Entropy-based radiogenomic descriptions on magnetic resonance imaging (MRI) for molecular characterization of breast cancer

Patent metadata

Field	Value
Publication number	US-10055842-B2
Application number	US-201715397266-A
Country	US
Kind code	B2
Filing date	Jan 3, 2017
Priority date	Jul 29, 2016
Publication date	Aug 21, 2018
Grant date	Aug 21, 2018

Abstract

Methods, apparatus, and other embodiments distinguish disease phenotypes and mutational status using co-occurrence of local anisotropic gradient orientations (CoLIAGe) and Laws features. One example apparatus includes a set of circuits that acquires a radiologic image (e.g., MRI image) of a region of tissue demonstrating breast cancer, computes a gradient orientation for a pixel in the MRI image, computes a significant orientation for the pixel based on the gradient orientation, constructs a feature vector that captures a discretized entropy distribution for the image based on the significant orientation, extracts a set of texture features from the MRI image, and classifies the phenotype of the breast cancer based on the feature vector and the set of texture features. Embodiments of example apparatus may generate and display a heatmap of entropy values for the image. Example methods and apparatus may operate substantially in real-time, or may operate in two, three, or more dimensions.

First claim

What is claimed is:

1. A non-transitory computer-readable storage device storing computer-executable instructions that when executed by a computer controls the computer to perform a method for distinguishing breast cancer phenotypes using co-occurrence of local anisotropic gradient orientations (CoLIAGe) and Laws energy features, the method comprising:
accessing a region of interest (ROI) in a volume illustrated in a magnetic resonance image (MRI), the ROI having a set of pixels, where a pixel in the set of pixels has an intensity;
obtaining an x-axis gradient for a first pixel in the set of pixels based, at least in part, on the intensity of the pixel;
obtaining a y-axis gradient for the first pixel based, at least in part, on the intensity of the pixel;
computing an x-axis gradient vector for a second pixel in an N pixel by N pixel neighborhood centered around the first pixel, N being a number;
computing a y-axis gradient vector for the second pixel in the N pixel by N pixel neighborhood;
constructing a localized gradient vector matrix based, at least in part, on the x-axis gradient vector for the second pixel, and the y-axis gradient vector for the second pixel;
computing a dominant orientation for the first pixel based, at least in part, on the localized gradient vector matrix;
constructing a co-occurrence matrix from the dominant orientation;
computing an entropy for the first pixel based, at least in part, on the co-occurrence matrix;
obtaining a distribution of the entropy;
constructing a feature vector based, at least in part, on the distribution of the entropy;
extracting a set of texture features from the ROI; and
controlling a phenotype classifier to classify the ROI based, at least in part, on the feature vector and the set of texture features.

2. The non-transitory computer-readable storage device of claim 1 , where the volume illustrated in the MRI is associated with a dynamic contrast enhanced (DCE) MRI image of a patient demonstrating breast cancer pathology.

3. The non-transitory computer-readable storage device of claim 2 , where the set of texture features includes a Laws feature, where the Laws feature is based on a two dimensional (2D) filter targeting speckling, rippling, waves, spottiness, or edges represented in the volume illustrated in the MRI.

4. The non-transitory computer-readable storage device of claim 3 , where controlling the phenotype classifier to classify the ROI includes distinguishing human epidermal growth factor receptor enriched (HER2-E) breast cancer subtype from human epidermal growth factor receptor positive (HER2+) breast cancer in the ROI.

5. The non-transitory computer-readable storage device of claim 4 , where controlling the phenotype classifier to classify the ROI includes distinguishing HER2-E from human epidermal growth factor receptor luminal (HER2-L) breast cancer subtype or human epidermal growth factor receptor basal (HER2-B) breast cancer subtype, or distinguishing HER2-B from HER2-E or HER2-L.

6. The non-transitory computer-readable storage device of claim 5 , where the phenotype classifier distinguishes HER2-E from HER2-L or HER2-B based, at least in part, on the feature vector, the entropy, or the distribution of the entropy.

7. The non-transitory computer-readable storage device of claim 5 , where the phenotype classifier distinguishes HER2-B from HER2-E or HER2-L based, at least in part, on a kurtosis value or a skewness value of pixel-wise entropy or the Laws feature.

8. The non-transitory computer-readable storage device of claim 3 , where controlling the phenotype classifier to classify the ROI includes predicting a TP53 mutational status of breast cancer tissue represented in the ROI.

9. The non-transitory computer-readable storage device of claim 8 , where predicting a TP53 mutational status comprises:
computing a skewness measure of correlation and a kurtosis measure of correlation based on the co-occurrence matrix; and
classifying the ROI as TP53 MUT or TP53 WT based, at least in part, on the feature vector, the skewness measure, and the kurtosis measure.

10. The non-transitory computer-readable storage device of claim 9 , where the skewness measure includes an inertia feature, a sum average feature, or a difference feature variance, and where the kurtosis measure includes an energy feature, an entropy feature, or a correlation feature.

11. A non-transitory computer-readable storage device storing computer-executable instructions that when executed by a computer controls the computer to perform a method for distinguishing disease phenotypes, the method comprising:
accessing a region of interest (ROI) in a volume illustrated in a radiologic image, where the ROI has a set of pixels, and where a pixel in the set of pixels has an intensity;
computing a local dominant gradient orientation for a first pixel in the set of pixels;
constructing a co-occurrence matrix for the set of pixels, based, at least in part, on the local dominant orientation;
computing an entropy measure for the set of pixels based, at least in part, on the co-occurrence matrix;
constructing a feature vector based on a distribution of the entropy measure;
extracting a set of texture features from the ROI; and
controlling a disease phenotype classification system to classify the ROI as human epidermal growth factor receptor enriched (HER2-E) breast cancer subtype or human epidermal growth factor receptor positive (HER2+) breast cancer based, at least in part, on the feature vector and the set of texture features.

12. The non-transitory computer readable storage device of claim 11 , where computing the local dominant gradient orientation for the first pixel includes:
obtaining a set of gradients for the first pixel along a plurality of axes;
computing a gradient orientation for the pixel based, at least in part, on the set of gradients;
computing a set of gradient vectors for a plurality of pixels in a neighborhood, where the neighborhood is centered on the first pixel;
constructing a localized gradient vector matrix from the set of gradient vectors; and
computing the local dominant gradient orientation for the first pixel using principal component analysis (PCA) based, at least in part, on the localized gradient vector matrix.

13. The non-transitory computer readable storage device of claim 12 , where constructing the co-occurrence matrix includes:
discretizing the local dominant gradient orientation for the first pixel; and
populating the co-occurrence matrix with local dominant gradient orientation pairs that co-exist between pixels in the neighborhood.

14. The non-transitory computer readable storage device of claim 13 , where computing the entropy measure includes:
aggregating the entropy measure for a subset of the set of pixels, where the entropy measure is based, at least in part, on the co-occurrence matrix; and
constructing a histogram of the entropy measure for set of pixels, where the histogram is divided into bins, where a bin has a threshold size.

15. The non-transitory computer readable storage device of claim 14 , where constructing the feature vector includes extracting entropy measure values from a distribution of the histogram.

16. The non-transitory computer readable storage device of claim 15 , where controlling the disease phenotype classification system includes constructing a heatmap based on the feature vector and the set of texture features, where hot areas of the heatmap represent high entropy values, and where cool areas of the heatmap represent low entropy values.

17. An apparatus

Assignees

• Univ Case Western Reserve

Inventors

• Prasanna Prateek
• Braman Nathaniel
• Madabhushi Anant
• Varadan Vinay
• Harris Lyndsay
• Singh Salendra

Classifications

• A61B5/055

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

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G06T7/11

  Region-based segmentation · CPC title

• G06T2207/30068

  Mammography; Breast · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title