U.S. patent application number 13/411041 was filed with the patent office on 2012-11-15 for atlas and methods for segmentation and alignment of anatomical data.
This patent application is currently assigned to CORTECHS LABS, INC.. Invention is credited to Anders Dale, Sebastian Gicquel.
Application Number | 20120288174 13/411041 |
Document ID | / |
Family ID | 21996708 |
Filed Date | 2012-11-15 |
United States Patent
Application |
20120288174 |
Kind Code |
A1 |
Dale; Anders ; et
al. |
November 15, 2012 |
ATLAS AND METHODS FOR SEGMENTATION AND ALIGNMENT OF ANATOMICAL
DATA
Abstract
The present invention provides an atlas comprising values
representative of magnetic resonance properties of a magnetic
resonance (MR) scan and optionally, prior probability data relating
to tissue type. Further embodiments of the invention involve a
system including an MR scanner and the atlas for use in alignment
of an MR scan, such as a localizer scan, to obtain a specific
geometry of the data acquired during a subsequent scan. Also, a
system includes an MR scanner and the atlas for automatic
segmentation of an MR scan. Methods of making and using the atlas
and system are also provided.
Inventors: |
Dale; Anders; (La Jolla,
CA) ; Gicquel; Sebastian; (Merignac, FR) |
Assignee: |
CORTECHS LABS, INC.
La Jolla
CA
|
Family ID: |
21996708 |
Appl. No.: |
13/411041 |
Filed: |
March 2, 2012 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
11999887 |
Dec 7, 2007 |
8140144 |
|
|
13411041 |
|
|
|
|
10055256 |
Jan 22, 2002 |
7324842 |
|
|
11999887 |
|
|
|
|
Current U.S.
Class: |
382/131 |
Current CPC
Class: |
A61B 5/055 20130101 |
Class at
Publication: |
382/131 |
International
Class: |
G06K 9/00 20060101
G06K009/00 |
Claims
1. A prior probability MRI based atlas embodied on a
computer-readable medium comprising: a plurality of nodes
corresponding to a plurality of voxels representing spatial
locations of a subject, each of the nodes configured to store: at
least two magnetic property values for each of the voxels as
determined by magnetic resonance imaging of a plurality of
subjects, wherein the at least two magnetic property values
correspond to tissue type at one or more voxels; and prior
probability data relating to tissue type at one or more voxels,
wherein the tissue type prior probability is determined from
measurements of at least one subject.
2. The atlas of claim 1, wherein the plurality of nodes represent
the subject divided in three-dimensional space.
3. The atlas of claim 1, wherein each node in the plurality of
nodes corresponds to a voxel in the plurality of voxels.
4. The atlas of claim 1, wherein each node in the plurality of
nodes corresponds to more than one voxel in the plurality of
voxels.
5. The atlas of claim 1, wherein the at least two magnetic property
values of each of the voxels comprise T1 values and T2 values.
6. The atlas of claim 5, wherein the at least two magnetic property
values of each of the voxels further comprise proton density
values.
7. The atlas of claim 1, wherein the at least two magnetic property
values represent different magnetic properties.
8. The atlas of claim 1, wherein the at least two magnetic property
values correspond to a first tissue type at one voxel and a second
tissue type at a second voxel.
9. The atlas of claim 1, wherein the at least two magnetic property
values correspond to a first tissue type at one voxel and a second
tissue type at a second voxel.
10. A prior probability MRI based atlas embodied on a
computer-readable medium compromising: a plurality of nodes
corresponding to a plurality of voxels representing spatial
locations of a subject, each of the nodes configured to store: at
least one magnetic property value as determined by magnetic
resonance imaging of at least one subject; and at least one tissue
type prior probability value corresponding to a tissue type of one
or more voxels in the plurality of voxels as determined from
measurements of the at least one subject.
11. The atlas of claim 10, wherein each of the nodes is also
configured store a second tissue type prior probability value
corresponding to a second tissue type of one or more voxels in the
plurality of voxels as determined from measurements of the at least
one subject.
12. The atlas of claim 10, wherein the at least one magnetic
property value as determined by magnetic resonance imaging of at
least one subject is determined by magnetic resonance imaging of a
plurality of subjects; and wherein the at least one tissue type
prior probability value corresponding to a tissue type of one or
more voxels in the plurality of voxels as determined from
measurements of the at least one subject is determined from
measurements of the plurality of subjects.
13. A prior probability MRI based atlas embodied on a
computer-readable medium comprising: a plurality of nodes, each
node configured to store: statistical values calculated from a
statistical representation of measured values of at least one
magnetic property as determined by magnetic resonance imaging; and
statistical values calculated from a statistical representation of
prior probability values corresponding to a tissue type for each of
a plurality of corresponding voxels of a plurality of subjects.
14. The atlas of claim 13, wherein the statistical values
calculated from the statistical representation of measured values
comprise: a mean and a variance of intensities of each of a
plurality of magnetic property values at each corresponding voxel
of said plurality of subjects.
15. The atlas of claim 13, wherein the statistical values are
determined for each tissue type.
16. A prior probability MRI based atlas embodied on a
computer-readable medium comprising: a plurality of nodes, each
node configured to store: statistical values calculated from a
statistical representation of at least two magnetic property values
as determined by magnetic resonance imaging for each of a plurality
of corresponding voxels of a plurality of subjects; and at least
one tissue type prior probability value corresponding to a tissue
type for each of a plurality of corresponding voxels of a plurality
of subjects.
17. The atlas of claim 16, wherein the statistical values
calculated from the statistical representation comprise: a mean and
a variance of intensities of each of a plurality of magnetic
property values at each corresponding voxel of the plurality of
subjects.
18. The atlas of claim 17, wherein the mean and the variance of
intensities of the magnetic property values at each corresponding
voxel are determined for the tissue type for each of the plurality
of corresponding voxels of the plurality of subjects.
19. The atlas of claim 16, wherein the statistical values
calculated from the statistical representation of at least two
magnetic property values are scanner-specific.
20. The atlas of claim 16, wherein the statistical values
calculated from the statistical representation of at least two
magnetic property values are acquisition-specific.
21. The atlas of claim 16, wherein the statistical values
calculated from the statistical representation of at least two
magnetic property values comprise magnetic resonance sequence
parameters.
22. A prior probability MRI based atlas embodied on a
computer-readable medium comprising: a plurality of nodes
corresponding to a plurality of voxels of at least one subject;
wherein at least one node of the plurality of nodes is configured
to store: a prior probability of a tissue type located at a voxel
corresponding to the at least one node, calculated for a plurality
of tissue types; and a statistical value of a measured magnetic
property of the voxel, calculated for each tissue type located at
the voxel.
23. The atlas of claim 21, in which in said statistical value
comprises a mean and a variance of the measured magnetic
property.
24. The atlas of claim 21, wherein the plurality of tissue types
comprise a labeled anatomical structure.
Description
RELATED APPLICATIONS
[0001] This application is a continuation of U.S. patent
application Ser. No. 10/055,256, filed Jan. 22, 2002, the entire
contents of which are incorporated herein by reference.
TECHNICAL FIELD
[0002] The present invention generally relates to magnetic
resonance and other biological scan data.
BACKGROUND
[0003] Magnetic resonance imaging is a complex interaction between
protons in biological tissues, a static and alternating magnetic
field (the magnet), and energy in the form of radio-frequency waves
of a specific frequency (RF), introduced by coils placed next to
the subject. The energy state of the hydrogen protons is
transiently increased. The subsequent return to equilibrium
(relaxation) of the protons results in the release of RF energy
which can be measured by the same surface coils that delivered the
RF pulses. The RF energy, also referred to as the RF signal or
echo, is complex and is thus transformed by Fourier analysis into
useful information used to form an MR image.
SUMMARY
[0004] The present invention provides apparatus and methods for
processing data associated with magnetic resonance (MR) scanning.
In particular, in one embodiment, the present invention provides an
atlas comprising at least one value representative of a magnetic
property and, optionally, at least one value representative of
tissue type prior probability. In a further embodiment, the present
invention provides an atlas comprising a plurality of values
representative of magnetic properties of a plurality of spatial
locations of a plurality of subjects. In one embodiment, a system
is provided having both an MR scanner and an atlas of the present
invention. In a further embodiment, the invention provides methods
of making and using the atlas and system.
[0005] The apparatus and methods of the present invention provide a
model having data representative of one or more subjects. The data
includes magnetic property values, optionally, tissue type prior
probability values. The atlas can be used to automatically align an
MR scan, such as a localizer scan, to obtain a specific geometry of
the data acquired during a subsequent scan. The atlas may also be
used to automatically identify, or segment, tissue type of a
subject based on MR scan data of the subject.
[0006] According to one embodiment of the invention, an atlas is
provided comprising a plurality of values representative of a
magnetic property of a plurality of spatial locations of a subject
as determined by magnetic resonance. According to a further
embodiment, an atlas is provided comprising values representative
of a statistical representation of a magnetic property of a
plurality of spatial locations of a plurality of subjects. The
present invention also provides a system comprising an MR scanner
and an atlas. For example, the atlas may contain magnetic property
data. The system can be used to automatically align an MR scan,
such as a localizer scan, to obtain a specific orientation of the
data acquired during a subsequent scan. The system may also be used
to automatically identify, or segment, tissue type of a subject
based on MR scan data of the subject.
[0007] Methods of using the atlas are further provided herein. In
one embodiment, a method of using the atlas having magnetic
property values to obtain a specific geometry of data to be
acquired during a subsequent scan is provided. In a variation of
this embodiment, a method of using the atlas may additionally
involve tissue type probabilities.
[0008] Methods of using the atlas are further provided herein. In
one embodiment, a method of using the atlas having magnetic
property values to determine tissue type is provided. In a
variation of this embodiment, a method of using the atlas may
additionally involve tissue type probabilities.
[0009] According to a further embodiment of the invention, a method
is provided for obtaining information about a subject having the
steps of providing a magnetic resonance scanner, providing an atlas
having magnetic resonance data derived from at least one other
subject and processing information received from the scanner
pertaining to the subject. Also included are the steps of reading
the atlas and determining alignment of the magnetic resonance scan
to obtain a specific geometry of a subsequent magnetic resonance
scan.
[0010] According to another embodiment of the invention, another
method is provided for obtaining information about a subject. This
method involves the steps of providing magnetic property values
corresponding to tissue types pertaining to the subject, providing
an atlas having magnetic property values derived from at least one
other subject, along with labeling tissue types of a tissue
corresponding to the magnetic resonance property values pertaining
to the subject by using the atlas having the magnetic resonance
values derived from at least one other subject.
[0011] According to a further embodiment of the invention, a method
is provided for creating an atlas by providing a first magnetic
resonance modality volume pertaining to a subject, divided into
voxels, and recording a magnetic property value in a node of the
atlas corresponding to a voxel of the first magnetic resonance
modality volume.
[0012] Another embodiment of the invention involves a method for
creating an atlas. A first magnetic resonance modality volume is
provided pertaining to a subject and divided into voxels. A labeled
volume is provided indicating tissue types of tissue corresponding
to the voxels. Distortion of the first magnetic resonance modality
volume is corrected. Magnetic property distribution parameters are
extracted for each tissue type identified at each voxel. Also,
magnetic property data is recorded corresponding to each tissue
type in a node of the atlas corresponding to a voxel of the first
magnetic resonance modality volume.
[0013] According to another embodiment, a method for creating an
atlas is provided wherein a voxel intensity is obtained from an
image representative of at least one magnetic modality of a voxel
of a subject, a magnetic property value is derived from the voxel
intensity, and the magnetic property value is written to a node of
the atlas corresponding to the voxel.
[0014] A further embodiment of the invention provides a method for
processing an image of a subject. An atlas is provided having
magnetic property values derived from at least one other subject.
The image is aligned to the atlas, and the image is segmented into
segments. The segments are labeled to designate a tissue type of a
tissue corresponding to the magnetic property values pertaining to
the subject by the use of the atlas. An image is thus obtained
pertaining to magnetic property values of the subject.
[0015] It will further be appreciated that in the methods of the
present invention, distortion may be corrected prior to entering
the data into the atlas, as well as prior to processing newly
acquired data in conjunction with the atlas.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The invention will be apparent from the description herein
and the accompanying drawings, in which like reference characters
refer to the same parts throughout the different views.
[0017] FIG. 1 provides a subject and a grid pattern illustrating
voxels of a subject;
[0018] FIG. 2 illustrates an atlas;
[0019] FIGS. 3-8 illustrate nodes of an atlas according to various
embodiments of the invention;
[0020] FIGS. 9A and 9B illustrate sample data for determination of
the content of a node of an atlas;
[0021] FIG. 10 provides a sample method for the creation of an
atlas;
[0022] FIG. 11 provides a sample method for the registration of MR
data to an atlas; and
[0023] FIG. 12 provides a functional schematic of a system
according to an embodiment of the invention.
DETAILED DESCRIPTION
[0024] The present invention, in various embodiments, involves an
atlas containing values representative of magnetic properties of a
magnetic resonance (MR) scan and optionally prior probability data
relating to tissue type. Further embodiments of the invention
involve a system including an MR scanner and the atlas for use, for
example, in alignment of an MR scan and for automatic segmentation
of an MR scan. Methods of creating and using the atlas and system
are also provided.
[0025] As used herein, the following terms are defined as
follows:
[0026] T1 and T2 relaxation times: The rate of return to
equilibrium of perturbed protons is referred to as the relaxation
rate. The relaxation rate is different for different normal and
pathologic tissues. The relaxation rate of a hydrogen proton in a
tissue is influenced by surrounding molecular environment and
atomic neighbors. Two relaxation rates, the T1 and T2 relaxation
times, may be measured. The T1 relaxation rate is the time for 63%
of the protons to return to their normal equilibrium state, while
the T2 relaxation rate is the time for 63% of the protons to become
dephased owing to interactions among adjacent protons. The
intensity of the signal and thus the image contrast can be
modulated by altering certain parameters, such as the interval
between RF pulses (TR) and the time between the RF pulse and the
signal reception (TE). So-called T1-weighted (T1W) images are
produced by keeping the TR and TE relatively short. Under these
conditions, contrast between structures is based primarily on their
T1 relaxation differences. T2-weighted (T2W) images are produced by
using longer TR and TE times.
[0027] TR: The time between repetitions of RF in an imaging
sequence.
[0028] TE: The time between the RF pulse and the maximum in the
echo in a spin-echo sequence.
[0029] Flip Angle: The angle that the magnetic moment vector
rotates when applying a B1 RF pulse field.
[0030] T1: The time to reduce the difference between the
longitudinal magnetization and its equilibrium magnetization by an
exponential factor.
[0031] T2: The time to reduce the transverse magnetization by an
exponential factor.
[0032] PD: The concentration of spins.
[0033] T1-weighted: A magnetic resonance image where the contrast
is predominantly dependent on T1.
[0034] T2-weighted: A magnetic resonance image where the contrast
is predominantly dependent on T2.
[0035] PD-weighted: A magnetic resonance image where the contrast
is predominantly dependent on PD.
[0036] Diffusion-weighted: A magnetic resonance image where the
contrast is predominantly dependent on diffusion weighting
gradient.
[0037] Magnetization Transfer-weighted: A magnetic resonance image
where the contrast is predominantly dependent on magnetization
transfer saturation effect.
[0038] Tissue Type: As used herein, "tissue type" can be used to
designate a classification or characteristic of a tissue, such as
tissue within a voxel. For example, when used with a human brain as
the subject, tissue type can include, without limitation, gray
matter, white matter and cerebral spinal fluid. Optionally, the
tissue type can be more specific, such as referring to anatomical
structure. For example, in the case of a brain as the subject, the
tissue type may designate gray matter and/or, more specifically,
hippocampus, or other appropriate anatomical structure label. In
another example, in the case of a spine as a subject, the tissue
type may designate bone, and/or more specifically vertebral bodies,
or other appropriate anatomical structure labels. In yet another
example, in the case of the kidney as a subject, the tissue type
may designate the cortex, and/or more specifically nephrons, or
other appropriate anatomical structure labels.
[0039] Localizer scan: A low-resolution scan acquired at the
beginning of a scanning procedure to estimate the precision of the
acquisition geometry relative to the subject to be imaged.
[0040] Subsequent scan: A high-resolution scan acquired on the
basis of the localizer geometry, such as orientation, dimensions,
or voxel size.
[0041] Magnetic property: A magnetic property of protons, such as
T2, T1, PD, diffusion or magnetization transfer.
[0042] The present invention is applicable to a wide variety of MR
scans of a subject including mammals (e.g. humans), as well as
specific portions of a subject (e.g., organ, limb, or a portion of
an organ or limb), also referred to herein as the "subject". Each
subject is divided in three-dimensional space into voxels. With
reference to FIG. 1, a subject 100, such as a human brain, is shown
with an illustrative grid pattern 200 signifying the locations of
voxels 210. Each voxel 210 represents a three-dimensional portion
of the subject 100. A voxel 210 may be of various dimensions and
can have different dimensions along different axes within the
subject 100.
[0043] As shown in FIG. 2, an atlas 300 is provided according to an
embodiment of the invention. While illustrated as a
three-dimensional structure, the invention is not so limited, as
the atlas 300 may be formed of any of a variety of data structures
as will be apparent to one of ordinary skill in the art. The atlas
300 includes nodes 310. According to an embodiment of the
invention, each node 310 corresponds to a voxel 210 (cf. FIG. 1)
representing a portion of the subject 100. Alternatives of the
invention may involve fewer nodes 310 than voxels 210. In such a
case, a node 310 may be configured to represent a plurality of
voxels 210 or the nodes 310 may represent only a subset of the
overall voxels 210.
[0044] FIGS. 3-8 provide various configurations of the nodes 310
according to alternative embodiments of the invention. Each node
310 is configured to store information relating to the
corresponding voxel 210. As shown in FIG. 3, the node 310 may be
configured to have a magnetic property 320 corresponding to the
voxel 210. magnetic properties may include, but are not limited to,
T1, T2, proton density (PD), T2*, magnetization transfer, diffusion
tensor and derived variables, such as anisotropy and diffusivity.
According to one embodiment of the invention, the magnetic
properties may be computed from the images, based on a forward
model, and the MR acquisition parameters, including, but not
limited to, TR, TE, and flip angle. Determination of such magnetic
properties and details regarding the MR acquisition parameters can
be found in Magnetic Resonance Imaging, Physical Principle and
Sequence Design, E. M. Haacke et al., Wiley-Liss, 1999, pp.
637-667, which is incorporated herein by reference.
[0045] Optionally, a second magnetic property 330 corresponding to
the voxel 210 may also be stored in the node 310. Additional
magnetic properties may also be stored in the node 310.
[0046] A tissue type prior probability 340 corresponding to a
tissue type found in the voxel 210 may optionally be stored in the
node 310. When used with a human brain as the subject, tissue type
can include, without limitation, gray matter, white matter and
cerebral spinal fluid. Optionally, the tissue type can be more
specific, such as referring to anatomical structure. For example,
in the case of a brain as the subject, the tissue type may
designate gray matter and/or, more specifically, the hippocampus,
or other appropriate anatomical structure label. In another
example, in the case of the spine as a subject, the tissue type may
designate bone, and/or more specifically vertebral bodies, or other
appropriate anatomical structure labels. In yet another example, in
the case of the kidney as a subject, the tissue type may designate
the cortex, and/or more specifically nephrons, or other appropriate
anatomical structure labels. It will be appreciated that the tissue
type of the voxel 210 may be determined by human labeling or may be
determined by other known methods such as an algorithm (e.g.
Adaptive Segmentation of MRI Data, Wells W M, at al., IEEE
Transactions on Medical Imaging, 1996;15:429-442 (corrected version
available at
http://citeseer.nj.nec.com/cache/papers/cs/19782/http:zSzzSzsplweb.bwh.ha-
rvard.edu:8000zSzpageszSzpplzSzswzSzpaperszSztmi-96.pdf/wells96adaptive.pd-
f), Statistical Approach to Segmentation of Single-Channel Cerebral
MR Images, Rajapakse J C, et al., IEEE Transactions on Medical
Imaging, 1997, Vol. 16, No. 2: 176-86, and Automated Model-Based
Bias Field Correction of MR Images of the Brain, Van Leemput, K. et
al., IEEE Transactions on Medical Imaging, 1999, Vol. 18, No. 10),
which are incorporated herein by reference.
[0047] According to a further embodiment of the invention, a node
310 may include a tissue type prior probability 340 corresponding
to a tissue type found in the voxel 210, as illustrated in FIG. 4.
According to this embodiment, a first magnetic property 320 is also
stored. Optionally, a second magnetic property 330, or additional
magnetic properties, may also be stored in the node 310.
[0048] According to a further embodiment of the invention, as shown
in FIG. 5, one or multiple magnetic properties may be determined
for each of the tissue types located at the corresponding voxel
210. Therefore, as shown by way of example in FIG. 5, if a voxel
210 has two tissue types located at the voxel 210, as determined
from a plurality of subjects, one or more magnetic properties 320,
330 may be stored for each of the tissue types. As shown in FIG. 5,
a value of a first magnetic property 320 may be stored for the
tissue type 1 at the corresponding voxel. Optionally, a value of a
second magnetic property 330 may also be stored for tissue type 1.
Separate magnetic properties 320, 330 may also be stored for the
values corresponding to the tissue of tissue type 2. This variation
of the invention is useful in conjunction with an atlas 300 formed
of information from more than one subject 100. A tissue type
probability 340 may also be optionally stored in the node 310 for
one or more of the tissue types detected at the corresponding voxel
210.
[0049] In a further embodiment, illustrated by way of example in
FIG. 6, a tissue type prior probability 340 may be stored at a node
310 for each tissue type located at a corresponding voxel 210. A
magnetic property 320 is also stored at the node 310 for each
tissue type. Optionally, one or more further magnetic properties
330 may also be stored at the node 310.
[0050] A further embodiment of a node 310 is illustrated in FIG. 7.
The node 310 of FIG. 7 provides a tissue type prior probability 340
and statistical data pertaining to a magnetic property of the
tissue of a corresponding voxel 210, relative to a plurality of
subjects. As shown by way of example in FIG. 7, a mean 322 of the
values of a first magnetic property for a first tissue type at the
corresponding voxel 210 is provided. A variance 324 of the values
of a first magnetic property for the first tissue type at the
corresponding voxel 210 is also provided.
[0051] The node 310 of FIG. 7 may also optionally include
statistical data pertaining to one or more additional magnetic
properties, such as a mean 332 and variance 334 of a second
magnetic property.
[0052] The node 310 of FIG. 7 is also optionally suitable for use
with an atlas 300 containing information from a plurality of
subjects 100. Any of the data 322, 324, 332, 334, 340 as described
above in relation to a first tissue type, may also be determined in
relation to a second tissue type and stored.
[0053] FIG. 8 illustrates a node 310 of a further embodiment of the
invention providing statistical data, such as a mean 322 and a
variance 324, of the values of a first magnetic property for a
first tissue type at a corresponding voxel 210. Optionally, further
statistical data 332, 334 or a tissue type prior probability 340
may be provided. Similar information 322, 324, 332, 334, 340 may
also be optionally provided relating to further tissue types at a
corresponding voxel 210.
[0054] As illustrated by way of example in FIGS. 9A and 9B, the
determination of a mean 322 and a variance 324 for a first magnetic
property can be determined. FIG. 9A provides a table 400 having the
sample magnetic property values for an analogous voxel of each of
three subjects. FIG. 9B illustrates the three steps 410, 420, 430
involved in determining the content of the node 310 corresponding
to the illustrative voxel. As shown in step 1, 410, the tissue type
is 1, the mean of the value is 100 and the variance is 0. The prior
probability of this node corresponding to a voxel of tissue type 1
is 1. Step 2, 420, adds the data of the second subject to the data
already tabulated from the first subject. Therefore, the mean now
rises to 150, while the remaining data is unchanged, as the tissue
type is 1 for both subjects, leaving the prior probability at
1.
[0055] Step 3, 430, illustrates a node configuration illustrated in
FIG. 7 or 8 by the tabulation of statistical data per tissue type
for each node. Because the tissue type for the third subject is 2,
a second set of statistical data is tabulated for the new tissue
type, while the first set of data is updated in view of the third
subject. The mean and variance of tissue type 1 remain unchanged.
The prior probability of tissue type 1 however, now changes to 2/3.
The mean of tissue type 2 is 50, and the variance 0. The prior
probability of tissue type 2 is 1/3.
[0056] In another embodiment, additional data may be stored at each
node relating to the corresponding voxel or a representation
thereof. For example, image intensity data, expressed in arbitrary
units, may be stored. Alternatives include those apparent to one of
skill in the art.
[0057] In another embodiment, global prior probabilities may be
stored in the atlas of the present invention. Global probabilities
indicate the overall prior probability of something, such as a
tissue type appearing in a particular area, or anywhere, in a
subject. The global mean and variance of various magnetic
properties may also be determined and stored for each tissue type.
Such global values may be stored at a variety of locations in the
atlas, such as in a header, or alternatively, at each node.
[0058] As shown by way of example in FIG. 10, a method 500 is
provided according to an embodiment of the invention for the
creation of an atlas 300. The atlas is built from one or more
subject data sets 510, 512, 514. A subject data set may contain at
least one MR scan 516, 518, 520 of a subject (e.g. an organ or a
portion of an organ). The MR scans can be, but are not limited to,
T1, T2, proton density (PD), T2*, magnetization transfer, diffusion
tensor or derived variables such as anisotropy and diffusivity.
[0059] Distortions are then corrected in the MR scan 516, step 530.
Corrections of distortion are known to one of ordinary skill in the
art and are discussed in more detail in relation to FIG. 11
herein.
[0060] According to one embodiment, a subject's data set used in
creating or adding to an atlas can also contain a labeled
representation 522, 524, 526 of the MR scan(s), such as a segmented
volume identifying each tissue type/anatomical structure. The
labeled representation can be obtained by way of manual labeling
(e.g. by experienced anatomists) and/or by way of automatic
segmentation methods as described by way of example in Wells,
supra, Statistical Approach to Segmentation of Single-Channel
Cerebral MR Images, Rajapakse J C, et al., IEEE Transactions on
Medical Imaging, 1997, Vol. 16, No. 2: 176-86, and Automated
Model-Based Bias Field Correction of MR Images of the Brain, Van
Leemput, K. et al., IEEE Transactions on Medical Imaging, 1999,
Vol. 18, No. 10, which are incorporated herein by reference.
[0061] Next, the tissue type and corresponding magnetic property
statistical distribution data is extracted from the corrected
subject data set 510, step 540.
[0062] A high-resolution temporary atlas 560, step 550, is then
created by storing the tissue type and corresponding magnetic
property statistical data of each voxel 210 of the subject, in each
corresponding node 310 of the atlas 300.
[0063] The high-resolution temporary atlas 560 may then be used as
the atlas 300 if the atlas 300 is to only have data pertaining to a
single subject.
[0064] However, if additional subjects are to be added, the method
500 continues with the subject data set 512 of a second subject,
and, optionally subject data sets 514 of additional subjects.
Correction of distortion, step 530, and extraction of statistical
data 540 is conducted as in relation to the first subject data set
510.
[0065] After each additional subject data set 512, 514 is
processed, the tissue type and corresponding magnetic property
statistical data of each voxel 210 of the subject is registered, or
aligned, with the existing node structure of the atlas 300, step
570. During registration, the data, such as tissue type and
magnetic statistical data, corresponding to the voxels 210 of the
subject, is manipulated to correspond to the analogous voxels 210
represented by the node 310 structure of the atlas. Further details
of registration, step 570, are discussed in detail in relation to
FIG. 11 herein.
[0066] Next, the additional data, such as tissue type and magnetic
statistical data, is then added to the atlas 300 by updating the
atlas parameters, step 580. As shown in FIG. 10, a high-resolution
atlas 565 is produced after the addition of two subject data sets
510, 512 to the atlas 300. This high-resolution atlas 565 may be
used as an atlas 300, or additional subject data sets 514 may be
added.
[0067] When the desired N subject data sets have been added to the
atlas, the atlas may optionally be subsampled, step 590 to create
the atlas 300. As discussed herein, alternatives of the invention
may involve fewer nodes 310 than voxels 210. In such a case, a node
310 may be configured to represent a plurality of voxels 210 or the
nodes 310 may represent only a subset of the overall voxels 210.
Such a reduced resolution may also be generated by the subsampling,
step 590, by combining data from multiple voxels into one node.
Also, only a portion of the voxels representing a portion of the
subject may be used in the atlas 300.
[0068] An atlas 300 of the present invention may be customized for
a specialized purpose. The atlas may have values of a statistical
representation that are population-specific (e.g., related to age,
sex and/or pathology), scanner-specific (e.g., related to
manufacturer and/or scanner model), and/or acquisition
sequence-specific (e.g., related to flash and/or inversion
recovery). Acquisition sequences can include including, without
limitation, at least one from the group of PD-, T2-, T1-,
diffusion-, and magnetization transfer-weighted. Acquisition
sequence-specific values may involve magnetic resonance sequence
parameters, including, without limitation, at least one from the
group of TR, TE and flip angle.
[0069] An atlas of the present invention may be oriented to various
coordinate systems. One such example of a coordinate system is a
Cartesian coordinate system, such as a Right Anterior Superior
(RAS) coordinate system, used in orienting an image relative to a
subject, or an arbitrarily determined coordinate system.
[0070] An atlas of the present invention may be created at various
spatial resolutions. An atlas may further be sub-sampled to reduce
the resolution and data required and time required for
calculations. The resolution may also vary within an atlas,
allowing greater resolution at areas of interest.
[0071] According to one embodiment of the invention, an atlas may
be constructed as shown in FIGS. 10 and 11. Optionally, an atlas
may be formed by data from only one subject. An atlas may be formed
by N subjects, which may be determined by monitoring the change of
values within the atlas upon the addition of each additional
subject. According to another embodiment, when the values stored in
the nodes of the atlas no longer vary within a statistical range of
confidence, the addition of further subjects is no longer
required.
[0072] The registration of data onto the atlas may comprise the
determination of at least 6 parameters. For example, those
parameters can be 3 translation shifts, 3 scaling factors and 3
rotation angles relatively to the 3 orthogonal directions of the
atlas coordinate system.
[0073] Further detail regarding registration of data onto an atlas,
or temporary atlas as described in FIG. 10, is illustrated by way
of example in the method 700 of FIG. 11. In FIG. 11, a method 700
is provided according to an embodiment of the invention for the
registration of MR data to an atlas 300. The example method 700 of
FIG. 11 is also applicable to prior probability data or any other
data types for association to nodes 310 of the atlas 300.
[0074] An initial set of registration parameters is provided, step
710, along with an initial bias estimate, step 720, according to
methods known to one of skill in the art in relation to atlases
having other types of data. See, for example, Wells, supra,
Automated Model-Based Bias Field Correction of MR Images of the
Brain, Van Leemput, K. et al., IEEE Transactions on Medical
Imaging, 1999, Vol. 18, No. 10, and Automatic Scan Prescription for
Brain MRI, Itti, L. et al., Magnetic Resonance in Medicine, 2001,
Vol. 45: 486-494, which are incorporated herein by reference. The
initial bias estimate of step 720 adjusts for intensity fall-off in
the portions of the image away from the image center.
[0075] A magnetic resonance (MR) volume is also provided, step 730.
The magnetic resonance volume can be generated by deriving a
magnetic property value for a voxel from a voxel intensity value of
a corresponding voxel of an image containing magnetic resonance
data.
[0076] A bias correction is applied to the MR volume, step 740.
With regard to step 740, and step 530 of FIG. 10, distortion and
bias can be caused by a variety of factors. For example, the
distortion and bias can be subject-dependent, such as from, but not
limited to, chemical shift, magnetic susceptibility, and/or
per-acquisition motion. Alternatively or in addition, distortion
and bias can be scanner-dependent, such as from, but not limited
to, gradients non-linearity, main magnetic field non-homogeneity
and/or eddy currents. Maxwell effects are a further source of
potential distortion or bias. Correction of such distortion and
bias are known to one of ordinary skill in the art.
[0077] As shown in FIGS. 10 and 11, bias and distortion are
corrected prior to incorporating the data into the atlas. According
to a further embodiment of the invention, distortion and bias are
corrected prior to processing data in conjunction with the
atlas.
[0078] A transform is applied to the bias-corrected MR volume, step
750. Linear transformations (e.g. translation, rotation, scaling)
are applied to images via homogeneous matrices. According to one
embodiment, they are 4.times.4 matrices, wherein the 3 first bottom
elements always equal 0 and the last bottom elements always equals
1. Any transformation can be decomposed into a translation, a
rotation and a scaling matrices. The final homogeneous matrix is
then a multiplication of those 3 matrices. Details are given by way
of example below: [0079] tx, ty and tz being the translation
parameters in the x, y and z directions, the translation
homogeneous matrix is given by:
[0079] ( 1 0 0 t x 0 1 0 t y 0 0 1 t z 0 0 0 1 ) ##EQU00001##
[0080] xs, ys and zs being the scaling parameters in the x, y and z
directions, the scaling homogeneous matrix is given by:
[0080] ( x s 0 0 0 0 y s 0 0 0 0 z s 0 0 0 0 1 ) ##EQU00002##
[0081] .theta., .phi. and .phi. being the rotation parameters
relatively to the x, y and z axis, the rotation homogeneous matrix
is given by:
[0081] ( cos .PHI.cos .phi. + sin .PHI. sin .theta.sin .phi. sin
.PHI.cos .theta. - cos .PHI.sin .theta.sin .phi. cos .theta.sin
.PHI. 0 - sin .PHI.cos .theta. cos .PHI. cos .theta. sin .theta. 0
sin .PHI.sin .theta.cos .phi. - cos .PHI.sin .theta. - cos .PHI.sin
.theta. - sin .PHI. sin .theta. cos .theta.cos .PHI. 0 0 0 0 1 )
##EQU00003##
[0082] The voxels, or MRI points, corresponding to nodes 310 of the
atlas 300 are segmented based on a Maximum A Posteriori (MAP)
estimator, step 760. The MAP estimator is a probability computation
with statistical information stored in the atlas. The MAP estimator
and its use with other types of data are known to one of ordinary
skill in the art, as illustrated by way of example in Wells, supra,
Statistical Approach to Segmentation of Single-Channel Cerebral MR
Images, Rajapakse J C, et al., IEEE Transactions on Medical
Imaging, 1997, Vol. 16, No. 2: 176-86, and Automated Model-Based
Bias Field Correction of MR Images of the Brain, Van Leemput, K. et
al., IEEE Transactions on Medical Imaging, 1999, Vol. 18, No. 10,
which are incorporated herein by reference.
[0083] The registration parameters and bias estimation are then
updated, step 770, as is known to one of ordinary skill in the art,
as illustrated by way of example in Wells, supra, Automated
Model-Based Bias Field Correction of MR Images of the Brain, Van
Leemput, K. et al., IEEE Transactions on Medical Imaging, 1999,
Vol. 18, No. 10, and Multimodality Image Registration by
maximization of Mutual Information, Maes, F. et al., IEEE
Transactions on Medical Imaging, 1997, Vol. 16, No. 2, which are
incorporated herein by reference. If the target MAP is not reached,
the process repeats, step 780, beginning again with application of
bias correction to the MR volume, step 740.
[0084] If the target MAP is reached, the registration matrix is
provided, step 790. The registration matrix can include sixteen
(16) values, including translation parameters, scaling parameters,
and a combination of the sines and cosines of rotation parameters.
The registration matrix can be used to obtain a specific geometry
(e.g. orientation and/or dimensions) of the data acquired during a
subsequent scan, as discussed herein.
[0085] The MR volume corrected for bias is also provided, step 800,
allowing more accurate computation of the magnetic property values
for each node of the atlas.
[0086] Further information regarding the details of the steps of
FIG. 11 can be found in Wells, supra, Automated Model-Based Bias
Field Correction of MR Images of the Brain, Van Leemput, K. et al.,
IEEE Transactions on Medical Imaging, 1999, Vol. 18, No. 10, and
Multimodality Image Registration by maximization of Mutual
Information, Maes, F. et al., IEEE Transactions on Medical Imaging,
1997, Vol. 16, No. 2, which are incorporated herein by
reference.
[0087] According to a further embodiment of the invention, a system
900 is provided as shown by way of example in FIG. 12. A scanner
910 is provided to capture magnetic images. A processor 920 is
provided to interface with the scanner 910 and the atlas 300 in
order to conduct the methods according to various embodiments of
the present invention.
[0088] The atlas and system 900 of the present invention may be
used in a variety of applications. In one embodiment, a method of
using the atlas with magnetic property data and optionally with
tissue (or anatomical structure) type prior probabilities is
provided, automatically align an MR scan, such as a localizer scan,
to obtain a specific geometry of the data acquired during a
subsequent scan (auto-slice prescription). Further details of this
implementation can be found in U.S. Pat. No. 6,195,409, issued Feb.
27, 2001, to Chang et al., which is incorporated herein by
reference. In an additional embodiment, a method of using the atlas
with magnetic property data to determine anatomical structure or
detect abnormal tissue (auto-segmentation) is provided. Further
details of this implementation can be found in Wells, supra,
Statistical Approach to Segmentation of Single-Channel Cerebral MR
Images, Rajapakse J C, et al., IEEE Transactions on Medical
Imaging, 1997, Vol. 16, No. 2: 176-86, and Automated Model-Based
Bias Field Correction of MR Images of the Brain, Van Leemput, K. et
al., IEEE Transactions on Medical Imaging, 1999, Vol. 18, No. 10,
which are incorporated herein by reference.
[0089] It will further be appreciated that in the methods of the
present invention, including the applications described herein,
distortion of newly obtained data may optionally be corrected prior
to processing data in conjunction with the atlas. Further details
of distortion correction can be found in Sources of Distortion in
Functional MRI Data, Jezzard, P. et al., Human Brain Mapping, 1999,
Vol. 8:80-85, which is incorporated herein by reference.
[0090] The present invention has been described by way of example,
and modifications and variations of the described embodiments will
suggest themselves to skilled artisans in this field without
departing from the spirit of the invention. Aspects and
characteristics of the above-described embodiments may be used in
combination. The described embodiments are merely illustrative and
should not be considered restrictive in any way. The scope of the
invention is to be measured by the appended claims, rather than the
preceding description, and all variations and equivalents that fall
within the range of the claims are intended to be embraced therein.
The contents of all references, databases, patents and published
patent applications cited throughout this application are expressly
incorporated herein by reference.
* * * * *
References