Automatic segmentation and quantitative parameterization of brain tumors in MRI

Patent metadata

Field	Value
Publication number	US-10169685-B2
Application number	US-201515323530-A
Country	US
Kind code	B2
Filing date	Jul 7, 2015
Priority date	Jul 7, 2014
Publication date	Jan 1, 2019
Grant date	Jan 1, 2019

Abstract

Techniques, systems, and devices are described for implementing automatic segmentation and quantitative parameterization of MRI images. For example, the disclosed method includes processing the MRI image to correct any distortions; performing a preliminary segmentation of the MRI image to assign a tissue label of a set of tissue labels to one or more preliminary volumes of voxels of the MRI image; comparing each voxel of the MRI image with the one or more preliminary volumes of voxels with an assigned tissue label and assigning each voxel of the MRI image a probability of being associated with each tissue label of the set of tissue labels; and assigning each voxel of the MRI image a tissue label according to its greatest probability among probabilities for each voxel being associated with the set of tissue labels.

First claim

What is claimed are techniques and structures as described and shown, including:

1. A method for processing a magnetic resonance imaging (MRI) image, comprising:
preprocessing the MRI image to correct distortions;
performing, after the preprocessing, a preliminary segmentation of the MRI image to assign a tissue label of a set of tissue labels to one or more preliminary volumes of voxels of the MRI image;
comparing each voxel of the MRI image with the one or more preliminary volumes of voxels with an assigned tissue label and assigning each voxel of the MRI image a probability of being associated with each tissue label of the set of tissue labels; and
assigning each voxel of the MRI image a tissue label according to its greatest probability among probabilities for each voxel being associated with the set of tissue labels,
wherein the comparing and assigning is performed using a classifier trained on a subject-by-subject basis that segments each patient according to his or her own subject-specific classifier by using both intensity and spatial data from each voxel, and
wherein a weighted random sampling of the one or more preliminary volumes of voxels of the MRI image is used to train the classifier such that weights for sampling reflect a relative distribution of voxels assigned to tissue labels from preliminary segmentation.

2. The method of claim 1 , wherein the comparing and assigning comprises:
training the classifier using the voxels of the one or more preliminary volumes with the assigned tissue label; and
assigning by the classifier each voxel of the MRI image a probability of membership for each tissue label.

3. The method of claim 1 , further comprising:
filtering the MRI image to remove any clusters of voxels that have fewer than a certain number of continuous voxels.

4. The method of claim 1 , wherein the set of tissue labels includes: contrast enhancing (CE), fluid-attenuated inversion recovery hyperintensity (FH), gray matter (GM), white matter (WM), cerebral spinal fluid (CSF), and blood vessel (BV).

5. The method of claim 1 , wherein the preprocessing the MRI image to correct any distortions includes skull stripping and registration to Montreal Neurological Institute (MNI) templates.

6. The method of claim 2 , wherein the classifier comprises a k-nearest neighbor (KNN) probabilistic classifier and a Gaussian mixture model (GMM) probabilistic classifier.

7. The method of claim 1 , wherein the one or more preliminary volumes are derived by trained human operators and wherein the comparing includes:
assessing a similarity between volumes by dividing a sum of intersection between a voxel being compared and the one or more preliminary volumes with values of volumes of the voxel being compared and the one or more preliminary volume.

8. A system for processing a magnetic resonance imaging (MRI) image, including:
an MRI machine that obtains an MRI image of a subject; and
a processor that includes multiple components for processing the MRI image, including:
a component that corrects distortions in the MRI image;
a component that performs a preliminary segmentation of the MRI image to assign a tissue label of a set of tissue labels to one or more preliminary volumes of voxels of the MRI image;
a component that compares each voxel of the MRI image with the one or more preliminary volumes of voxels with an assigned tissue label and assigning each voxel of the MRI image a probability of being associated with each tissue label of the set of tissue labels; and
a component that assigns each voxel of the MRI image a tissue label according to its greatest probability among probabilities for each voxel being associated with the set of tissue labels,
wherein the comparing and assigning is performed using a classifier trained on a subject-by-subject basis that segments each patient according to his or her own subject-specific classifier by using both intensity and spatial data from each voxel, and
wherein a weighted random sampling of the one or more preliminary volumes of voxels of the MRI image is used to train the classifier such that weights for sampling reflect a relative distribution of voxels assigned to tissue labels from preliminary segmentation.

9. The system of claim 8 , wherein the component that compares includes:
a component that trains the classifier using the voxels of the one or more preliminary volumes with the assigned tissue label; and
a component that assigns by the classifier each voxel of the MRI image a probability of membership for each tissue label.

10. The system of claim 8 , wherein the processor further includes:
a component that filters the MRI image to remove any clusters of voxels that have fewer than a certain number of continuous voxels.

11. The system of claim 8 , wherein the set of tissue labels includes:
contrast enhancing (CE), fluid-attenuated inversion recovery hyperintensity (FH), gray matter (GM), white matter (WM), cerebral spinal fluid (CSF), and blood vessel (BV).

12. The system of claim 8 , wherein the component that processes the MRI image to correct any distortions includes a component for skull stripping and registration to Montreal Neurological Institute (MNI) templates.

13. The system of claim 9 , wherein the classifier comprises a k-nearest neighbor (KNN) probabilistic classifier and a Gaussian mixture model (GMM) probabilistic classifier.

14. The system of claim 8 , wherein the one or more preliminary volumes are derived by trained human operators and wherein the component that compares includes:
a component that assesses a similarity between volumes by dividing a sum of intersection between a voxel being compared and the one or more preliminary volumes with values of volumes of the voxel being compared and the one or more preliminary volume.

15. A computer program product comprising a non-transitory computer readable medium having code stored thereupon, the code, when executed by a processor, causing the processor to implement a method for processing a magnetic resonance imaging (MRI) image, comprising:
code for preprocessing the MRI image to correct distortions;
code for performing, after the preprocessing, a preliminary segmentation of the MRI image to assign a tissue label of a set of tissue labels to one or more preliminary volumes of voxels of the MRI image;
code for comparing each voxel of the MRI image with the one or more preliminary volumes of voxels with an assigned tissue label and assigning each voxel of the MRI image a probability of being associated with each tissue label of the set of tissue labels; and
code for assigning each voxel of the MRI image a tissue label according to its greatest probability among probabilities for each voxel being associated with the set of tissue labels,
wherein the comparing and assigning is performed using a classifier trained on a subject-by-subject basis that segments each patient according to his or her own subject-specific classifier by using both intensity and spatial data from each voxel, and
wherein a weighted random sampling of the one or more preliminary volumes of voxels of the MRI image is used to train the classifier such that weights for sampling reflect a relative distribution of voxels assigned to tissue labels from preliminary segmentation.

16. The computer program product of claim 15 , wherein the code for comparing and assigning comprises:
code for training the classifier using the voxels of the one or more preliminary volumes with the assigned tissue label; and
code for assigning by the classifier each voxel of the MRI image a probability of membe

Assignees

• Univ California

Inventors

• Chen Clark
• Steed Tyler
• Carter Bob S
• Dale Anders M

Classifications

• G01R33/5608 Primary

  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 title

• G06V10/7788

  the supervisor being a human, e.g. interactive learning with a human teacher · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06F18/24147

  Distances to closest patterns, e.g. nearest neighbour classification · CPC title

• G06F18/2415

  based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate · CPC title