Difference between revisions of "Projects:DTIFiberRegistration"

From NAMIC Wiki
Jump to: navigation, search
 
(25 intermediate revisions by 6 users not shown)
Line 1: Line 1:
 +
Back to [[NA-MIC_Internal_Collaborations:DiffusionImageAnalysis|NA-MIC Collaborations]], [[Algorithm:MIT|MIT Algorithms]]
 +
__NOTOC__
 +
= Joint Registration and Segmentation of DWI Fiber Tractography =
 +
 +
The purpose of this work is to jointly register and cluster DWI fiber tracts obtained from a group of subjects. We formulate a
 +
maximum likelihood problem which the proposed method solves using a
 +
generalized Expectation Maximization (EM) framework. Additionally,
 +
the algorithm employs an outlier rejection and denoising strategy to
 +
produce sharp probabilistic maps (an atlas) of certain bundles of interest.
 +
This atlas is potentially useful for making diffusion measurements in
 +
a common coordinate system to identify pathology related changes or
 +
developmental trends.
 +
 
= Description =
 
= Description =
  
Collaborators: Ulas Ziyan (MIT), Mert Sabuncu (MIT), Lauren O'Donnell (BWH), Carl-Fredrik Westin (BWH)
+
''Initial Registration''
  
Image registration is necessary to compare and combine information
+
A spatial normalization is necessary to obtain a group-wise
from a group of subjects. Affine registration
+
clustering of the resulting fibers. This initial normalization is performed
(e.g. Talairach normalization) is common for group analysis, but
+
on the Fractional Anisotropy (FA)
this typically yields poor alignment accuracy in certain local
+
images generated for each subject. This initial normalization aims
regions of interest.  
+
to remove gross differences across subjects due to global head size
 +
and orientation. It is thus limited to a 9 parameter affine
 +
transformation that accounts for scaling, rotation and translation.
 +
The resulting transformations are then applied to each of the
 +
computed fibers to map them into a common coordinate frame for
 +
clustering.
  
To address this, there has been interest in
 
developing nonlinear registration methods. In nonlinear
 
registration, the description and regularization of the deformation
 
is critical, since an unconstrained registration algorithm suffers
 
form the potential problem of overfitting to the images, which may
 
undermine consecutive analyses. A popular approach is to employ a
 
model-based regularization which may have no biological meaning.
 
Yet, we believe that the regularization and description of the
 
deformation should be grounded in the application.
 
  
In this project, we
+
[[Image:MIT_DTI_JointSegReg_ourapproach.jpg |thumb|400px|Our Approach]]
explore the use of anatomical structures (called fiber
 
bundles) extracted from Diffusion MR Images of a group of subjects
 
for regularizing a non-linear registration algorithm.
 
  
'''Segmentation'''
+
''Initial Fiber Clustering''
  
 
Organization of tract fibers into bundles, in the entire white
 
Organization of tract fibers into bundles, in the entire white
Line 33: Line 39:
 
anatomical labels to form an atlas. In this work,  
 
anatomical labels to form an atlas. In this work,  
 
we take advantage of automatically segmented tractography that has been labeled (as
 
we take advantage of automatically segmented tractography that has been labeled (as
bundles) with such an atlas.
+
bundles) with such an atlas for initialization.
 +
 
 +
 
 +
''Joint Registration and Segmentation''
 +
 
 +
Once we obtain an initial affine registration and clustering results using the high dimensional atlas, we iteratively fine-tune the registration and clustering results using a maximum likelihood framework, which is solved through a generalized EM algorithm. For the registration we use one set of affine parameters per fiber bundle, and combine these affine registrations into a single smooth and invertable warp field using a log-Euclidian poly-affine framework. Additionally, the algorithm employs an outlier rejection and denoising strategy while
 +
producing sharp probabilistic maps of certain bundles of interest.
 +
 
 +
We tested the registration component of this algorithm without updating the clustering with 26 major fiber bundles. The poly-affine warp fields with relatively limited number of components resulted in similar quality registrations when compared with results from a benchmark non-linear registration algorithm that was run on FA images:
 +
 
 +
{|
 +
|
 +
[[Image:FiberBundleReg.jpg|thumb|800px|Top Row: 3D renderings of the registered tracts of a subject (in green) and the template (in red) within 5mm of the central axial slice overlayed on the central FA slice of the template. ''Aff'' (left) stands for the FA based global affine, ''Dem'' (middle) for the demons algorithm and ''PA'' (right) for the proposed framework in this work. Arrows point to an area of differing qualities of registration. Overlapping of the red and green fibers is indicative of better registration. Bottom Row: Jacobian determinant images from the central slice of the volume: Yellow represents areas with small changes in size, and the shades of red and blue represent enlargement and shrinking, respectively. The Jacobian of the global affine registration is constant. The Jacobian of the demons algorithm is smooth due to the Gaussian regularization. The Jacobian of the new algorithm reflects the underlying anatomy because of the fiber bundle-based definition of the deformation.]]
 +
|
 +
|}
 +
 
 +
 
 +
Corpus Callosum, Cingulum and the Fornix were selected for further investigation because of the specific challenges they present. These three structures are in close proximity with each other, and that results in many mislabeled fibers when labeled using a high dimensional atlas (see figure below (left)). Their close proximity also results in a number of trajectories deviating from one structure to another. These are precisely the sorts of artifacts we wish to reduce through learning common spatial distributions of fiber bundles from a group of subjects.
 +
 
 +
{|
 +
|
 +
[[Image:MIT_DTI_JointSegReg_beforeandafter.jpg|thumb|1000px|Tracts from Fornix (in green) and Cingulum (in purple) bundles along with a few selected tracts from Corpus Callosum (in black) as labeled using the high dimensional atlas (left) and after the EM algorithm with tract cuts (right). The tractography noise is evident in the images on the left as tracts deviating from one bundle to another. Also, these images contain instances where the high dimensional atlas failed to label the tracts correctly. The EM algorithm is able to remove the segments of tract bundles that are not consistent from subject to subject.]]
 +
|
 +
|}
 +
 
 +
We also constructed two different atlases to compare the effects of labeling algorithms on the quality of resulting group maps. The first one is constructed using the initial labels from the high dimensional atlas. A second one is built using the proposed algorithm:
 +
 
 +
{|
 +
|
 +
[[Image:MIT_DTI_JointSegReg_atlas2D.jpg|thumb|600px|Spatial distributions of Corpus Callosum, Cingulum and Fornix bundles from three single slices overlaid on their corresponding FA images. These maps are constructed using two different methods. a)High dimensional atlas, c) Proposed algorithm.  The colorbars indicate the probability of each voxel in the spatial distribution of the corresponding fiber bundle. Note that the probabilities become higher in the central regions of the bundles and the number of sporadical voxels with non-zero probabilities decrease from left to right, indicating a sharper atlas through better registration and more consistent labeling of the subjects.]]
 +
|
 +
|}
  
'''Registration of Fiber Bundles'''
 
  
To align the bundles from two subjects, we
+
[[Image:MIT_DTI_JointSegReg_atlas3D.jpg|thumb|400px|Isoprobability surfaces of the spatial distributions of Fornix (in green) and Cingulum (in purple) bundles constructed from 15 subjects using the EM algorithm with tract cut operations. A few selected tracts from Corpus Callosum (in black) are also drawn to highlight the spatial proximity of the three bundles. These spatial distributions retain very little of the tractography noise that is apparent in the individuals' tract bundles.]]
compute 3 dimensional spatial probability maps for each fiber tract
 
bundle in each subject. The goal of our registration is to maximize
 
the correlation between two corresponding bundles' probability maps.
 
To achieve that, we employ a sequential
 
quadratic programming method to find the 9 affine
 
parameters that maximize the correlation between corresponding
 
probability maps in different subjects. For a pair of subjects, this
 
yields a set of affine transformations that relate each
 
corresponding bundle pair. The last step of our algorithm is to fuse
 
these affine transformations to achieve a global, invertible
 
non-linear deformation. There are several ways to achieve this. In
 
this study, we employed the log-Eucledian poly-affine framework which guarantees invertibility.
 
  
[[Image:Thalamus_algo_outline.png|640px]]
+
''Project Status''
  
This image is a schematic outline of spectral segmentation algorithm. (A) DTI data from an individual thalamic hemisphere, shown here as a single slice cuboid map (B) Initial graph corresponding to the sparse affinity matrix (C) Unordered affinity matrix (D) Ordered and clustered affinity matrix (E) Clusters in the original data space (F) Clusters in 3D.
+
* Working 3D implementation in Matlab and C.
  
== Publications ==
+
= Key Investigators =
  
Submitted for publication.
+
* MIT: Ulas Ziyan, Mert R. Sabuncu
 +
* Harvard DBP1: Carl-Fredrik Westin, Lauren O'Donnell
  
<br />
+
= Publications =
 +
''In Print''
  
== Software ==
+
* [http://www.na-mic.org/publications/pages/display?search=Projects%3ADTIFiberRegistration&submit=Search&words=all&title=checked&keywords=checked&authors=checked&abstract=checked&sponsors=checked&searchbytag=checked| NA-MIC Publications Database on DTI Fiber Registration]
  
The algorithms now are implemented in matlab.
+
[[Category: Registration]] [[Category: Segmentation]]

Latest revision as of 20:20, 11 May 2010

Home < Projects:DTIFiberRegistration
Back to NA-MIC Collaborations, MIT Algorithms

Joint Registration and Segmentation of DWI Fiber Tractography

The purpose of this work is to jointly register and cluster DWI fiber tracts obtained from a group of subjects. We formulate a maximum likelihood problem which the proposed method solves using a generalized Expectation Maximization (EM) framework. Additionally, the algorithm employs an outlier rejection and denoising strategy to produce sharp probabilistic maps (an atlas) of certain bundles of interest. This atlas is potentially useful for making diffusion measurements in a common coordinate system to identify pathology related changes or developmental trends.

Description

Initial Registration

A spatial normalization is necessary to obtain a group-wise clustering of the resulting fibers. This initial normalization is performed on the Fractional Anisotropy (FA) images generated for each subject. This initial normalization aims to remove gross differences across subjects due to global head size and orientation. It is thus limited to a 9 parameter affine transformation that accounts for scaling, rotation and translation. The resulting transformations are then applied to each of the computed fibers to map them into a common coordinate frame for clustering.


Our Approach

Initial Fiber Clustering

Organization of tract fibers into bundles, in the entire white matter, reveals anatomical connections such as the corpus callosum and corona radiata. By clustering fibers from multiple subjects into bundles, these common white matter structures can be discovered in an automatic way, and the bundle models can be saved with expert anatomical labels to form an atlas. In this work, we take advantage of automatically segmented tractography that has been labeled (as bundles) with such an atlas for initialization.


Joint Registration and Segmentation

Once we obtain an initial affine registration and clustering results using the high dimensional atlas, we iteratively fine-tune the registration and clustering results using a maximum likelihood framework, which is solved through a generalized EM algorithm. For the registration we use one set of affine parameters per fiber bundle, and combine these affine registrations into a single smooth and invertable warp field using a log-Euclidian poly-affine framework. Additionally, the algorithm employs an outlier rejection and denoising strategy while producing sharp probabilistic maps of certain bundles of interest.

We tested the registration component of this algorithm without updating the clustering with 26 major fiber bundles. The poly-affine warp fields with relatively limited number of components resulted in similar quality registrations when compared with results from a benchmark non-linear registration algorithm that was run on FA images:

Top Row: 3D renderings of the registered tracts of a subject (in green) and the template (in red) within 5mm of the central axial slice overlayed on the central FA slice of the template. Aff (left) stands for the FA based global affine, Dem (middle) for the demons algorithm and PA (right) for the proposed framework in this work. Arrows point to an area of differing qualities of registration. Overlapping of the red and green fibers is indicative of better registration. Bottom Row: Jacobian determinant images from the central slice of the volume: Yellow represents areas with small changes in size, and the shades of red and blue represent enlargement and shrinking, respectively. The Jacobian of the global affine registration is constant. The Jacobian of the demons algorithm is smooth due to the Gaussian regularization. The Jacobian of the new algorithm reflects the underlying anatomy because of the fiber bundle-based definition of the deformation.


Corpus Callosum, Cingulum and the Fornix were selected for further investigation because of the specific challenges they present. These three structures are in close proximity with each other, and that results in many mislabeled fibers when labeled using a high dimensional atlas (see figure below (left)). Their close proximity also results in a number of trajectories deviating from one structure to another. These are precisely the sorts of artifacts we wish to reduce through learning common spatial distributions of fiber bundles from a group of subjects.

Tracts from Fornix (in green) and Cingulum (in purple) bundles along with a few selected tracts from Corpus Callosum (in black) as labeled using the high dimensional atlas (left) and after the EM algorithm with tract cuts (right). The tractography noise is evident in the images on the left as tracts deviating from one bundle to another. Also, these images contain instances where the high dimensional atlas failed to label the tracts correctly. The EM algorithm is able to remove the segments of tract bundles that are not consistent from subject to subject.

We also constructed two different atlases to compare the effects of labeling algorithms on the quality of resulting group maps. The first one is constructed using the initial labels from the high dimensional atlas. A second one is built using the proposed algorithm:

Spatial distributions of Corpus Callosum, Cingulum and Fornix bundles from three single slices overlaid on their corresponding FA images. These maps are constructed using two different methods. a)High dimensional atlas, c) Proposed algorithm. The colorbars indicate the probability of each voxel in the spatial distribution of the corresponding fiber bundle. Note that the probabilities become higher in the central regions of the bundles and the number of sporadical voxels with non-zero probabilities decrease from left to right, indicating a sharper atlas through better registration and more consistent labeling of the subjects.


Isoprobability surfaces of the spatial distributions of Fornix (in green) and Cingulum (in purple) bundles constructed from 15 subjects using the EM algorithm with tract cut operations. A few selected tracts from Corpus Callosum (in black) are also drawn to highlight the spatial proximity of the three bundles. These spatial distributions retain very little of the tractography noise that is apparent in the individuals' tract bundles.

Project Status

  • Working 3D implementation in Matlab and C.

Key Investigators

  • MIT: Ulas Ziyan, Mert R. Sabuncu
  • Harvard DBP1: Carl-Fredrik Westin, Lauren O'Donnell

Publications

In Print