Computer-aided image analysis

• US 7,383,237 B2
• Filed: 02/06/2006
• Issued: 06/03/2008
• Est. Priority Date: 05/01/1998
• Status: Expired due to Fees

• Current Assignee: Curtis Anderson, Health Discovery Corporation, James Roberts, Joe Mckenzie, Jules B. Paderewski, Julian N. Stern, Memorial Health Systems Incorporated, Timothy P. O'Hayer
• Original Assignee: Health Discovery Corporation
• Inventors: Zhang, Hong, Carls, Garry, Barnhill, Stephen D.
• Primary Examiner(s): GURLEY, LYNNE ANN
• Application Number:
  US11/349,542
• Publication Number:
  US 20060224539A1
• Time in Patent Office: 848 Days
• Field of Search: None
• US Class Current: 706/20
• CPC Class Codes: G06F 18/2113 by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
• CPC Class Codes: G06F 18/2411 based on the proximity to a decision surface, e.g. support vector machines
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images
• CPC Class Codes: G16H 30/40 for processing medical images, e.g. editing
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

Digitized image data are input into a processor where a detection component identifies the areas (objects) of particular interest in the image and, by segmentation, separates those objects from the background. A feature extraction component formulates numerical values relevant to the classification task from the segmented objects. Results of the preceding analysis steps are input into a trained learning machine classifier which produces an output which may consist of an index discriminating between two possible diagnoses, or some other output in the desired output format. In one embodiment, digitized image data are input into a plurality of subsystems, each subsystem having one or more support vector machines. Pre-processing may include the use of known transformations which facilitate extraction of the useful data. Each subsystem analyzes the data relevant to a different feature or characteristic found within the image. Once each subsystem completes its analysis and classification, the output for all subsystems is input into an overall support vector machine analyzer which combines the data to make a diagnosis, decision or other action which utilizes the knowledge obtained from the image.

Claims

1. A method for automated analysis of a digitized microscope image of a medical specimen having a plurality of features of interest, the method comprising:

storing a plurality of detection subsystems in a memory device in communication with a processor, wherein each of the plurality of detection subsystems comprises a pre-processing component for detecting and extracting one of the features of interest within the microscope image, a classification component comprising at least one first-level support vector machine for classifying the extracted feature of interest into at least one of a plurality of possible feature classes, an output for outputting the classified feature of interest, and an overall analyzer for combining the outputs of the plurality of detection subsystems and generating an analysis of the microscope image, the overall analyzer comprising a second-level support vector machine;

receiving digitized microscope image data to be processed in the processor; and

processing the digitized microscope image data to classify the features of interest and generate an output comprising an analysis of the digitized microscope image.

2. The method of claim 1, wherein the medical specimen is a cytology specimen.

3. The method of claim 1, wherein the medical specimen is a tissue sample.

4. The method of claim 1, wherein the microscope image comprises a plurality of a single type of feature of interest and the pre-processing component comprises a segmentation function for separating overlapping features of interest in the microscope image.

5. The method of claim 4, wherein the segmentation function comprises a gravitation model that is applied to each overlapping feature of interest to contract each feature into a distinct body.

6. The method of claim 1, wherein the pre-processing component comprises a segmentation function for separating the feature of interest from a background and generating a numerical value for the segmented feature of interest.

7. The method of claim 6, wherein segmenting comprises identifying local extremes corresponding to each segmented feature of interest in the image data.

8. The method of claim 7, wherein the feature of interest comprises a spot having a brightness and identifying local extremes comprises classifying the brightness of the spot into one or more of a plurality of brightness levels.

9. The method of claim 8, wherein geometry is a possible feature characteristic and geometry is determined measuring a change in slope between borders of the spot at two different brightness levels.

10. The method of claim 1, wherein the pre-processing component comprises a segmentation function for segmenting the feature of interest and transforming the segmented feature to a fixed dimensional vector.

11. The method of claim 10, wherein transforming comprises:

computing a centroid of the feature of interest;

sampling a contour of the feature of interest using a polar coordinate system having an origin at the centroid to provide a plurality of radial measures;

forming a vector using the plurality of radial measures; and

applying a Fourier transform to the vector to provide the fixed dimensional vector.

12. A computer system for analysis of a digitized microscope image of a medical specimen having a plurality of features of interest, the computer system comprising:

a processor programmed for executing a plurality of support vector machines;

an input device for receiving image data to be processed;

a memory device in communication with the processor having a plurality of detection subsystems stored therein for execution by the processor, each of the plurality of detection subsystems comprising;

a pre-processing component for detecting and extracting one of the features of interest within the image data;

a classification component comprising at least one first-level support vector machine for classifying the feature of interest into at least one of a plurality of possible features characteristics;

an output for outputting the classified feature of interest;

an overall analyzer for combining the outputs of the plurality of detection subsystems and generating an analysis of the digitized microscope image, the overall analyzer comprising a second-level support vector machine.

13. The computer system of claim 12, wherein pre-processing component applies a segmenting routine to separate the feature of interest from a background and generates a numerical value for the segmented feature of interest.

14. The computer system of claim 13, wherein segmenting routine identifies local extremes corresponding to each segmented feature of interest in the image data.

15. The computer system of claim 14, wherein the feature of interest comprises a spot having a brightness and local extremes are identified by classifying the brightness of the spot into one or more of a plurality of brightness levels.

16. The computer system of claim 15, wherein geometry is a possible feature characteristic and geometry is determined by measuring a change in slope between borders of the spot at two different brightness levels.

17. The computer system of claim 12, wherein the pre-processing component segments the feature of interest and applies a transform to the segmented feature to a fixed dimensional vector.

18. The computer system of claim 17, wherein transform comprises:

computing a centroid of the feature of interest;

sampling a contour of the feature of interest using a polar coordinate system having an origin at the centroid to provide a plurality of radial measures;

forming a vector using the plurality of radial measures; and

applying a Fourier transform to the vector to provide the fixed dimensional vector.

19. The computer system of claim 12, wherein each digitized image includes a plurality of a single type of feature of interest and the pre-processing component segments a first feature of interest from a second, at least partially overlapping feature of interest by applying a gravitation model to each feature of interest to contract each feature into a distinct body.

20. The method of claim 12, wherein the medical specimen is a cytology specimen.