Leveraging on local and global textures of brain tissues for robust automatic brain tumor detection

Patent metadata

Field	Value
Publication number	US-10055839-B2
Application number	US-201615061465-A
Country	US
Kind code	B2
Filing date	Mar 4, 2016
Priority date	Mar 4, 2016
Publication date	Aug 21, 2018
Grant date	Aug 21, 2018

Abstract

A method for performing cellular classification includes generating a plurality of local dense Scale Invariant Feature Transform (SIFT) features based on a set of input images and converting the plurality of local dense SIFT features into a multi-dimensional code using a feature coding process. A first classification component is used to generate first output confidence values based on the multi-dimensional code and a plurality of global Local Binary Pattern Histogram (LBP-H) features are generated based on the set of input images. A second classification component is used to generate second output confidence values based on the plurality of LBP-H features and the first output confidence values and the second output confidence values are merged. Each of the set of input images may then be classified as one of a plurality of cell types using the merged output confidence values.

First claim

The invention claimed is:

1. A method for performing cellular classification, the method comprising:
generating a plurality of local dense Scale Invariant Feature Transform (SIFT) features based on a set of input images;
converting the plurality of local dense SIFT features into a multi-dimensional code using a feature coding process;
using a first classification component to generate first output confidence values based on the multi-dimensional code;
generating a plurality of global Local Binary Pattern Histogram (LBP-H) features based on the set of input images;
using a second classification component to generate second output confidence values based on the plurality of LBP-H features;
merging the first output confidence values and the second output confidence values using a fusion algorithm to yield merged output confidence values; and
classifying each of the set of input images as one of a plurality of cell types using the merged output confidence values, wherein the fusion algorithm is a multiplicative fusion algorithm.

2. The method of claim 1 , further comprising:
acquiring a plurality of input images;
calculating an entropy value for each of the plurality of input images, each entropy value representative of an amount of texture information in a respective image;
identifying one or more low-entropy images in the set of input images, wherein the one or more low-entropy images are each associated with a respective entropy value below a threshold value; and
generating the set of input images based on the plurality of input images, wherein the set of input images excludes the one or more low-entropy images.

3. The method of claim 2 , wherein the plurality of input images are acquired using an endomicroscopy device during a medical procedure.

4. The method of claim 2 , wherein the plurality of input images are acquired using a digital holographic microscopy device during a medical procedure.

5. The method of claim 1 , wherein the feature coding process performs a bag of words coding process.

6. The method of claim 5 , wherein the bag of words coding process comprises:
randomly selecting a predetermined percentage of descriptors from a training split;
performing k-means clustering to construct a plurality of different vocabularies;
using the plurality of different vocabularies to quantize the SIFT features in each input image included in the set of input images to yield a quantized representation of the SIFT features,
wherein the multi-dimensional code is the quantized representation.

7. The method of claim 1 , wherein the first classification component is a Support Vector Machine (SVM) classifier.

8. The method of claim 7 , wherein the multi-dimensional code is used to train the SVM classifier with a radial basis function (RBF) kernel.

9. The method of claim 7 , wherein the multi-dimensional code is used to train the SVM classifier with a linear kernel.

10. The method of claim 1 , wherein the second classification component is a Random Forest classifier with a plurality of trees and a maximum depth of a plurality of levels for each of the plurality of trees.

11. A method for performing cellular classification during a medical procedure, the method comprising:
prior to the medical procedure, performing a training process comprising:
training a first classification component based on a plurality of local dense SIFT features associated with a plurality of training images, and
training a second classification component based on a plurality of global LBP-H features associated with the plurality of training images; and

during the medical procedure, performing a cell classification process comprising:
acquiring an input image using an endomicroscopy device,
using the first classification component to generate first output confidence values based on SIFT features associated with the input image,
using the second classification component to generate second output confidence values based on LBP-H features associated with the input image,
multiplying the first output confidence values and the second output confidence values to yield merged output confidence values; and
identifying a class label corresponding to the input image based on the merged output confidence values, and
presenting the class label on a display operably coupled to the endomicroscopy device.

12. The method of claim 11 , wherein the class label provides an indication of whether biological material in the input image is healthy, malignant, or benign.

13. The method of claim 11 , wherein the first classification component is an SVM classifier.

14. The method of claim 13 , wherein the second classification component is a Random Forrest classifier.

15. A system performing cellular classification, the system comprising:
an endomicroscopy device including a probe configured to acquire a set of input images during a medical procedure;
an imaging computer configured to perform a cellular classification process during the medical procedure, the cellular classification process comprising:
using a first classification component to generate first output confidence values based on SIFT features associated with the set of input images,
using a second classification component to generate second output confidence values based on LBP-H features associated with the set of input images,
multiplying the first output confidence values and the second output confidence values to yield merged output confidence values; and
identifying one or more cellular class labels corresponding to the set of input images based on the merged output confidence values, and

a display configured to present the one or more cellular class labels during the medical procedure.

16. The system of claim 15 , wherein the endomicroscopy device is a Confocal Laser Endo-microscopy device.

17. The system of claim 15 , wherein the endomicroscopy device is a Digital Holographic Microscopy device.

18. The system of claim 15 , wherein the first classification component is a SVM classifier.

19. The system of claim 18 , wherein the second classification component is a Random Forrest classifier.

Assignees

• Siemens Ag

Inventors

• Bhattacharya Subhabrata
• Sun Shanhui
• Chen Terrence
• Kamen Ali

Classifications

• G06V10/809

  of classification results, e.g. where the classifiers operate on the same input data · CPC title

• G06V20/698 Primary

  Matching; Classification · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06F18/254

  of classification results, e.g. of results related to same input data · CPC title

• G06F18/2411

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