Disease Characterization From Fused Pathology And Radiology Data

Patent metadata

Field	Value
Publication number	US-2016196648-A1
Application number	US-201514964665-A
Country	US
Kind code	A1
Filing date	Dec 10, 2015
Priority date	Jan 5, 2015
Publication date	Jul 7, 2016
Grant date	—

Abstract

Methods and apparatus associated with distinguishing invasive adenocarcinoma (IA) from in situ adenocarcinoma (AIS) are described. One example apparatus includes a set of logics, and a data store that stores three dimensional (3D) radiological images of tissue demonstrating IA or AIS. The set of logics includes a classification logic that generates an invasiveness classification for a diagnostic 3D radiological image, a training logic that trains the classification logic to identify a texture feature associated with IA, an image acquisition logic that acquires a diagnostic 3D radiological image of a region of tissue demonstrating cancerous pathology and that provides the diagnostic 3D radiological image to the classification logic, and a prediction logic that generates an invasiveness score based on the diagnostic 3D radiological image and the invasiveness classification. The training logic trains the classification logic using a set of 3D histological reconstructions combined with the set of 3D radiological images.

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 distinguishing invasive adenocarcinoma (IA) from non-invasive adenocarcinoma (AIS), the method comprising:
accessing a computed tomography (CT) image of a region of tissue demonstrating cancerous pathology; extracting a set of texture features from the CT image; providing the set of texture features to an automated IA classifier; receiving, from the IA classifier, a classification of the set of texture features; and classifying the region of tissue as IA or AIS, based, at least in part, on the set of texture features and the classification.

2 . The non-transitory computer-readable storage device of claim 1 , where the IA classifier is trained on a set of composite images, where the composite images are formed from a set of three dimensional (3D) histology reconstructions of a region of tissue demonstrating IA or AIS combined with a set of 3D CT images of the region of tissue demonstrating IA or IAS.

3 . The non-transitory computer-readable storage device of claim 2 , the method comprising:
combining the set of 3D histology reconstructions with the set of 3D CT images by:
accessing a set of histology slices of a region of tissue demonstrating cancerous pathology, where a member of the set of histology slices includes an invasive component and a non-invasive component;
generating a 3D histology reconstruction of the set of histology slices;
accessing a set of 3D radiological images of the region of tissue, where a member of the set of 3D radiological images includes a texture feature or a shape feature;
identifying an optimal translation of the set of 3D radiological images relative to the 3D histology reconstruction;
identifying an optimal rotation of the set of 3D radiological images relative to the 3D histology reconstruction;
registering the 3D histology reconstruction with the set of 3D radiological images based, at least in part, on the optimal translation or the optimal rotation;
mapping the extent of cancerous invasion from the 3D histology reconstruction onto the set of 3D radiological images; and
identifying, based, at least in part on the mapping, a texture feature or a shape feature associated with an invasive component.

4 . The non-transitory computer-readable storage device of claim 3 , where generating the 3D histology reconstruction includes performing a group-wise registration of at least two members of the set of histology slices.

5 . The non-transitory computer-readable storage device of claim 3 , where the optimal translation is an affine translation or a deformable translation.

6 . The non-transitory computer-readable storage device of claim 3 , where mapping the extent of cancerous invasion from the 3D histology reconstruction onto the set of 3D radiological images includes using an inverse of the optimized affine translation or an inverse of the optimized deformable translation.

7 . The non-transitory computer-readable storage device of claim 1 , where the region of tissue includes a ground glass nodule (GGN).

8 . The non-transitory computer-readable storage device of claim 1 , where the set of 3D CT images is acquired with an image size of 512 by 512 pixels.

9 . The non-transitory computer-readable storage device of claim 8 , where a member of the set of 3D CT images has an in-plane resolution of 0.57 mm to 0.87 mm.

10 . The non-transitory computer-readable storage device of claim 9 , where a distance between a first member of the set of 3D CT images and a second, different member of the set of 3D CT images is 1 mm to 5 mm.

11 . The non-transitory computer-readable storage device of claim 3 , where the texture feature associated with the invasive component is a first order statistic.

12 . The non-transitory computer-readable storage device of claim 11 , where the first order statistic is computed from a mean of CT intensity of a sliding 3×3×3 voxel window obtained from a member of the set of 3D CT images.

13 . The non-transitory computer-readable storage device of claim 3 , the method including refining the 3D histology reconstruction using a local search constrained by the set of 3D radiological images.

14 . A method for distinguishing invasive cancerous pathology from non-invasive pathology, the method comprising:
accessing a set of histology slices of a region of tissue demonstrating invasive cancerous pathology and non-invasive cancerous pathology; generating a three dimensional (3D) histology reconstruction of the set of histology slices; accessing a set of 3D radiological images of the region of tissue, where a member of the set of 3D radiological images includes a texture feature or a shape feature; registering the 3D histology reconstruction with the set of 3D radiological images; generating a mapping of invasive cancerous pathology from the 3D histology reconstruction onto the set of 3D radiological images based, at least in part, on the registration; identifying, in a member of the set of 3D radiological images, a texture feature or a shape feature associated with invasive cancerous pathology based, at least in part, on the mapping; and training a classifier to distinguish invasive cancerous pathology from non-invasive cancerous pathology based, at least in part, on the mapping.

15 . The method of claim 14 , comprising:
accessing a diagnostic 3D radiological image of a region of tissue demonstrating cancerous pathology; extracting a texture feature or a shape feature from the diagnostic 3D radiological image; providing the extracted texture feature or the extracted shape feature to the classifier; receiving, from the classifier, a first classification; and distinguishing invasive cancerous pathology represented in the diagnostic 3D radiological image from non-invasive cancerous pathology represented in the diagnostic 3D image based, at least in part, on the first classification and the texture feature or the shape feature.

16 . An apparatus, comprising:
a processor; a memory; a data store that stores a set of three dimensional (3D) radiological images of tissue demonstrating invasive adenocarcinoma (IA) or non-invasive adenocarcinoma (AIS), where a member of the set of 3D radiological images includes a texture feature or a shape feature; an input/output interface; a set of logics; and an interface to connect the processor, the memory, the data store, the input/output interface and the set of logics, where the set of logics includes:
a classification logic that generates an invasiveness classification for a diagnostic 3D radiological image;
a training logic that trains the classification logic to identify a texture feature or a shape feature associated with IA or AIS;
an image acquisition logic that accesses a diagnostic 3D radiological image of a region of tissue demonstrating cancerous pathology and that provides the diagnostic 3D radiological image to the classification logic, where the diagnostic 3D radiological image includes a texture feature or a shape feature; and
an IA prediction logic that generates an invasiveness score based, at least in part, on the diagnostic 3D radiological image and the invasiveness classification.

17 . The apparatus of claim 16 , where the classification logic generates the invasiveness classification by distinguishing IA from AIS represented in the diagnostic 3D radiological image based, at least in part, on a texture feature extracted from t

Assignees

• Univ Case Western Reserve

Inventors

• Madabhushi Anant
• Rusu Mirabela

Classifications

• A61P37/00

  Drugs for immunological or allergic disorders · CPC title

• A61P37/06

  Immunosuppressants, e.g. drugs for graft rejection · CPC title

• C07K2317/71

  Decreased effector function due to an Fc-modification · CPC title

• C07K2317/41

  Glycosylation, sialylation, or fucosylation · CPC title

• C07K2317/92

  Affinity (KD), association rate (Ka), dissociation rate (Kd) or EC50 value · CPC title