Apparatus for automated identification of thick cell groupings on a biological specimen

• US 5,987,158 A
• Filed: 11/13/1997
• Issued: 11/16/1999
• 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: Meyer, Michael G., Lee, Shih-Jong J., Wilhelm, Paul S.
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): Prikockis, Larry J.
• Application Number:
  US08/969,970
• Time in Patent Office: 733 Days
• Field of Search: 382/133, 382/128, 382/134, 382/159, 382/160, 382/224, 382/225, 382/283, 356/39, 356/335, 356/336, 877/11
• US Class Current: 382/133
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/30024 Cell structures in vitro; Tissue sections in vitro
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 20/69 Microscopic objects, e.g. biological cells or cellular parts

Abstract

A thick group of cells classifier. Image data acquired from an automated microscope from a cytological specimen is processed by a computer system. The computer applies filters at different stages. Obvious artifacts are eliminated from analysis early in the processing. The first stage of processing is image segmentation where objects of interest are identified. The next stage of processing is feature calculation where properties of each segmented thick group object are calculated. The final step is object classification where every segmented thick group object is classified as being abnormal or as belonging to a cellular or non-cellular artifact.

Claims

1. In an automated system for analyzing a biological specimen having a thick group of cells, a thick group of cells classification apparatus comprising:

(a) an automated microscope having at least one image output;

(b) image segmentation means for identifying objects of interest connected to the image output, wherein the image segmentation means has a segmented image output;

(c) object feature calculation means for measuring properties of each segmented object having a segmented object output wherein the object feature calculation means has a thick group of cells object feature output; and

(d) an object classification means for classifying thick group objects for analysis having an input connected to the thick group of cells object feature output, where the object classification means has a thick group of cells classification output, and where the object classification means comprises a multiple stage algorithmic classifier, wherein every stage is an algorithmic classifier and wherein at least one of the stages comprises a Fisher'"'"'s binary decision tree classifier.

2. The apparatus of claim 1 wherein the biological specimen is a specimen prepared by the Papanicolaou method.

3. The apparatus of claim 1 wherein the biological specimen is a gynecological specimen.

4. The apparatus of claim 1 wherein the object classification means further comprises a digital computer.

5. An apparatus for training a system to recognize thick groups of interest comprising:

(a) means for acquiring at least one image from at least one biological specimen;

(b) means for detecting at least one thick group of cells for analysis from the at least one image, wherein the means for detecting has a thick group of cells output;

(c) means for obtaining diagnostic truth connected to the thick group of cells output, wherein the means for obtaining diagnostic truth has a truth established thick group of cells output;

(d) a means for computing features connected to the truth established thick group of cells output, wherein the means for computing features has a feature output;

(e) means for selecting features that best discriminate between thick groups of interest and other groups, wherein the means for selecting features has a discrimination output; and

(f) classification means for separating thick groups of interest from other groups connected to the discrimination output, wherein the classification means has a classification output, where the object classification means comprises a multiple stage algorithmic classifier wherein every stage is an algorithmic classifier, and wherein at least one of the stages comprises a Fisher'"'"'s binary decision tree classifier.

6. The apparatus for training a system to recognize thick groups of interest of claim 5 further comprising means for determining an overall gain for the classification means.

7. The apparatus of claim 5 wherein the thick group of interest comprises an abnormal group of cells.

8. The apparatus of claim 5 wherein the thick group of interest comprises an artifact.

9. The apparatus of claim 5 wherein the thick group of interest comprises a thick group of normal cells.

10. The apparatus of claim 5 wherein the biological specimen is a specimen prepared by the Papanicolaou method.

11. The apparatus of claim 5 wherein the biological specimen is a gynecological specimen.

12. The apparatus of claim 5 wherein the thick group detection means further comprises a digital computer.

13. In an automated system for analyzing a biological specimen having a thick group of cells, a thick group of cells classification process comprising the steps of:

(a) obtaining at least one image output;

(b) segmenting the at least one image output to identify objects of interest;

(c) measuring properties of each segmented object of interest to identify a plurality of object features indicative of a thick group of cells; and

(d) processing the plurality of object features indicative of a thick group of cells to classify thick group objects for analysis by applying a multiple stage algorithmic classifier, wherein every stage is an algorithmic classifier and wherein at least one of the stages comprises a Fisher'"'"'s binary decision tree classifier.

14. The process of claim 13 wherein the biological specimen is a specimen prepared by the Papanicolaou method.

15. The apparatus of claim 13 wherein the biological specimen is a gynecological specimen.

16. The apparatus of claim 13 wherein the object classification means further comprises a digital computer.