System and method for unsupervised detection and gleason grading of prostate cancer whole mounts using NIR fluorscence

• US 8,139,831 B2
• Filed: 12/02/2008
• Issued: 03/20/2012
• Est. Priority Date: 12/06/2007
• Status: Active Grant

• Current Assignee: BETH Israel Deaconess Medical Center (Caregroup Incorporated), Siemens Healthineers AG (Siemens AG)
• Original Assignee: Siemens AG
• Inventors: Khamene, Ali, Ehtiati, Tina, Frangioni, John V., Kianzad, Vida, Azar, Fred S.
• Primary Examiner(s): Lee, John
• Application Number:
  US12/326,135
• Publication Number:
  US 20090161928A1
• Time in Patent Office: 1,204 Days
• Field of Search: 382/128, 382/133
• US Class Current: 382/128
• CPC Class Codes: G06F 18/2415 based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/10064 Fluorescence image
• CPC Class Codes: G06T 2207/20076 Probabilistic image processing
• CPC Class Codes: G06T 2207/20152 Watershed segmentation
• CPC Class Codes: G06T 2207/30024 Cell structures in vitro; Tissue sections in vitro
• CPC Class Codes: G06T 2207/30081 Prostate
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/66 of image moments or centre of gravity
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 20/698 Matching; Classification

Abstract

A method for unsupervised classification of histological images of prostatic tissue includes providing histological image data obtained from a slide simultaneously co-stained with NIR fluorescent and Hematoxylin-and-Eosin (H&E) stains, segmenting prostate gland units in the image data, forming feature vectors by computing discriminating attributes of the segmented gland units, and using the feature vectors to train a multi-class classifier, where the classifier classifies prostatic tissue into benign, prostatic intraepithelial neoplasia (PIN), and Gleason scale adenocarcinoma grades 1 to 5 categories.

Claims

1. A method for unsupervised classification of histological images of prostatic tissue, comprising the steps of:

providing histological image data obtained from a slide simultaneously co-stained with NIR fluorescent and Hematoxylin-and-Eosin (H&

E) stains;

segmenting prostate gland units in the image data;

forming feature vectors by computing discriminating attributes of the segmented gland units; and

using said feature vectors to train a multi-class classifier within a Bayesian framework, wherein said classifier is arranged to classify prostatic tissue into benign, prostatic intraepithelial neoplasia (PIN), and Gleason scale adenocarcinoma grades 1 to 5 categories and to use Bayesian posterior probabilities to determine a strength of a diagnosis, wherein a borderline prognosis between two categories is provided to a second phase classifier using a classification model whose parameters are tuned to the two categories of the borderline prognosis.

2. The method of claim 1, wherein said classifier is trained to detect a most prominent and a second most prominent pattern in said image data, and to compute a Gleason score as a sum of Gleason grades of said patterns.

3. The method of claim 1, wherein said classifier is trained using a multi-class support vector machine using a probabilistic interpretation of the classifier output.

4. The method of claim 1, wherein said classifier is trained using a multi-class boosting algorithm using a probabilistic interpretation of the classifier output.

5. The method of claim 1, wherein said slide is co-stained with an AMACR biomarker.

6. The method of claim 1, wherein said discriminating attributes include boundary and region descriptors, structural descriptors, and texture descriptors.

7. A program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method steps for unsupervised classification of histological images of prostatic tissue, said method comprising the steps of:

providing histological image data obtained from a slide simultaneously co-stained with NIR fluorescent and Hematoxylin-and-Eosin (H&

E) stains;

segmenting prostate gland units in the image data;

forming feature vectors by computing discriminating attributes of the segmented gland units; and

using said feature vectors to train a multi-class classifier within a Bayesian framework, wherein said classifier is arranged to classify prostatic tissue into benign, prostatic intraepithelial neoplasia (PIN), and Gleason scale adenocarcinoma grades 1 to 5 categories and to use Bayesian posterior probabilities to determine a strength of a diagnosis, wherein a borderline prognosis between two categories is provided to a second phase classifier using a classification model whose parameters are tuned to the two categories of the borderline prognosis.

8. The computer readable program storage device of claim 7, wherein said classifier is trained to detect a most prominent and a second most prominent pattern in said image data, and to compute a Gleason score as a sum of Gleason grades of said patterns.

9. The computer readable program storage device of claim 7, wherein said classifier is trained using a multi-class support vector machine using a probabilistic interpretation of the classifier output.

10. The computer readable program storage device of claim 7, wherein said classifier is trained using a multi-class boosting algorithm using a probabilistic interpretation of the classifier output.

11. The computer readable program storage device of claim 7, wherein said slide is co-stained with an AMACR biomarker.

12. The computer readable program storage device of claim 7, wherein said discriminating attributes include boundary and region descriptors, structural descriptors, and texture descriptors.