Dynamic control and decision making method and apparatus

• US 6,198,839 B1
• Filed: 09/05/1997
• Issued: 03/06/2001
• Est. Priority Date: 09/05/1997
• Status: Expired due to Term

• Current Assignee: Tripath Imaging Incorporated (Becton, Dickinson & Co)
• Original Assignee: Tripath Imaging Incorporated (Becton, Dickinson & Co)
• Inventors: Bannister, Wendy R., Meyer, Michael G., Oh, Seho, Lee, Shih-Jong J., Kuan, Chih-Chau L.
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): Wu, Jingge
• Application Number:
  US08/924,351
• Time in Patent Office: 1,278 Days
• Field of Search: 128/920, 128/923, 128/925, 128/922, 382/133, 382/134, 382/224, 382/128, 356/92
• US Class Current: 382/133
• CPC Class Codes: G01N 15/1433 using image recognition
• CPC Class Codes: G06V 20/69 Microscopic objects, e.g. biological cells or cellular parts
• CPC Class Codes: Y10S 128/922 including image analysis

Abstract

Dynamic control of the processing flow of an image analyzer such as a biological specimen analyzer as processing proceeds. Data collected and processed from a specimen under analysis, such as a biological specimen on a microscope slide, determines the fate of further processing. If there is enough evidence, based on the data collected from a slide, to make a decision with sufficient confidence, the processing of the slide can be stopped and a decision may be rendered. By avoiding unnecessary additional computation system throughput may be enhanced. Otherwise, data collection and computation continues until either certain termination criteria are met or no more data is left to acquire. This slide-dependent control and decision making method flexibly limits the amount of computation required to reach a system decision about a specimen. By evaluating analysis processing continuously a maximum signal to noise ratio may be achieved by preventing additional noise from entering the analysis and thus swamping signal information.

Claims

1. An automated cytological analysis screening method for characterizing a biological specimen comprising the steps of:

(a) acquiring an image data set representative of an image of a biological specimen within a field of view of said specimen, wherein said selected image is selected in accordance with a predetermine scan sequence of fields of view of said biological specimen;

(b) operating on said image data in accordance with a first image processing module for determining a first feature set having at least a one measurement value, M1, indicative of the magnitude of presence of at least a first feature, F1, of said selected image;

(c) selectively, in response to said first feature set, executing step (x)

operating on said image data by at least one different image processing module selected from a plurality of image processing modules, N, for determining a second feature set including at least one additional measurement value representative of a corresponding image feature, where said selected image processing module is selected based on said first feature set, and selectively executing step (x) in response to said second feature set;

operating on said image data by other processing modules selected from a plurality of image processing modules, N, for determining other measurement values representative of a corresponding image features, where said selected image processing modules are selected based on previously obtained feature set results of a previously image data set of a previous field of view;

selectively changing said selected scanning sequence and performing steps (a), (b), and (c), and

repeating step (a) if images remain, otherwise terminate analysis of said biological specimen; and

(d) characterizing said biological specimen and terminate further image processing of said biological specimen.

2. The method of claim 1 further includes the step of:

operating on said image data by at least another different image processing module selected from said plurality of image processing modules for determining an additional feature set including at least one additional measurement value representative of a corresponding image feature, where said selected image processing module is selected based on feature set results of at least two different images corresponding to two different fields of view, respectively.

3. The method of claim 1 further includes the step of:

terminating operating on said image data set by previously selected ones of said processing modules based on said feature set of a previously processed image and a last obtained feature set result of a current image data set.

4. The method of claim 1 further including the step of separately accumulating selected measurement values associated with mutually exclusive features, and analyzing said accumulated measurement values for subsequent decision making to selectively terminate or continue image processing.

5. The method of claim 1 wherein said plurality image processing modules include a group of modules consisting of a single cell classification module, a group classification module, a thick group classification module, an endocervical classification module, a cellular object classification module, and a poly detection module.

6. A method of image processing a biological specimen on a slide by an automated cytological analysis screening system for classifying a biological specimen as being normal or requiring subsequent review, where said automated cytological analysis screening system includes a plurality of unique image processing modules where each image processing module serves to determine a measurement value indicative of a unique feature of a processed image, said method comprising the steps of:

(a) image scanning said slide so as to determine a prioritization scan sequence identifying those areas on said slide where an image scanning field of view of those areas have selected cellular characteristics indicative of a likelihood of abnormal cells;

(b) fetching, in accordance with said prioritized scan sequence, an image data set representative of an image in a field of view;

(c) image processing said image data set for determining a first feature set, including at least one feature, where said feature set is representative of presence of any selected feature in the acquired image data set;

(d) selectively, in response to said feature set, executing step (x),

opeating on said image data set by first selected additional ones of said plurality of unique image processing modules, for determining a second feature set, where said selected image processing modules are selected based on said first feature set, and choosing to go to step (x), or go to another image processing step based on said second feature set,

enabling and disabling selected ones of said plurality of processing modules in response to said first feature set, operating on said image data set by enabled additional selected ones of said plurality of processing modules, determining a second feature set including at least a measurement value associated with at least one feature, and choosing to go to step (x), or go to sstep (b) so as to acquire image data set associated with the next field of view in accordance with said prioritized scan sequence;

selectively changing said prioritization scan sequence and repeating above steps (b), (c), and (d);

and

(e) characterize said biological specimen as abnormal and terminate image fetching and image processing further fields of view.

7. The method of claim 6 further includes the step of:

selectively enabling and disabling second addition selected ones of said plurality of processing modules in response to said second feature set selectively based on said feature sets based on image data associated with two sequential ones of said field of view;

operating on said image data by said enabled second additional selected ones of said plurality of processing modules, and choosing, based on a feature set associated with said enabled second additional set of processing modules to go to step (x), or go to the (b) so as to acquire image data associated with the next field of view in accordance with said prioritized scan sequence.

8. The method of claim 6 further includes the step of separately accumulating selected measurement values associated with mutually exclusive features, and determining after each fetching of an image data set associated with an individual field of view whether to continue processing images, terminate image processing and characterize said specimen as being normal or needing review.

9. The method of claim 6 wherein said plurality image processing modules include at least a group of modules consisting of a single cell classification module, a group classification module, a thick group classification module, an endocervical classification module, a cellular object classification module, and a poly detection module.

10. The method of claim 6 wherein said selected prioritization scan sequence is a function of the quantity of endocervical cell groups and abnormal or glandular cells, where said prioritization scan sequence begins with those having high score values of both endocervical cell groups and abnormal or glandular cells.

11. The method of claim 6 wherein said selected prioritization scan sequence is selected as a function of (i) a Group Score, Z, indicative of the quantity of endocervical cells, and a SIL Score, A, indicative of the quantity of abnormal or glandular cells, determined to be present in an image of a field of view.

12. The method of claim 11 wherein said selected prioritization scan sequence is selectively chosen from the group where,

(i) highest priority is assigned to those field of views having both a high Group Score and a high SIL score,

(ii) highest priority is assigned to those field of views having a high Group Score, or

(iii) highest priority is assigned to those field of views having a high SIL Score.

13. The method of claim 11 further where,

a selected one image processing module determines (i) a Group Score, Z, indicative of the quantity of endocervical cells, and a SIL Score, A, indicative of the quantity of abnormal or glandular cells, within a processed image of a filed of view, and

step (c) includes the step of comparing said Group Score Z to a predetermined threshold Z_th, and comparing said SIL Score to a predetermined threshold A_th, and selectively choosing said selected prioritization scan sequence based on said threshold comparisons.

14. The method of claim 12 wherein said selected prioritization scan sequence is selectively chosen from the group where,

(iv) highest priority is assigned to those field of views having both a high Group Score and a high SIL score,

(v) highest priority is assigned to those field of views having a high Group Score, or

(vi) highest priority is assigned to those field of views having a high SIL Score.