step 1,
volumetric radiological images or imaging data of a patient are
transmitted to the pulmonary imaging system. Alternatively, the
volumetric radiological images or imaging data may already be
stored within the memory of the system and may be accessed by the
processor. The volumetric radiological images or imaging data may
be CT scans, MM scans, and/or PET scans, for example, from which a
series of two-dimensional planar images (referred to herein as
two-dimensional volumetric images or two-dimensional images) can be
produced in multiple planes, for example.
[0042] At step 2, the lungs, airways, and/or blood vessels are
segmented using the 3D image data. The methods of performing lung,
airway and vessel segmentation from the volumetric images or
imaging data may be those employed by the Pulmonary Workstation of
Vida Diagnostics, Inc. (Coralville, Iowa) and as described in the
following references, each of which is incorporated herein by
reference: United States Patent Publication 2007/0092864, which is
entitled: Treatment Planning Methods, Devices and Systems; United
States Patent Publication 2006/0030958, which is entitled: Methods
and Devices for Labeling and/or Matching; Tschirren et al.,
"Intrathoracic airway trees: segmentation and airway morphology
analysis from low-dose CT scans," IEEE Trans Med Imaging. 2005
December; 24 (12):1529-39; Tschirren et al., "Matching and
anatomical labeling of human airway tree," IEEE Trans Med Imaging.
2005 December; 24 (12):1540-7; Tschirren, Juerg, "Segmentation,
Anatomical Labeling, Branchpoint Matching, and Quantitative
Analysis of Human Airway Trees in Volumetric CT Images," Ph.D.
Thesis, The University of Iowa, 2003; Tschirren, Juerg,
Segmentation, Anatomical Labeling, Branchpoint Matching, and
Quantitative Analysis of Human Airway Trees in Volumetric CT
Images, Slides from Ph.D. defense, The University of Iowa, 2003;
and Li, Kang, "Efficient Optimal Net Surface Detection for Image
Segmentation--From Theory to Practice," M.Sc. Thesis, The
University of Iowa, 2003, for example. Segmentation of the lungs,
airways, and vessels results in identification of the lungs,
airways, and vessels as distinct from the surrounding tissues and
of separation of the lungs, airways, and vessels into smaller
distinct portions which may be individually identified in
accordance with standard pulmonary anatomy.
[0043] At step 3, lobar segmentation is performed. The segmentation
of the lungs, airways, and vessels obtained in step 2 can be used
to identify and delineate the lobes, again by applying standard
pulmonary anatomy. For example, using the identified segments of
the airway and/or vessel trees obtained in step 2, the lobes may be
segmented and identified by extracting the portions of the airway
tree corresponding to particular lobes based on known air way tree
structures and connectivity information. The extracted lobar airway
tree portions may be further divided into portions corresponding to
sub-lobes, again based on known airway and/or vessel tree structure
and connectivity information. In this way, the portions of the
volumetric images corresponding to lobes and/or sub-lobes can be
identified.
[0044] In step 4, the lobar fissures portions of the volumetric
images are identified by the system. The lobar fissures, as formed
by the abutting pleural lining of the lobes, can be seen
radiologically on X-ray as well as on two-dimensional, volumetric
images such as CT scans. As revealed by the tissues lining the
fissures. The fissures may be automatically detected by the system
in the volumetric images using known methods or other methods. In
some embodiments, identification of the lobar fissures begins with
enhancing the fissures to ensure accurate detection. In some
embodiments, Hessian-matrix or structure tensor based approaches
may be used for identification and enhancement of the fissures, as
described in A.F. et al., "Multiscale vessel enhancement
filtering," MICCAI. 1998; 1496 (3):130-7, for example. The
identified fissures may be enhanced and shown to the user on the
volumetric image. An example of this is shown in FIG. 2, which is a
sagittal CT scan 10 including enhanced fissure lines 12.
[0045] In step 5, the fissures may be characterized. This may be
accomplished by combining the information about the lobar
segmentation obtained in step 3 with the fissure identification
obtained in step 4. The locations at which the lobar regions abut
each other may be used to identify the location where a fissure
would normally be present. However, in some individuals, portions
of the fissure (the tissue lining the fissure) may be absent.
Therefore, the normal fissure locations as determined from the
lobar anatomy can be compared to the actual fissure locations
identified in step 4. If there is a location where a fissure would
normally be present as determined by the abutting lobe surfaces,
but the fissure identification indicated that there was no fissure
present in a portion of or all of that location, then the fissure
is described as missing, absent or incomplete in that location. In
this way, the pulmonary imaging system not only can identify and
highlight existing fissures for users and present them in
two-dimensional images and three-dimensional models, but can also
identify locations where the fissure is absent. The extent and
location of absent fissures can then be used to characterize the
patient's disease and to determine appropriate therapeutic
approaches. This method differs from existing methods in which
absent portions are calculated by either fitting the existing
portions of the fissures to a reference atlas (van Rikxoort et al,
2011) or by an extrapolation of the existing portion of the fissure
to the absent portion (Pu, et al., 2010). In the lobar atlas
approach, a reference atlas is created using the fissure locations
of a group of subjects. The fissures of an individual patient can
be compared to the reference atlas to predict the locations of
absent portions of the fissures. This method relies on consistency
of anatomy among individuals, which may not be accurate,
particularly in the presence of severe disease which can
dramatically change fissure patterns. In the extrapolation based
method, the location of missing fissures is estimated by extending
existing fissures into the missing spaces. This method may cause
unpredictable errors, particularly in patients having low fissure
completeness. Therefore, although these and other alternative
methods of identifying missing fissures may be used, these other
approaches do not make full use of the anatomic information
available in the CT image data in the way that the identification
of abutting fissures does.
[0046] Once the locations of existing fissures and absent fissures
have been identified, they can be presented visually to a user in
two dimensions, such as on a CT scan, or in three dimensions, such
as in a three-dimensional model. This step of fissure visualization
is indicated at step 6 on FIG. 1. In some embodiments, the visual
presentations can either show only the existing fissures, with gaps
where the fissures are absent. In other embodiments, only the areas
of missing fissure may be shown. In still other embodiments, the
areas of missing fissures can be shown, with the missing fissure
being shown as the way the fissure would look if it were present.
In some embodiments, the missing fissure is shown in a way that
contrasts with the existing fissure, to clearly indicate that,
although a fissure is shown, the displayed fissure actually
represents an area of missing fissure. For example, the missing
fissure portions may be shown in a different color than the
existing fissure portions. The fissures may be shown as a
three-dimensional model in isolation or in combination with other
components of the lungs such as the airway tree, parenchyma, and/or
the vessels.
[0047] An example of a three-dimensional model of a patient's
fissures 30 in isolation is shown in FIG. 3, with the existing
fissure portions 32 shown in a first color represented by dark gray
and the missing fissure portions 34 shown in a second color
represented by light gray. In this patient, who suffers from
emphysema, the left oblique fissure 36 is more than 95% complete,
while the horizontal fissure 38 is only about 70% complete.
[0048] In FIG. 4B, the existing 42 and missing portions 42, 44 of a
patient's fissures are shown in two dimensions, overlaid on a
sagittal view CT image 20 of the right lung of a patient. The
existing fissures 42 are shown in a first color represented by dark
gray, while the missing portions 44 are shown in a second color
represented by white. For purposes of comparison, the same CT image
is shown in FIG. 4A without the fissure overlay. It can be
appreciated how much more difficult it is to determine the location
of the fissures, and what portions are absent, without the
assistance of the fissure visualization provided in FIG. 4b.
[0049] FIG. 5 is an example of fissure visualization in multiple
views, as it may be presented to a user in a graphical user
interface and therefore represents a screen shot 50 that may be
provided by the pulmonary imaging system. It can be seen that the
screenshot 50 includes CT images 20 in the three orthogonal views:
a sagittal view, an axial view, and a transverse view. In each of
the CT images 20, the existing and missing portions of the fissures
42, 44 are enhanced using a different color, with a first color
represented by dark gray indicating the existing fissure 42 and a
second color represented by white indicating the missing fissure
44. The user may have the option to select different images to be
presented on the display, such as by moving from one image to
another in a series for a particular view. The screenshot 50 also
includes a three-dimensional model of the fissures 30 along with a
model of the airway tree 60, constructed from the analysis of the
two-dimensional volumetric data, with the areas existing and
missing fissures 62, 64 shown in different colors corresponding to
the colors used in the two dimensional images and represented by
dark gray and light gray for purposes of visualization in this
figure.
[0050] In addition to using the fissure information to visually
enhance or display the fissures, the fissure information can also
be used to characterize the fissures, as indicated at step 5 of
FIG. 1. Such fissure characterization can include characterizing
the location of disease, disease heterogeneity, and/or extent of
disease (such as the Global Initiative for Chronic Lung Disease, or
GOLD, classification system), for example.
[0051] In some embodiments, a fissure integrity score may be
calculated to characterize the fissure of a portion thereof. The
fissure integrity score may be calculated as the incompleteness
percentage (IP) or conversely as the completeness percentage (CP).
These values may be calculated using the total area of existing
fissure and of the absent fissure portions determined as described
above using the following equations:
IP (%)=100*[1-ExistingFissure/(ExistingFissure+AbsentFissure)]
CP (%)=100*ExistingFissure/(ExistingFissure+AbsentFissure)
[0052] These measurements can be made for a single fissure, for a
selected portion of a fissure such as only a portion abutting a
particular lobe or sub-lobe, or for a combination of fissures or
selected portions of fissures. The choice of which portion of the
fissure to assess may be determined by the possible locations of
therapeutic interventions such as BLVR surgery. That is, the
fissure integrity score may be calculated for those fissures or
portions thereof which abut a lobe or sub-lobe for which BLVR
therapy is being considered. For example, if bronchoscopy guided
BLVR therapy is being considered for either the left upper lobe or
the left lower lobe, the fissure integrity score may be calculated
for the entire left oblique fissure, because this fissure abuts
both of these lobes along its entire length. If the use of BLVR
therapy is being considered in the right lower lobe, the fissure
integrity score may be calculated based on the entire right oblique
fissure. If BLVR surgery is being considered for the right upper
lobe, the fissure integrity score may be calculated from the
combination of the upper part of the oblique fissure (only the
portion of the fissure abutting the right upper lobe) and the
entire horizontal fissure. If BLVR surgery is being considered for
the right middle lobe, the fissure integrity score may be
calculated for a combination of the lower part of the oblique
fissure (only the portion of the fissure abutting the right middle
lobe) and the entire horizontal fissure.
[0053] Because the fissure integrity score provides a numerical
assessment of how intact (or not intact) the fissures are, it
provides a global quantitative assessment of possible collateral
ventilation. For example, if the completeness percentage is 100%,
the fissure is intact and there is likely no collateral ventilation
between adjacent lobes. BLVR therapy is therefore more likely to be
successful. On the other hand, of the fissure integrity score
indicates that the fissure completeness is low, collateral
ventilation may occur through the missing areas of fissure and the
outcome of BLVR therapy may be less successful.
[0054] In some embodiments, the fissure integrity score may be used
to decide whether or not to proceed with BLVR therapy and in which
lobes or sub-lobes to perform such therapy. For example, a fissure
integrity score cut-off or threshold may be used for therapeutic
decision making. A patient with a completeness percentage below the
threshold may be ineligible for BLVR surgery for the corresponding
portion of the lung. Likewise a patient with an incompleteness
percentage above the threshold may be ineligible for BLVR surgery
for the corresponding portion of the lung. The fissure integrity
score may therefore be used to triage patients as being ineligible
for, or possibly eligible for, BLVR therapy.
[0055] In some embodiments, the relationship between the existing
and absent fissures and other normal or abnormal lung structures
can also be evaluated and measured. For example, the lobes of the
human lungs can be further dived into bronchopulmonary segments,
also called sub-lobes. Each sub-lobe is supplied by one bronchus.
There are typically 10 sub-lobes in the right lung (3 in upper
lobe, 2 in middle lobe, 5 in lower lobe) and 8-10 sub-lobes in the
left lung (4-5 in upper lobe, 4-5 in lower lobe). Depending on
their locations, the surfaces of some sub-lobes may be located at
fissure locations, thus contacting the fissures at such locations,
or they may not abut the fissures. In some embodiments, the portion
of a fissure contacting a sub-lobe may be identified and
characterized as separate from the remainder of the fissure. For
example, characterization of a fissure in a sub-lobe contacting
area can be performed (such as the completeness percent or the
incompleteness percent) and the portions of fissures in contact
with different sub-lobes can be visually distinguished from each
other when displayed for user.
[0056] A visual presentation of the portions of fissures which
contact various sub-lobes can be provided to clinicians as an
indication of the fissure integrity at a sub-lobar level. An
example of this is shown in FIGS. 6(a)-(d) in which 3 dimensional
models of portions of a fissure are shown in a variety of ways that
such models may be provided to a clinician, with the sub-lobe
labels having been identified and displayed with the fissure
portions based on the sub-lobe associated with (contacting) that
portion of the fissure. In these figures, the portions of the
fissures contacting different sub-lobes are each colored
differently, represented by different shades of gray in the
figures, in order to distinguish them from each other, and
different colors are also used to distinguish existing from missing
fissure portions. In FIG. 6(a) the existing portions 62 left
oblique fissure 70 are shown in dark gray while the absent portions
64 shown in light gray. In FIG. 6(b), the left oblique fissure 70
is again shown, with each area of contact of the sub-lobes 76 of
the left upper lobe with the fissure distinctly colored and
labeled. Similarly, in FIG. 6(c), the left oblique fissure is shown
(as seen from below, the opposite side as shown in FIG. 6(b)) with
each area of contact of the sub-lobes 78 of the left lower lobe
distinctly colored and labeled. In FIG. 6(d), the entire left lung
is shown as a three-dimensional model 83, with each of the
sub-lobes 82 separately and distinctly colored (shown in shades of
gray) and labeled with a sub-lobe label 84 and with the fissure
completeness score 86 for each portion of the fissure in contact
with that sub-lobe.
[0057] This information relating to the completeness percentage of
the portion of a fissure contacting a sub-lobe may be used in
combination with other information, such as density based emphysema
measurements, which may be specific to the lobes or sub-lobes, for
example, to guide BLVR treatment planning. This sub-lobe fissure
information can then be used as a degree of the influence of
fissure integrity on sub-lobes. If treatment is being planned for a
particular lung volume such as a lobe or sub-lobe, and if a portion
of the fissure contact with that lung volume has a low fissure
integrity, the treatment of that lung volume may not be effective
or may be less effective than desired due to collateral ventilation
from across the fissure. In such cases, the treatment plan may be
modified to manage the portion of the fissure having low fissure
integrity. For example, the treatment plan may include targeted
treatment of the particular lung volume as well as a sub-lobe or
sub-lobes on the contralateral side of the fissure from the
particular lung volume and adjacent to the portion of the fissure
having low fissure integrity. In this way, collateral ventilation
of the particular treated lung volume can be prevented by targeted
treatments to lung lobes or sublobes across from each other on both
sides of a portion of the fissure having low fissure integrity.
[0058] Other useful information which can be determined based on
the fissure identification includes the spatial relationship
between fissure locations and the regions of the lung affected by
emphysema, for example. For example, the distance of fissures (both
intact and missing portions) from the centroids of regions of
emphysema can be calculated. The orientation of the fissures
relative to the regions of emphysema can also be determined. This
distance and orientation information can be used to predict the
impact of fissure integrity on treatments in the corresponding
regions of emphysema. Additionally, the information may help with
characterizing the emphysema and understanding the impact of
fissure integrity on the progress of the emphysema.
[0059] FIG. 7 is an example of a visual representation of the
spatial relationship between fissures 70 and regions of emphysema
as it may be provided to a clinician in various embodiments. The
visual representation can be used by the clinician to visually
assess the local influence of emphysema on fissure integrity. In
FIG. 7, the regions of emphysema are symbolically represented by
spheres 90 with radii reflecting the sizes of those regions, though
other types of visual representations may alternatively be
used.
[0060] Other information which can be determined using the fissure
identification includes the spatial relationship between fissures
and tumors, which may have an impact on patient prognosis. For
example, recent findings suggest that the presence of tumor
invasion through a fissure has a significant negative impact on
long-term survival, due perhaps to the rapid spread of such tumors.
Thus, it is useful to know the relative locations of fissures and
tumors, the distance between them, and whether or not the tumors
invade the fissures. Various embodiments therefore identify the
locations of tumors and fissures, provide images such as the 3
dimensional model of the fissures 70 and airways tree 60 shown in
FIG. 8 in which the tumor 92, fissures 70, and the airway tree 60
can be seen, and/or calculate the nearest distance between the
fissures 70 and the tumor 92. Since tumors invading through the
fissures have a significant effect on long-term survival, it is
important to visualize the spatial relationship between fissures
and tumors. In the example shown in FIG. 8, it can be seen that
both tumors 92 are confined to a single lobe and they do not invade
the fissures 70.
[0061] In addition, local and global measurement of fissure
integrity can also be utilized to predict the spread of diseases
such as cancerous tumors. Other measurements which may be made by
the system in various embodiments include the distance between the
fissures and anatomical landmarks or locations such as the lung
apex, the diaphragm, and the ribs, for example. In addition, these
measurements can be performed at different levels of lung
inflation, to provide information about, and to help better
understand, lung mechanics in both normal and diseased lungs.
[0062] As discussed above, the fissures are the interface between
the lobes of the lungs and they are lined by the pleura. An
analysis of the fissures can therefore include characterization of
the pleura itself. For example, pleural thickening can occur in
certain disease conditions, and in some cases is due to
inflammation. Such pleural thickening can result in changes in the
intensity distributions and thickness of fissure surface. For
example, portions of the fissure may have an abnormal intensity on
volumetric imaging which may be indicative of the presence of
disease or fluid. Various embodiments may therefore identify the
intensity, such as in Hounsfield Units (HU), of the fissures and of
the various portions of the fissures if the intensity is not
uniform. Various embodiments may therefore provide measurements of
the intensity distribution and the thickness of the pleura, or can
assist a clinician in making these measurements, to provide further
information about and characterization of the associated
disease.
[0063] In some embodiments, the shape of the fissure may be
determined by the system. Fissure shape can be changed due to lung
disease, such as emphysema. Thus, analysis of fissure shape can
also be useful in characterizing lung disease. The shape analysis
may include, but is not limited to, principal component analysis
and surface curvature measurement, for example. These results may
be provided in comparison to normal results, for example, to help
identify areas of abnormality since the normal shape can be altered
due to some diseases.
[0064] In some embodiments, the topology of the fissure surface may
be characterized. The topological information may include, for
example, the number of holes (incomplete portions) in the fissure,
which could be caused by or associated with a vein crossing the
fissure.
[0065] In some embodiments, a clinician may interact with the
visual display to identify the fissures manually or to edit the
fissures that were automatically identified by the system. An
example of an editing tool icon 94 is shown in FIG. 9, in which a
sagittal CT image 20 of the lungs is shown. The editing tool 94 can
be used to edit the enhanced fissure line 22, such as to change the
characterization of the identified fissure from existing fissure 42
to incomplete fissure 44 or vice versa. The editing tool, the icon
for which may appear differently from that shown in FIG. 9, may
allow a user to change the identification of the voxels at the
fissure location, relabeling them as either existing fissure or
incomplete/missing fissure.
[0066] In some embodiments, the process of editing a fissure using
a fissure editing tool may include the following steps. First, a
user may select a fissure editing tool for use in a two-dimensional
image. The two-dimensional image may include identification of the
fissure locations as existing or incomplete, as automatically
identified by the system, which may be shown enhancing the fissure
by using colors such as blue for existing and green for missing
fissure. The user may then position to the editing tool icon at a
selected a location in the two-dimensional image including the
automatically identified existing and missing fissure. The user may
then direct the system to change the fissure identification (from
existing to incomplete, or from complete to existing) using the
tool. For example, the user may click and drag a mouse to move the
corresponding tool icon on the display, at the location of the
portion of the fissure for which the user wishes to change the
fissure identification. During use, the tool editing icon may
appear in a color matching the color of the new (revised) state of
the fissure, such as a first color or shade of gray such as light
gray for intact or a second color or shade of gray such as dark
gray for missing fissure, for better visualization of the
underlying CT data. The fissure label (existing or missing) in the
edited image and neighboring images will be automatically updated
according to the size of the 3D sphere. An example of this is shown
in FIG. 9 in which circle 96 represents the central cross-section
of the 3-dimensional volume within which the fissure identification
will be changed, if so directed by the user. The color change may
occur immediately while the user is interacting with the image, or
may occur when the user indicates that editing is complete, such as
by unclicking the mouse.
[0067] Fissure editing may also be performed by a user by
interacting with a three-dimensional model of the fissures produced
by the system. An example of this is shown in FIG. 10, in which the
editing tool icon 94 is shown in the 3D model of the fissures 30.
The model also includes a 3D model of the airway tree 60. The user
may edit the fissure characterization using the following steps.
First, the user may select the editing tool for use in the
three-dimensional model, which displays both existing and
incomplete portions 62, 64 of the fissures as automatically
identified by the system and/or previously edited by a user. The
user may place the editing tool icon 94 on a selected location on
the fissure model 30. The user may then apply the tool to the
fissure to change the identification of the fissure location as
existing or incomplete, such as by clicking and dragging the icon
using a mouse, at the fissure location as described above for
editing the two-dimensional image. In response, the system may
change the fissure characterization, and likewise change the
fissure color shown in the model 30, to indicate the revised
fissure characterization.
[0068] The use of the pulmonary visualization system which includes
automatic lobar fissure identification, visualization, and
characterization as described herein provides several advantages.
The system may provide a priori knowledge to predict the response
of a patient to a bronchoscopically-guided procedure such as a BLVR
procedure. It may also provide an easily recognizable visual
display of completeness and incompleteness of the lobar fissure,
such as through the use of color coding. It may also provide an
easily recognizable visual display of the spatial relationships
between fissures and normal and abnormal lung structures including
the airway tree, the lobes, the sub-lobes, the fissures, regions of
emphysema, and tumors, for example. In addition, it may detect and
identify normal and abnormal regions of the lungs and fissures and
link two-dimensional data and images to multidimensional
visualization and measurements. In some embodiments, it may offer
"on-demand" measurement of fissures for purpose of immediate
evaluation of normal and diseased states, determination of the
appropriateness of a proposed procedure, and procedure planning.
The automation of the measurement of the fissure integrity may
provide enhanced clinical utility by allowing easier, faster, and
more accurate decisions, thereby saving time, money and potentially
lives.
[0069] Various embodiments may be used by physicians to predict the
response of a patient with emphysema or other lung disease to a
proposed procedure, such as the implantation of a device or other
BLVR treatment. Treatment planning and determination of the most
appropriate device therapy may be optimized by predicting response.
For example, thoracic surgeons may use the information for
treatment planning for lung volume reduction surgery. Radiologists
and pulmonary clinicians may use these characterizations to
determine the appropriate patients to triage to endobronchial BLVR
therapy. Pulmonary clinicians may use the information to plan
procedures for BLVR therapies and to evaluate treatment
response.
[0070] With regard to the prediction of collateral ventilation, it
is noted that such a prediction may be considered as intralobar or
interlobar collateral ventilation. Intra-lobar collateral
ventilation may occur through the accessory pathways of the lungs
including the intra-alveolar pores of Kohn (the Pores of Kohn take
their name from the German physician Hans Kohn [1866-1935] who
first described them in 1893), the bronchioalveolar communication
of Lambert and the intrabronchiolar pathways of Martin. In 1955,
Lambert discovered that there were accessory bronchioloalveolar
communications extending from respiratory bronchioles to alveolar
ducts and sacs subtended by the bronchiole. Later, Martin was able
to pass polystyrene spheres up to 120 microns in diameter from one
segment of the canine lung to another, through the collateral
channels. (See, e.g., Menkes H, Traystman R, Terry P., Collateral
ventilation. Fed Proc. 1979 January; 38(1):22-26, hereby
incorporated by reference). In certain conditions, such as
emphysema, these accessory pathways can become enlarged and airway
obstruction can increase expiratory resistance, leading to the
passage of air as intra-lobar collateral ventilation from one
lobule to another. Interlobar collateral ventilation may occur when
portions of the interlobar fissures are absent or when the adjacent
lobes become fused to each other, resulting in an incomplete
fissure and allowing air communication between the lobes at those
locations.
[0071] In some embodiments, fissure integrity data as herein
described can be analyzed in conjunction with treatment efficacy
data of patients that have undergone a procedure such as LVR
therapy. Statistical analysis of such data can provide greater
insight as to whether particular fissure sections may be more
reliable at predicting treatment efficacy than others. That is,
based on air-flow physics, various portions of the fissure surface
may contribute differently to collateral ventilation. If so, the
prediction of response to LVR treatments may be improved based on
more regional information of fissure completeness.
[0072] However, analysis of a plurality of patients may be
difficult, as the overall lung structure (e.g., shape and size of
lungs, lobes and sub-lobes) may vary from person to person and
change based on the level of inspiration. As previously described,
a collection of data from a variety of patients may be used to
build a reference atlas or a set of reference atlases that are
created using the fissures or other pulmonary structures from one
or various groups of subjects. For instance, in some embodiments,
3D pulmonary data of lungs or fissures of a given patient can be
registered to the reference domain on a voxel-by-voxel basis.
[0073] Each entry in the atlas may be registered to the reference
domain so that voxel-based statistical analysis may be performed
among entries in the atlas while referring to the same physical
location within the lungs. That is, after registration, a location
identified in one patient in an atlas will have a corresponding
location in any other subjects in the atlas. In some examples, the
reference domain can have pre-defined lung structures including
lungs, lobes, sub-lobes, fissures, and the like of a reference
patient. In some such examples, the reference patient may be a
healthy patient or a patient with mild disease and nicely defined
fissure surfaces or other lung structures.
[0074] FIGS. 11 and 12 illustrate exemplary registration of 3D
fissure surface models from an atlas into a reference domain. FIG.
11 shows a series of exemplary 3D fissure surface models 102-112
from different patients. In each model, dark gray surface indicates
the complete portions of a fissure, while the light gray surface
indicates the incomplete portions. It is evident from the models
that each patient has a unique fissure makeup. It is also evident
that the overall lung structure (e.g., shape, size, etc.) varies
widely from patient to patient. Accordingly, it may be difficult to
directly compare fissure information between, for example, models
106 and 112 based on the images of FIG. 11. However, as previously
described the fissure surface models 102-112 may be registered to a
reference domain 120 to allow direct voxel-by-voxel comparison of
fissure characteristics.
[0075] FIG. 12 shows a series of atlas entries registered to a
single reference model. As shown, a variety of models in the atlas
include fissure information as described above with reference to
FIG. 11. However, in the illustrative example of FIG. 12, each
entry has been registered on a voxel-by-voxel basis to the
reference domain 120. That is, the overall shape and structure of
each of the registered models was warped into the reference domain
120, but retains its own fissure integrity data. By establishing
the spatial correspondences among fissures from different subjects,
one can evaluate the statistical differences in fissure integrity,
and other fissure properties amongst a given population. For
example, such data can be statistically analyzed among various
populations including but not limited to: LVR responders, LVR non
responders, subjects with or without collateral ventilation,
subjects with various degrees of pulmonary disease (COPD, asthma,
etc.). It will be appreciated that, with respect to FIGS. 11 and
12, while various portions of the image represent fissure
completeness, other fissure characteristics, such as curvatures,
distance map to other lung structures, etc., may be represented in
the fissure atlas.
[0076] Once the atlas is built (e.g., once plurality of scan data
has been registered to the reference domain), statistics from a
given population may be analyzed. For example, in some embodiments,
entries in the atlas can be divided into two categories: patients
who responded to a particular treatment (responders) and patients
who did not (non-responders). In other examples, responders and
non-responders can be divided into separate atlases. Fissure
characteristics such as integrity data can be analyzed on a
voxel-by-voxel basis to determine the likelihood that, among
patients in one particular category, certain fissure
characteristics are present. In some examples, the likelihood that
the fissure in a given voxel is intact can be determined. For
example, given a group of 100 responders to a treatment and a voxel
or region V.sub.1 located on the fissure, it can be determined how
many of the 100 responders had a complete fissure represented at
V.sub.1. This number may be referred to as "fissure integrity
likelihood". Other likelihood determinations related to additional
fissure characteristics are possible.
[0077] FIG. 13 is a visual representation of the relationship
between the fissure integrity likelihood and response to a
treatment. For example, a series of patient scans can be organized
into a series of atlases with each of the atlases containing scans
having like properties. Such properties can include responders to a
particular therapy (or non-responders), diseased patients, healthy
patients, patients with specific diseases, patients with certain
stages or severity of a disease, and the like. In some instances
scans can be divided into atlases corresponding to such categories
for direct comparisons with other categories, such as healthy vs.
diseased patients, patients with a first disease vs. patients with
a second disease, or patients at various stages of a disease. In
the illustrated example, the upper section of FIG. 13 includes
atlases of the left lung (LL, either left lower or left upper
lobe), the right lower lobe (RLL), and the right upper lobe (RUL)
for responders to endobronchial valve therapy treatment. Each atlas
is statistically analyzed to determine the fissure integrity
likelihood at each voxel. The lower section of FIG. 13 shows
similar data for non-responders. In the illustrated embodiment,
dark gray sections indicate voxels representing portions of the
fissure that were intact in most of the patients in the group.
Light gray sections indicate voxels representing portions of the
fissure that were not intact in most of the patients in the
group.
[0078] Voxel-by-voxel comparisons of fissure integrity likelihood
scores for responders and non-responders may allow for
visualization of which portions of a fissure are important for the
efficacy of therapy. For instance, it can be seen that there is a
low fissure integrity likelihood corresponding to a first portion
of the LL in non-responders. At first, this may suggest that the
integrity of the first portion of the LL includes an important
fissure section for the efficacy of the therapy. However, analysis
of the LL in responders to the therapy reveals that patients may in
fact respond to the therapy despite a low fissure integrity
likelihood at the first portion. Accordingly, a patient having low
fissure integrity at the first portion of the left lobe may still
respond to endobronchial valve therapy. It will be appreciated
that, while the representation of FIG. 13 relates to patients who
are or who are not responsive to endobronchial valve therapy, the
same technique may be used for any of a variety of LVR therapies
seeking atelectasis.
[0079] In another example, it can be seen that there is low fissure
integrity likelihood at to a second portion in the RUL in
non-responders, while there is high fissure integrity likelihood at
the second portion in the RUL in responders. Accordingly, it may be
that the second portion of the RUL is an important fissure portion
for efficacy of endobronchial valve therapy. In general,
voxel-by-voxel probabilistic maps of fissure integrity likelihood
such as that shown in FIG. 13 provide a visual representation of
which fissure portions may be important for efficacy of certain
treatments.
[0080] With further reference to FIG. 13, global analysis of the
RUL scan of a patient in the group of responders may yield a
relatively low overall fissure integrity score. That is, as shown
in the upper right image of FIG. 13, a patient in the group of
responders may show an incomplete fissure at the third portion of
the RUL. As a result, a global fissure integrity measurement of the
RUL may be low enough to suggest that LVR therapy will be
ineffective based on the global analysis. However, FIG. 13 suggests
that local fissure incompleteness at the third portion of the RUL
may not prohibit a patient from responding to a treatment.
Accordingly, voxel-by-voxel analysis of fissure integrity
likelihood data of responders and non-responders may improve the
ability to predict the likelihood of therapy efficacy when compared
to global fissure integrity measurements.
[0081] In some instances, LVR treatment procedures may be
ineffective for reasons other than fissure incompleteness. For
example, one or more valves placed in one or more portions of a
patient's lungs may be inappropriately sized for the position in
which they are placed. Additionally or alternatively, one or more
valves may be misplaced or misaligned within the lungs.
Accordingly, one or more valves placed within the patient may not
properly provide the outcome intended by the LVR procedure.
Ineffectiveness of such therapy may not be due patient selection,
but rather may be caused by procedural errors. As a result, scans
of patients who received therapy including procedural errors may be
indexed into a non-responder atlas, even though the patient's
fissure integrity may be high and indicative of a potential
responder. This may skew statistical analysis of the effect of
fissure integrity on therapy efficacy such as the analysis
described above. In addition to treatment procedural issues
described above, other factors may affect fissure metrics, such as
local motion of proximate the lungs (e.g., of the heart or
diaphragm). Such motion may affect the appearance of the fissure in
scan data, and may make it appear incomplete in some regions,
falsely impacting the fissure completeness score.
[0082] In various embodiments, data can be weighted to account for
factors that may affect the reliability of the fissure metric
analysis. That is, when generating an atlas or compiling fissure
data from an atlas, data contributions from certain scans or voxels
may be weighted based on a confidence in their dependence on
fissure metrics. For instance, if a scan in a non-responder atlas
is from a patient having a misaligned valve from an LVR procedure,
the fissure data from that patient may not be reflected in the
determination of the fissure integrity likelihood for
non-responders. That is, in determining the fissure integrity
likelihood for a given voxel among non-responders, data from the
patient with the misaligned valve may be weighted less than other
data sets. The weighting can be done on the data set as a whole, or
on a subset of voxels affected by the misalignment. In general, a
confidence weight may be included in statistical analysis of the
fissure data in order to compensate for possible unrelated
contributors to therapy response or other fissure data. The
confidence weight may be applied on a scan-by-scan basis, a
voxel-by-voxel basis, or in other various volumetric subdivisions,
such as lobes, sub-lobes, or other definable sets of voxels. In
some examples, confidence weights can be applied to the data used
in the generation of fissure integrity likelihood maps such as
shown in FIG. 13. In general, a weighting mechanism or other
integration of prior information such as valve placement issue can
be interpreted in a statistical framework and seen as a way to
improve the classification accuracy. In various embodiments, other
prior information known to have an impact on fissure completeness
could be similarly considered to improve the maximum a posteriori
solution.
[0083] FIG. 14 is a visual representation of the relationship
between the fissure integrity likelihood and response to a
treatment including confidence weights based on valve issues. As
with FIG. 13, the upper section of FIG. 14 includes atlases of the
left lung (LL, either left lower or left upper lobe), the right
lower lobe (RLL), and the right upper lobe (RUL) for responders to
endobronchial valve therapy. Each atlas is statistically analyzed
to determine the fissure integrity likelihood among the atlas at
each voxel. The lower section of FIG. 14 shows similar data for
non-responders. However, FIG. 14 includes a confidence weight
associated with entries in the atlases to be combined and compared.
Weighting can be performed on a voxel-by-voxel basis, or may be
performed for sets of multiple voxels, such as lobes or
sub-lobes.
[0084] For instance, in the illustrative example, information about
lung segments associated with valve positioning issues (highlighted
RB1 and RB3 segments in the treated RUL) can be used to adjust the
fissure integrity atlas. That is, due to known information
regarding the lung segments associated with the valve positioning
issues, the data contributing to the fissure integrity likelihood
scores at voxels associated with those segments can be weighted
accordingly. In general, weighting can be performed according to
any voxel selection routine, such as voxels corresponding to lobes,
sub-lobes, or arbitrarily-shaped sub-segments of the fissures
(e.g., cylindrical, cubic, etc.).
[0085] In an exemplary embodiment, a Lobar Occlusion Score (LOS) is
computed a posteriori for each patient and individual treated lobe
to determine the segments for which valves have been improperly
positioned. In some such embodiments, the LOS score is simply the
percentage of the volume of affected sub-lobes relative to the
overall lobar volume. The LOS can then be used to weight the
contribution of each sub-lobe to the fissure atlas, such as by way
of the confidence weight of FIG. 14. In some examples, this
correction is particularly important for non-responders, since
improperly positioned valves may be more likely to falsely indicate
non-response than a positive response.
[0086] Various processes described above can be performed in order
to generate a final fissure atlas. FIG. 15 is a process flow
diagram illustrating an exemplary method of generating a final
fissure atlas. Fissures from a given population are mapped to a
reference domain (202). As described above, the reference domain
can include fissure data from a single patient or a set of
patients, for example. Once the fissure surfaces are mapped to a
singled reference domain, voxel-based statistics of fissure metrics
can be computed for a set of parameters (204). For example, a
fissure integrity likelihood score can be determined for each voxel
within the population of fissure surfaces. In some examples,
various metrics can be computed according to confidence weights as
described above (206). Additionally or alternatively, voxels can be
partitioned into regions of interest (208). Regions of interest can
be defined by lobes, sub-lobes, or other defining boundaries, such
as customized volumes of a defined size and shape. In some
examples, regions may be defined based on determined fissure
metrics. Voxel-based statistics among a population, which may be
weighted per step 206, can be combined into a final fissure atlas
(210). The final fissure atlas may include regional partitions as
defined in step 208.
[0087] In some embodiments, comparison data can be generated and
displayed for comparing statistical data between groups in an atlas
or between atlases. For instance, regarding FIGS. 13 and 14,
comparisons between atlases of responders (e.g., the upper row of
atlases) and non-responders (e.g., the lower row of atlases) to a
particular LVR therapy can be performed for determining areas with
significant differences between populations. These areas may have a
higher correlation to treatment efficacy than areas without
significant differences between populations when predicting whether
or not a patient will respond to a treatment.
[0088] FIG. 16 is an illustration highlighting the absolute
differences between the fissure probability maps of LVR responders
and LVR non-responders from FIG. 13. As shown in FIG. 16, the
lighter areas indicate differences in fissure integrity likelihood
nearing 50%, whereas darker areas indicate minimal differences in
fissure integrity likelihood between responders and non-responders.
Thus, since the darker areas show little difference between fissure
integrity likelihood between responders and non-responders, the
corresponding fissure portions may not be crucial to determining
the likelihood of responding to LVR therapy. While the absolute
difference between integrity likelihood in responders and
non-responders is illustrated in FIG. 16, it will be appreciated
that other mechanisms may be used to combine or compare sets of
data. In general, one or more such comparisons and combinations of
data may be selected in order to determine distinguishable
characteristics between responders and non-responders
[0089] FIG. 17 is a diagram illustrating an exemplary process for
determining differentiating features among populations. As shown,
in Step 1, a series of fissure databases representative of
different populations P.sub.1, P.sub.2. . . P.sub.p are defined.
Exemplary populations can include, for instance, responders or
non-responders to a LVR therapy. In general, any number of
populations may be defined. Next, in Step 2, each entry in each of
the databases is processed into a fissure atlas associated with
each population. Processing each entry in the database can include,
for instance, mapping each entry in the database to a reference
domain. In some example, creating the fissure atlas includes a
combination of each of the entries from the database being entered
into the atlas, such as the various weighted combining processes
described above. Processing can further include identifying various
regions of interest within the reference domain for comparative
analysis. For instance, the reference domain may be partitioned
into lobes, sub-lobes, or custom-defined volumetric regions. This
is illustrated in FIG. 17 by the division of the reference domain
(and similarly, each entry from the database that was mapped to the
reference domain) into regions R.sub.1, R .sub.2. . . R.sub.n. In
some embodiments, the generation of fissure atlases in Step 2 may
be performed in a similar manner to the method of FIG. 15.
[0090] Once the fissure atlases have been defined in Step 2,
fissure features can be determined in Step 3 and computed in the
reference domain. In various embodiments, fissure features may be
associated with each region defined by the atlas. For example,
fissure features associated with each region can include: a fissure
integrity score, the deviation of the fissure surface from the mean
model, distance between a fissure surface regional area and other
lung structures including but not limited to a series of airway
branch or a specific branch, vascular trees (arteries, veins), CT
density on fissure surfaces, and other quantitative measurements
projected on fissure surfaces, and the like. Such data can be
determined for each region in each fissure atlas. As shown, data
can be stored according to regions, for instance, in a vector.
[0091] According to Step 4, fissure features may be selected based
on their relative predictive value to procedural outcome or disease
contribution. In the particular example of FIG. 17, differentiating
features include the fissure integrity of regions 1 and 2, the
deformation fields of regions 5 and n, airway distances of regions
1 and 4, and the local vascularity of regions 1, 2, and n. The
selected differentiating features may vary depending on the
different populations being compared. In general, differentiating
features may be determined by a variety of methods, such as
comparisons or other statistical analysis of data in each atlas
(e.g., each population). Various features may be directly compared
to other atlases, or may be used to modify the comparison. For
instance, in some examples, the airway distance feature may be used
to weight the fissure integrity score from its respective region if
the airway distance is determined to affect the importance of the
fissure integrity of that region. Additionally or alternatively,
since in some examples it may be easier to analyze vascularity of a
patient than airway structure, and that vascularity may closely
mirror airway structure, the vascularity data may work to enhance
the airway distance data. In some embodiments, the selection of
best differentiating features can be performed using known feature
selection methods or by assessing the predictive ability of
features using various classification schemes, as will be described
below.
[0092] The selected best differentiating features may be used to
train a classifier, as shown in Step 5. A trained classifier may
help predict in which population a yet unclassified data set
belongs. For example, if populations are determined by whether or
not a patient is a responder to an LVR procedure, classification of
a scan for a new patient may comprise predicting whether or not the
patient is likely to be a responder to the LVR procedure.
Classification can be performed according to a variety of methods.
For example, for an unclassified set of data, the set of data can
be mapped to the reference domain and its associated fissure
features can be determined. The differentiating features of the
unclassified data can then be analyzed, and the data can be
classified according to its differentiating features. The
classified data can be added to an appropriate database.
[0093] In some embodiments, a variety of the steps of FIG. 17 may
be performed according to a variety of classification and selection
schemes. For instance, in some embodiments, Feature selection
approaches aim to select a small subset of features. For the
classification problem, feature selection aims to select subset of
highly discriminant features such as the features highlighted in
step 4 of FIG. 17. In other words, it selects features that are
capable of discriminating patient data that belong to different
classes (responders vs. non-responders, disease vs. healthy, etc.).
Classification in step 5 of FIG. 17 consist in identifying to which
of a set of categories (sub-populations) a new patient data
belongs, on the basis of a training set of data containing patient
data (or instances) whose class membership is known. Classification
is performed using discriminant features identified in step 4. For
the purpose of this invention, may combinations of feature
selection and classification methods can be used and explored. We
refer to state-of-the-art literature in machine learning (ex. M
Sonka, V Hlavac, R Boyle. Image Processing, Analysis, and Machine
Vision. Fourth Edition. 2014.)
[0094] As previously described, the method described in FIG. 17 may
be used to identify patients likely to respond to a treatment or
directly associated with a disease state based solely on scan data,
and prior to the procedure to be performed. FIG. 18 is a process
flow diagram demonstrating the classification of a patient. First,
a scan of the patient's lungs is received (302). The scan data can
be mapped to a reference domain (304). Mapping to the reference
domain may allow for consistent analysis with a database of
classification data. In some examples, the reference domain may be
defined by a fissure atlas 310 containing a series of patient
scans, each represented in the reference domain. Once mapped to the
reference domain, a variety of fissure related features of the
patient data can be determined (306). Finally, based on the
determined fissure related features of the patient data, the
patient may be classified (308) using a classifier 312 trained on
the fissure atlases 310.
[0095] The classifier 312 may identify a plurality of populations
(e.g., responders, non-responders) into which to group patients as
described with regard to FIG. 17. In some such examples, the
populations may each be identified according to a unique fissure
atlas in fissure atlases 310. Accordingly, the process described in
FIG. 18 may be used as a predictive method for determining, for
example, whether or not a patient is likely to respond to a LVR or
other treatment therapy based on scan results. As a result, therapy
may be selectively provided to those likely to respond to the
therapy. In another application, a similar procedure may be used to
determine which course of therapy is most likely to cause a patient
to respond, and therefore may be used in a treatment program
selection process.
EXPERIMENTAL
[0096] The following experimental description illustrates results
of predictive classification schemes. Treatments are broken down
into left upper/lower lobe treatment, right upper lobe treatment,
and right lower lobe treatment. Given a database of patient data
and knowledge of whether or not the patient responded to the given
therapy, the sensitivities and specificities of predictive models
are given for a variety of therapies. In addition, such metrics are
recorded for various regional divisions for analyzing fissure
integrities. For instance, classification was performed based on
lobar, sub-lobar, and regional (sub-segmental) fissure integrity
analysis.
[0097] BACKGROUND: Fissure integrity (FI) as a global measurement
of lobar collateral ventilation has been shown to correlate to
response of valve-based lung volume reduction (L VR) therapy.
[0098] OBJECTIVES: To determine if regional FI can further
influence outcome by providing a more localized predictor.
[0099] METHODS: Automated regional FI analysis of 253 subjects
treated with valves was analyzed in Apollo (VIDA Diagnostics, IA).
Individual fissures separating the treated lobe from its
ipsilateral lobe were spatially matched to the corresponding ones
in a reference patient, allowing voxel-based comparisons of fissure
properties between responders (n=141) and non-responders (n=112).
Regional FIs were derived based on sub-lobar contact areas or
further divided smaller sub-segmental regions on fissure surface.
Predictions of L VR outcome using lobar FL sub-lobar FI and
sub-segmental FI were obtained using a Naive Bayes classifier with
10-fold cross-validation to avoid over-optimistic results.
[0100] RESULTS: Lobar distribution of LVR treatment for
responders/non-responders is: RUL (36/21), RLL (19/14), LUL/LLL
(101/62). The sensitivity and specificity values are listed in
Table 1 below. Both methods of local FIs yield better
classification results than the global FI.
[0101] CONCLUSION: The results suggest regional FI biomarkers could
expand patient selection for valves and lead to more targeted,
personalized treatments for emphysema patients including other LVR
techniques.
TABLE-US-00001 TABLE 1 LVR outcome prediction using lobar FI,
sub-lobar FI, and sub-segmental FI Left upper/lower Right upper
lobe Right lower lobe lobe treatment treatment treatment (N = 163)
(N = 57) (N = 33) Lobar Sub- Sub-seg Lobar Sub- Sub-seg Lobar Sub-
Sub-seg FI lobar FI FI FI lobar FI FI FI lobar FI FI Sensitivity 96
98 97 90 95 95 58 95 100 (%) Specificity 12 21 26 9 92 78 57 50 79
(%)
[0102] As shown, when compared to a global, lobar FI metric,
sub-lobar and sub-segmental fissure integrity analysis often
results in improved sensitivity (true positives) and specificity
(true negative) when predicting the efficacy of an LVR
treatment.
[0103] In the foregoing detailed description, the invention has
been described with reference to specific embodiments. However, it
may be appreciated that various modifications and changes can be
made without departing from the scope of the invention.
* * * * *