Method and apparatus for robust biological specimen classification

• US 5,740,269 A
• Filed: 09/20/1994
• Issued: 04/14/1998
• Est. Priority Date: 09/20/1994
• Status: Expired due to Term

• Current Assignee: Tripath Imaging Incorporated (Becton, Dickinson & Co)
• Original Assignee: NeoPath, Inc. (Becton, Dickinson & Co)
• Inventors: Oh, Seho, Lee, Shih-Jong J., Kuan, Chih-Chau L., Bannister, Wendy R.
• Primary Examiner(s): Couso, Yon J.
• Assistant Examiner(s): Patel, Jayanti K.
• Application Number:
  US08/309,209
• Time in Patent Office: 1,302 Days
• Field of Search: 382/133, 382/159, 382/134, 382/224, 382/227, 382/128, 364/413.1
• US Class Current: 382/133
• CPC Class Codes: G06V 20/69 Microscopic objects, e.g. biological cells or cellular parts

Abstract

A robust biological specimen classifier. An automated microscopy system obtains features from an image of a biological specimen slide. A computer system in the automated microscopy system computes feature variations. Clusters are created that comprise samples of similar characteristics. A cluster membership generator generates membership values for each cluster. Classifiers specialized to certain feature sets are used to provide independent outputs. These outputs are offset and biased by the output of the membership generator. The output of the adjusted classification values are summed to create a slide score output.

Claims

1. An automated apparatus for classification of a Pap stained slide, the automated apparatus comprising:

(a) means for acquiring at least one image from the Pap stained slide having at least one image output;

(b) means for obtaining at least one slide feature from the at least one image having at least one slide feature output;

(c) a plurality of classifiers, each connected to the at least one slide feature output and each having a classification output;

(d) means for applying an offset to each classification output of the plurality of classifiers wherein the means for applying an offset has a plurality of offset outputs;

(e) means for membership generation connected to the at least one slide feature output wherein the means for membership generation has a plurality of membership value outputs;

(f) means for adjusting the plurality of offset outputs with the membership value outputs wherein the means for adjusting the plurality of offset outputs has a plurality of adjusted classification outputs; and

(g) means for aggregating the adjusted classification outputs wherein the means for aggregating the adjusted classification outputs has an aggregated classification output.

2. The apparatus of claim 1 wherein:

(a) the means for applying an offset has an offset terminal summer for each of the plurality of classifiers to offset the classification output of each of the plurality of classifiers with a predetermined offset value having an offsetted classifier output;

(b) the means for membership generation has a membership generator having a membership weighting factor output for each one of the plurality of classifiers;

(c) the means for adjusting the plurality of offset outputs has a multiplier for multiplying each of the membership weighting factor by each offsetted classifier output having membership adjusted outputs for each offsetted classifier output; and

(d) the means for aggregating the adjusted classification outputs has a summer to sum all of the membership outputs to produce a robust classification output.

3. The apparatus of claim 2 wherein the membership generator comprises a membership classifier.

4. The apparatus of claim 3 wherein the membership classifier comprises a K means clustering classifier.

5. The apparatus of claim 3 wherein the membership classifier comprises a fuzzy classifier.

6. The apparatus of claim 3 wherein the membership classifier comprises a Bayes belief network.

7. The apparatus of claim 5 wherein the fuzzy classifier comprises a fuzzy isodata clustering classifier.

8. The apparatus of claim 5 wherein the fuzzy classifier comprises a C means clustering classifier.

9. The apparatus of claim 5 wherein the fuzzy classifier comprises a fuzzy decision tree.

10. The apparatus of claim 2 wherein the membership generator comprises at least one membership feature extractor.

11. The apparatus of claim 10 wherein the at least one membership feature extractor comprises a specimen staining density extractor.

12. The apparatus of claim 10 wherein the at least one membership feature extractor comprises a specimen staining contrast extractor.

13. The apparatus of claim 10 wherein the at least one membership feature extractor comprises a reference cell nuclear size extractor.

14. The apparatus of claim 10 wherein the at least one membership feature extractor comprises at least one counter for objects of interest.

15. The apparatus of claim 1 wherein the Pap stained slide is a gynecological specimen.

16. An automated apparatus for classification of a biological specimen slide, the automated apparatus comprising:

(a) means for acquiring an image from the biological specimen slide having an image output;

(b) means for obtaining at least one slide feature from the image output having at least one slide feature output; and

(c) means for slide scoring using a plurality of classifiers, wherein each of the plurality of classifiers has a feature input connected to the at least one slide feature output and has a classified output wherein each classified output is offset to adjust for variations in at least one slide feature and adjusted by membership values, wherein the adjusted values are aggregated to generate an analysis score output.

17. The apparatus of claim 16 further comprising a means for deciding whether a slide is normal or requires review, having an input connected to the at least one analysis score output and having a decision output.

18. The apparatus of claim 16 wherein the means for slide scoring comprises a membership classifier.

19. The apparatus of claim 18 wherein the membership classifier comprises a K means clustering classifier.

20. The apparatus of claim 18 wherein the membership classifier comprises a fuzzy classifier.

21. The apparatus of claim 20 wherein the fuzzy classifier comprises a fuzzy isodata clustering classifier.

22. The apparatus of claim 20 wherein the fuzzy classifier comprises a C means clustering classifier.

23. The apparatus of claim 20 wherein the fuzzy classifier comprises a fuzzy decision tree.

24. The apparatus of claim 18 wherein the membership classifier comprises a Bayes belief network.

25. The apparatus of claim 16 wherein the biological specimen is a gynecological specimen.

26. The apparatus of claim 20 wherein the fuzzy classifier comprises a means for calculating a distance metric.

27. The apparatus of claim 26 wherein the means for calculating a distance metric comprises a means for calculating the distance metric in feature space.

28. The apparatus of claim 26 wherein the means for calculating a distance metric comprises a means for calculating the distance metric in classification performance space.

29. The apparatus of claim 28 wherein the means for calculating the distance metric in classification performance space further comprises a means for measuring an area under a receiver operating characteristic curve.