Identifying white matter fiber tracts using magnetic resonance imaging (MRI)

• US 9,568,580 B2
• Filed: 07/01/2009
• Issued: 02/14/2017
• Est. Priority Date: 07/01/2008
• Status: Active Grant

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Hagler, Donald J. Jr., Dale, Anders M.
• Primary Examiner(s): COOK, CHRISTOPHER L
• Application Number:
  US12/496,441
• Publication Number:
  US 20100004527A1
• Time in Patent Office: 2,785 Days
• Field of Search: 600/410, 600/415, 600/416
• US Class Current: 1/1
• CPC Class Codes: G01R 33/5602 by filtering or weighting based on different relaxation times within the sample, e.g. T1 weighting using an inversion pulse
• CPC Class Codes: G01R 33/5608 Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels image data processing or generation, in general G06T
• CPC Class Codes: G01R 33/56341 Diffusion imaging

Abstract

Systems, methods, and software are used for identifying fibers based at least in part on magnetic resonance imaging. A fiber tract atlas for a nervous system includes atlas voxels that each represent a different volume element of the nervous system; a first atlas voxel represents a first volume element of the nervous system. The fiber tract atlas also includes information on orientations of a first fiber tract in the first volume element of the nervous system. Magnetic resonance data is acquired from the nervous system of a subject. The magnetic resonance data includes data voxels; a first data voxel relates to the first atlas voxel. A diffusion vector is generated for the first data voxel based at least in part on the acquired magnetic resonance data. The fiber tract atlas is used to find a probability that the first data voxel represents the first fiber tract based at least in part on the generated diffusion vector and the information on the orientations of the first fiber tract in the first volume element.

Claims

1. A computer implemented method for determining how well magnetic resonance imaging data obtained from a subject corresponds to a given white matter fiber tract, the method comprising:

obtaining a fiber tract atlas for a nervous system, the fiber tract atlas comprising;

a plurality of atlas voxels that each represent a different volume element of the nervous system, the plurality of atlas voxels including a first atlas voxel that represents a first volume element of the nervous system, and

averaged information on probabilistic locations and orientations of a first fiber tract in the first volume element of the nervous system, wherein the averaged information on the probabilistic locations and the orientations of the first fiber tract includes a rank-two symmetric atlas diffusion tensor (T) having a first atlas diffusion tensor eigenvector (ν

), such that ν

′

Tν

is equal to a largest eigenvalue of T;

acquiring magnetic resonance data from the nervous system of a subject, the magnetic resonance data comprising a plurality of data voxels including a first data voxel that relates to the first atlas voxel;

calculating a diffusion tensor for the first data voxel;

generating a diffusion vector (ν

) for the first data voxel, the diffusion vector being a first eigenvector corresponding to a largest eigenvalue of the calculated diffusion tensor for the first data voxel; and

determining a probability P that the first data voxel of the magnetic resonance data of the subject corresponds to the first fiber tract, based at least in part on the generated diffusion vector (ν

) and the averaged information on the probabilistic locations and orientations of the first fiber tract in the first volume element, and without manual inspection of the generated diffusion vector, the orientation information, or magnetic resonance images associated with the magnetic resonance data, wherein the determined probability P is determined by combining the atlas diffusion tensor (T), the first atlas diffusion tensor eigenvector (ν

), and the diffusion vector (ν

) of the first data voxel according to the expression

2. The computer implemented method of claim 1, the method further comprising:

producing location information of a fiber tract of interest associated with the first data voxel using fiber density data, wherein the fiber density data is associated with the plurality of atlas voxels and includes streamline fiber counts to indicate a likelihood of an atlas voxel to the fiber tract of interest;

resampling the averaged information on the probabilistic locations and the orientations of the first fiber tract from an atlas space to a subject space for the subject, wherein the determined probability that the first data voxel corresponds to the first fiber tract is based in part on resampled location probability and resampled orientation information.

3. The computer implemented method of claim 1, wherein the nervous system includes a human nervous system, the determining the probability that the first data voxel corresponds to the first fiber tract is based at least in part on the generated diffusion vector and the fiber-specific average diffusion tensor based at least in part on measurements of a plurality of human nervous systems.

4. The computer implemented method of claim 1, wherein the averaged information on the probabilistic locations and the orientations of the first fiber tract comprises the first atlas diffusion tensor for the first atlas voxel, the fiber tract atlas further comprises an additional tensor for each other atlas voxel, the additional tensor for each other atlas voxel representing orientations of the first fiber tract in the atlas voxel, the method further comprising using the fiber tract atlas to determine an additional probability for each other data voxel, the additional probability for each other data voxel based at least in part on the generated diffusion vector for the data voxel and the additional tensor for an atlas voxel that relates to the data voxel, the additional probability for each other data voxel indicating a probability that the data voxel corresponds to the first fiber tract.

5. The computer implemented method of claim 1, further comprising determining that the first data voxel relates to the first atlas voxel by performing additional operations comprising:

registering a T₁-weighted image to an atlas space of the fiber tract atlas; and

resampling the data voxels into the atlas space.

6. The computer implemented method of claim 1, wherein the averaged information on the probabilistic locations and the orientations of the first fiber tract comprises the first atlas diffusion tensor, and the obtaining the fiber tract atlas comprises obtaining a fiber tract atlas that further includes a plurality of additional tensors, each additional tensor representing the orientations of additional fiber tracts in the first volume element of the nervous system.

7. The computer implemented method of claim 6, wherein the probability that the first data voxel corresponds to the first fiber tract comprises a first probability, the method further comprising using the fiber tract atlas to determine a plurality of additional probabilities based at least in part on the generated diffusion vector and the plurality of additional tensors, each additional probability representing a relative probability that the first data voxel corresponds to one of the additional fiber tracts.

8. The computer implemented method of claim 1, further comprising constructing the obtained fiber tract atlas by performing additional operations comprising:

acquiring additional magnetic resonance data from nervous systems of a plurality of additional subjects, the additional magnetic resonance data comprising a plurality of data voxels for each additional subject;

generating an additional diffusion tensor for each data voxel of each additional subject based at least in part on the additional magnetic resonance data; and

generating an average diffusion tensor for each atlas voxel of the fiber tract atlas by averaging the additional diffusion tensors across the additional subjects.

9. The computer implemented method of claim 1, wherein the generating the diffusion vector comprises generating an additional eigenvector (ν

) of a second diffusion tensor, the second diffusion tensor based at least in part on the acquired magnetic resonance data, and wherein the determining the probability comprises combining the rank-two symmetric tensor T, the eigenvector (ν

), and the additional eigenvector ν

according to the expression

10. The computer implemented method of claim 1, further comprising:

using the determined probability with the fiber tract atlas and the acquired magnetic resonance data to generate a fiber probability map for at least a first portion of the nervous system of the subject; and

using the probability map to determine at least one of a fractional anisotropy for the first fiber tract or a diffusivity for the first fiber tract.

11. The method of claim 1, wherein acquiring magnetic resonance data comprises:

applying a plurality of radio frequency pulses to the subject;

applying a plurality of gradient fields to the subject; and

detecting at least one echo signal from the subject.

12. A magnetic resonance imaging (MRI) system comprising:

a magnetic resonance data acquisition system adapted to acquire magnetic resonance data from a nervous system of a subject, the magnetic resonance data comprising a plurality of data voxels including a first data voxel; and

a data processor that is configured to;

receive the magnetic resonance data from the magnetic resonance data acquisition system;

obtain a fiber tract atlas for the nervous system from the magnetic resonance data, the fiber tract atlas comprising;

a plurality of atlas voxels that each represent a different volume element of the nervous system, the plurality of atlas voxels including a first atlas voxel that represents a first volume element of the nervous system, the first data voxel relating to the first atlas voxel; and

averaged information on probabilistic locations and orientations of a first fiber tract in the first volume element of the nervous system, wherein the averaged information on the probabilistic locations and the orientations of the first fiber tract includes a rank-two symmetric atlas diffusion tensor (T) having a first atlas diffusion tensor eigenvector (ν

), such that ν

′

Tν

is equal to the largest eigenvalue of T;

calculate a diffusion tensor for the first data voxel;

generate a diffusion vector (ν

) for the first data voxel, the diffusion vector being a first eigenvector corresponding to a largest eigenvalue of the calculated diffusion tensor for the first data voxel; and

process the fiber tract atlas and magnetic resonance data to determine a probability P that the first data voxel corresponds to the first fiber tract, based at least in part on the generated diffusion vector (ν

) and the averaged information on the probabilistic locations and the orientations of the first fiber tract in the first volume element and without manual inspection of the generated diffusion vector, the orientation information, or magnetic resonance images associated with the magnetic resonance data, wherein the determined probability includes P is determined by combining the atlas diffusion tensor (T), the first atlas diffusion tensor eigenvector (ν

), and the diffusion vector (ν

) of the first data voxel according to the expression

13. The MRI system of claim 12, the magnetic resonance data acquisition system comprising:

a magnet that generates a substantially static principal magnetic field;

a gradient sub-system that applies a plurality of gradient fields to the subject; and

a radio frequency sub-system that applies a plurality of radio frequency pulses to the subject and detects at least one echo signal from the subject.

14. The MRI system of claim 13, wherein the radio frequency sub-system includes a head coil that acquires magnetic resonance data from a human nervous system.

15. The MRI system of claim 12, wherein the data processor is further configured to:

process data corresponding to the determined probability and data corresponding to the fiber tract atlas and the acquired magnetic resonance data to generate a fiber probability map for at least a first portion of the nervous system of the subject; and

use the probability map to identify a region of interest of the nervous system of the subject.

16. The MRI system of claim 12, further comprising a data communication network communicatively coupling the magnetic resonance data acquisition system and a computing device that includes the data processor.

17. A computer program product comprising a non-transitory computer-readable medium having code stored thereon, the code, when executed by a data processing apparatus, causing the data processing apparatus to implement a method comprising:

obtaining a fiber tract atlas for a nervous system, the fiber tract atlas comprising;

a plurality of atlas voxels that each represent a different volume element of the nervous system, the plurality of atlas voxels including a first atlas voxel that represents a first volume element of the nervous system, and

averaged information on probabilistic locations and orientations of a first fiber tract in the first volume element of the nervous system, wherein the averaged information on the probabilistic locations and the orientations of the first fiber tract includes a rank-two symmetric atlas diffusion tensor (T) having a first atlas diffusion tensor eigenvector (ν

), such that ν

′

Tν

is equal to the largest eigenvalue of T;

acquiring magnetic resonance data from the nervous system of a subject, the magnetic resonance data comprising a plurality of data voxels including a first data voxel, the first data voxel relating to the first atlas voxel;

calculating a diffusion tensor for the first data voxel;

generating a diffusion vector (ν

) for the first data voxel, the diffusion vector being a first eigenvector corresponding to a largest eigenvalue of the calculated diffusion tensor for the first data voxel; and

determining a probability P that the first data voxel of the magnetic resonance data of the subject corresponds to the first fiber tract, based at least in part on the generated diffusion vector (ν

) and the averaged information on the probabilistic locations and the orientations of the first fiber tract in the first volume element, and without manual inspection of the generated diffusion vector, the orientation information, or magnetic resonance images associated with the magnetic resonance data, wherein the determined probability includes P is determined by combining the atlas diffusion tensor (T), the first atlas diffusion tensor eigenvector (ν

), and the diffusion vector (ν

) of the first data voxel according to the expression

18. The computer program product of claim 17, the operations further comprising:

producing location information of a fiber tract of interest associated with the first data voxel using fiber density data, wherein the fiber density data is associated with the plurality of atlas voxels and includes streamline fiber counts to indicate a likelihood of an atlas voxel to the fiber tract of interest;

resampling the averaged information on the probabilistic locations and the orientations of the first fiber tract from an atlas space to a subject space for the subject, wherein the determined probability that the first data voxel corresponds to the first fiber tract is based in part on resampled location probability and resampled orientation information.

19. The computer program product of claim 17, the operations further comprising:

using the determined probability with the fiber tract atlas and the acquired magnetic resonance data to generate a fiber probability map for at least a first portion of the nervous system of the subject; and

using the probability map to determine at least one of a fractional anisotropy for the first fiber tract or a diffusivity for the first fiber tract.

20. The computer program product of claim 17, the averaged information on the probabilistic locations and the orientations of the first fiber tract comprising the first atlas diffusion tensor, and the fiber tract atlas further comprising a plurality of additional tensors representing orientations of additional fiber tracts in the first volume element of the nervous system.

21. The computer program product of claim 20, the probability that the first data voxel corresponds to the first fiber tract comprising a first probability, the operations further comprising using the fiber tract atlas to find a plurality of additional probabilities based at least in part on the generated diffusion vector and the plurality of additional tensors, each additional probability representing a relative probability that the first data voxel corresponds to one of the additional fiber tracts.