Decision support for disease characterization and treatment response with disease and peri-disease radiomics

Patent metadata

Field	Value
Publication number	US-10004471-B2
Application number	US-201615226124-A
Country	US
Kind code	B2
Filing date	Aug 2, 2016
Priority date	Aug 6, 2015
Publication date	Jun 26, 2018
Grant date	Jun 26, 2018

Abstract

Methods, apparatus, and other embodiments associated with classifying a region of tissue using textural analysis are described. One example apparatus includes an image acquisition logic that acquires an image of a region of tissue demonstrating cancerous pathology, a delineation logic that distinguishes nodule tissue within the image from the background of the image, a perinodular zone logic that defines a perinodular zone based on the nodule, a feature extraction logic that extracts a set of features from the image, a probability logic that computes a probability that the nodule is benign or that the nodule will respond to a treatment, and a classification logic that classifies the nodule tissue based, at least in part, on the set of features or the probability. A prognosis or treatment plan may be provided based on the classification of the image.

First claim

What is claimed is:

1. A non-transitory computer-readable storage device storing computer executable instructions that when executed by a computer control the computer to perform a method for characterizing a nodule in a region of tissue, the method comprising:
accessing an image of a region of tissue demonstrating cancerous pathology, where accessing the image of the region of tissue comprises accessing a computed tomography (CT) image of a region of lung tissue, where the CT image is a no-contrast chest CT image;
segmenting a nodule in the image by extracting a nodule boundary from the image by automatically distinguishing nodule tissue in the image from the background of the image using a spectral embedding gradient vector flow active contour (SEGvAC) approach, the SEGvAC approach comprising:
generating an initial lung mask by separating a lung region represented in the image from surrounding lung anatomy;
generating a refined image by removing a non-granuloma structure from the initial lung mask using a rule-based classifier, where the rule-based classifier selects the non-granuloma structure to remove from the initial lung mask based on a convexity measure of the non-granuloma structure, or an elongation measure of the non-granuloma structure;
generating a spectral embedding (SE) representation by projecting at least one refined image into a three dimensional (3D) SE space; and
extracting a nodule boundary from the SE representation using a gradient vector flow field (GVF) active contour model, where a GVF force drives the active contour;

defining a perinodular region in the image;
generating a set of perinodular texture features;
computing a probability that the nodule is benign based, at least in part, on the set of perinodular texture features; and
controlling a computer aided diagnosis (CADx) system to generate a classification of the nodule based, at least in part, on the set of perinodular texture features, or the probability that the nodule is benign.

2. The non-transitory computer-readable storage device of claim 1 , where generating the initial lung mask further comprises refining the initial lung mask by applying morphological hole-filling to a logical complement of the initial lung mask.

3. The non-transitory computer-readable storage device of claim 1 , where extracting the nodule boundary from the SE representation comprises calculating a tensor gradient function derived from the SE representation.

4. The non-transitory computer-readable storage device of claim 1 , where a GVF force is calculated based on a generalized diffusion equation applied to a component of an image edge map of the CT image of the region of lung tissue.

5. The non-transitory computer-readable storage device of claim 1 , where defining the perinodular region comprises:
generating an outer perinodular boundary by dilating the nodule boundary a threshold amount;
generating a perinodular region by subtracting the nodule from the region defined by the outer perinodular boundary.

6. The non-transitory computer-readable storage device of claim 1 , where generating the set of perinodular texture features comprises:
extracting a first set of texture features from the perinodular region; and
reducing the first set of texture features using principal component analysis (PCA) variable importance on projections (VIP) feature ranking.

7. The non-transitory computer-readable storage device of claim 1 , the method further comprising:
extracting a set of nodule features from the nodule, where the set of nodule features comprises a set of shape features or a second set of texture features;
generating a reduced set of nodule features by reducing the set of nodule features using PCA-VIP feature ranking.

8. The non-transitory computer-readable storage device of claim 7 , where the set of perinodular texture features or the second set of texture features includes a gray-level statistical feature, a steerable Gabor feature, a Haralick feature, a Law feature, a Law-Laplacian feature, a local binary pattern (LBP) feature, an inertia feature, a correlation feature, a difference entropy feature, a contrast inverse moment feature, a gradient feature, a co-occurrence of local anisotropic gradient orientations (CoLIAGe) feature, or a contrast variance feature.

9. The non-transitory computer-readable storage device of claim 7 , where the set of shape features includes a size feature, an area feature, a perimeter feature, an eccentricity feature, an extend feature, a compactness feature, a radial distance feature, a roughness feature, an elongation feature, a convexity feature, an equivalent diameter feature, or a sphericity feature.

10. The non-transitory computer-readable storage device of claim 7 , where the CADx system generates the classification of the image of the nodule using a support vector machine (SVM) classifier, where the SVM classifier classifies the image of the nodule with an accuracy of at least 0.77 area under the curve (AUC).

11. The non-transitory computer-readable storage device of claim 10 , where the SVM classifier is trained using the set of perinodular texture features or the reduced set of nodule features with a radial-based kernel (RBF) function.

12. The non-transitory computer-readable storage device of claim 7 , where generating the classification of the nodule is based, at least in part, on the set of perinodular texture features, the reduced set of nodule features, or the probability that the nodule is benign.

13. The non-transitory computer-readable storage device of claim 1 , where accessing the image of the region of tissue comprises accessing a dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) image of a region of breast tissue.

14. The non-transitory computer-readable storage device of claim 13 , where defining the perinodular region comprises:
segmenting a breast wall from the image;
removing the breast wall from the image;
generating an outer perinodular boundary by dilating the nodule boundary a threshold amount;
generating a perinodular region by subtracting the nodule from the region defined by the outer perinodular boundary.

15. The non-transitory computer-readable storage device of claim 14 , the method further comprising computing a probability that the nodule will achieve pathological complete response (pCR) based, at least in part, on the set of perinodular texture features; and
controlling the CADx system to generate a classification of the nodule based, at least in part, on the set of perinodular texture features, or the probability that the nodule will achieve pCR.

16. An apparatus for classifying a region of tissue in an image, comprising:
a processor;
a memory;
an input/output interface;
a set of circuits; and
an interface to connect the processor, the memory, the input/output interface and the set of circuits, where the set of circuits includes:
an image acquisition circuit configured to acquire an image of a region of tissue demonstrating cancerous pathology;
a delineation circuit configured to distinguish nodule tissue in the image from the background of the image by defining a nodule boundary;
a perinodular zone circuit configured to define a perinodular zone based, at least in part, on the nodule tissue or the nodule boundary;
a feature extraction circuit configured to extract a set of features from the image, where the set of features comprises a first set of texture features extracted from the nodule tissue in the image or a set of shape features extracted from the nodule tissue in the image, and a second set of texture features ex

Assignees

• Univ Case Western Reserve

Inventors

• Madabhushi Anant
• Orooji Mahdi
• Rusu Mirabela
• Linden Philip
• Gilkeson Robert
• Braman Nathaniel Mason

Classifications

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title

• G06T12/30

  Image post-processing, e.g. metal artefact correction · CPC title

• G06T12/00

  Tomographic reconstruction from projections · CPC title

• G06F18/2411

  based on the proximity to a decision surface, e.g. support vector machines · CPC title

• G06F18/24

  Classification techniques · CPC title