Automated cancer detection using MRI

Patent metadata

Field	Value
Publication number	US-10215830-B2
Application number	US-201514971296-A
Country	US
Kind code	B2
Filing date	Dec 16, 2015
Priority date	Dec 16, 2015
Publication date	Feb 26, 2019
Grant date	Feb 26, 2019

Abstract

Methods and systems for diagnosing cancer in the prostate and other organs are disclosed. Exemplary methods comprises extracting texture information from MRI imaging data for a target organ, sometimes using two or more different imaging modalities. Texture features are determined that are indicative of cancer by identifying frequent texture patterns. A classification model is generated based on the determined texture features that are indicative of cancer, and diagnostic cancer prediction information for the target organ is then generated to help diagnose cancer in the organ.

First claim

The invention claimed is:

1. A method for facilitating cancer diagnosis, comprising:
extracting texture information from imaging data for a target organ;
determining texture features that are indicative of cancer by identifying frequent texture patterns in the extracted texture information by:
(1) identifying frequent texture patterns from the extracted texture information using frequent pattern mining by:
(a) identifying length-1 frequent patterns in the extracted texture information that are represented by a single texture pattern code that occurs more than a predetermined threshold percentage of the entire dataset,
(b) examining sub-datasets of the extracted texture information, wherein each sub-dataset contains a different one of the identified length-1 frequent patterns,
(c) generating frequent patterns with increasing pattern length, and
(d) examining additional sub-datasets of the extracted texture information, wherein each additional sub-dataset contains a different one of the generated frequent patterns having the increased pattern length;

(2) comparing occurrences of the frequent texture patterns between cancers and benign tissue using a Wilcoxon rank-sum test and selecting significant texture patterns based on the comparison; and
(3) ordering the significant texture patterns using minimum redundancy maximum relevance (mRMR) criterion, and then choosing the most discriminative texture features using forward feature selection;

generating a classification model based on the determined texture features that are indicative of cancer; and
based on the classification model, generating diagnostic cancer prediction information for the target organ.

2. The method of claim 1 , wherein the imaging data is derived from high-b-value diffusion weighted MRI.

3. The method of claim 1 , wherein the imaging data is derived from T2-weighted MRI.

4. The method of claim 1 , further comprising extracting texture information from first imaging data for a target organ and second imaging data for the target organ, the first and second imaging data being derived from different imaging modalities.

5. The method of claim 4 , wherein the first imaging data is derived from high-b-value diffusion weighted MRI and the second imaging data is derived from T2-weighted MRI.

6. The method of claim 4 , further comprising normalizing and registering the first imaging data and the second imaging data prior to applying the classification model.

7. The method of claim 1 , wherein the extraction of texture information is performed using local binary pattern (LBP), local direction derivative pattern (LDDP), and variance measure operator (VAR).

8. The method of claim 1 , wherein the classification model is generated using support vector machine (SVM).

9. The method of claim 1 , wherein generating diagnostic cancer prediction information comprises generating a cancer prediction map for the target organ.

10. A computing system for cancer diagnosis, the system comprising a processor and memory, the system operable to:
extract texture information from first imaging data for a target organ;
based on the extracted texture information, determine texture features that are indicative of cancer by:
identifying frequent texture patterns from the extracted texture information using frequent pattern mining by:
(a) identifying length-1 frequent patterns in the extracted texture information that are represented by a single texture pattern code that occurs more than a predetermined threshold percentage of the entire dataset,
(b) examining sub-datasets of the extracted texture information, wherein each sub-dataset contains a different one of the identified length-1 frequent patterns,
(c) generating frequent patterns with increasing pattern length, and
(d) examining additional sub-datasets of the extracted texture information, wherein each additional sub-dataset contains a different one of the generated frequent patterns having the increased pattern length;

comparing occurrences of the frequent texture patterns between cancers and benign tissue using a Wilcoxon rank-sum test and selecting significant texture patterns based on the comparison; and
ordering the significant texture patterns via minimum redundancy maximum relevance (mRMR) criterion;

generate a classification model based on the determined texture features that are indicative of cancer; and
based on the classification model, generate a diagnostic cancer prediction map of the target organ.

11. The system of claim 10 , wherein the first imaging data is derived from high-b-value diffusion weighted MRI.

12. The system of claim 10 , wherein the first imaging data is derived from T2-weighted MRI.

13. The system of claim 10 , wherein the system is operable to receive second imaging data for the target organ, the first and second imaging data being from different imaging modalities.

14. The system of claim 13 , wherein the first imaging data is derived from high-b-value diffusion weighted MRI and the second imaging data is derived from T2-weighted MRI.

15. The system of claim 13 , wherein the system is operable to normalize and register the first and second imaging data prior to extracting texture information.

16. The system of claim 10 , wherein the texture information is extracted using local binary pattern (LBP), local direction derivative pattern (LDDP), and variance measure operator (VAR).

17. The system of claim 10 , wherein the classification model is generated using support vector machine (SVM).

18. The system of claim 10 , wherein the target organ is a human prostate.

19. One or more non-transitory computer readable media storing computer-executable instructions, which when executed by a computer cause the computer to perform the method of claim 1 .

Assignees

• Us Health

Inventors

• Kwak Jin Tae
• Wood Bradford J
• Xu Sheng
• Turkbey Baris
• Choyke Peter L
• Pinto Peter A
• Summers Ronald M

Classifications

• A61B5/055

  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging · CPC title

• G01R33/5602

  by filtering or weighting based on different relaxation times within the sample, e.g. T1 weighting using an inversion pulse · CPC title

• G06T7/41

  based on statistical description of texture · CPC title

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G01R33/56341 Primary

  Diffusion imaging · CPC title