Segmentation and data mining for gel electrophoresis images

• US 20060257053A1
• Filed: 06/16/2004
• Published: 11/16/2006
• Est. Priority Date: 06/16/2003
• Status: Abandoned Application

• Current Assignee: Alexandre J. Boudreau, Claude Kauffmann, Dubé, Patrick, Khaldoune Zine El Abidine
• Original Assignee: Alexandre J. Boudreau, Claude Kauffmann, Dubé, Patrick, Khaldoune Zine El Abidine
• Inventors: El Abidine, Khaldoune Zine, Boudreau, Alexandre J., Dubé, Patrick, Kauffmann, Claude
• Application Number:
  US10/563,706
• Publication Number:
  US 20060257053A1
• Time in Patent Office: Days
• US Class Current: 382/305
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G16B 40/00 ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
• CPC Class Codes: G16B 40/20 Supervised data analysis

Abstract

A segmentation method is provided for the automated segmentation of spot-light structures into D images allowing precise quantification and classification of said structures and said images, based on a plurality of criteria, and further allowing the automated identification of multi-spot based patterns present in one or a plurality of images. In a preferred embodiment, the invention is used for the analysis of 2D gel electrophoresis images, with objective of quantifying protein expressions and for allowing sophisticated multi-protein pattern based image data mining, as well as image matching, registration, and automated classification.

Claims

1. An image and data management method, comprising the steps of:

displaying an image;

producing, displaying, and positioning at least one graphical marker in at least one context of said image;

selecting at least one external data to associate to at least one of said graphical marker, wherein said external data is selected in one or a plurality of local or remote repositories;

associating at least one of said external data to at least one of said graphical marker and displaying a visual indication of said association; and

saving information in one or a plurality of local or remote repositories, said information comprising at least data defining said association.

2. The method as claimed in claim 1 wherein said context is a region of interest, said region of interest being a user defined region composed of pixel values.

3. The method as claimed in claim 2 wherein defining a region of interest comprises the steps of:

providing a tool to the user for defining said region of interest;

interactively defining contour of said region of interest within said image using said tool, said contour being displayed in said image; and

automatically associating said pixel values of said user defined region to said graphical marker.

4. The method as claimed in claim 1 wherein said context is a region of interest, said region of interest being an automatically defined region composed of pixel values by means of an automated segmentation method.

5. The method as claimed in claim 4 further comprising automatically associating said graphical marker to said pixel values of said automatically defined region.

6. The method as claimed in claim 1 further comprising a means for displaying at least one of said external data.

7. The method as claimed in claim 1 wherein said step of producing, displaying and positioning said graphical marker is achieved automatically by means of a program.

19. The method as claimed in claim 4, wherein said automated segmentation method is provided by

computing statistical distribution of noise information in said image;

computing a multiscale analysis level N in accordance to said statistical distribution;

computing a multiscale image of said image up to said level N, and generating at least one type of regionalization of said multiscale image;

identifying objects of interest in said image in correspondence with said multiscale image and said regionalization;

identifying organized structures in said image said organized structures not objects of interest; and

characterizing and classifying said objects of interest.

8. A system for analyzing and managing image information, comprising:

image input means for inputting an image;

image analysis program for automatically identifying and quantifying objects of interest within said image, said program producing image information;

association program for associating multi-source information to said image and said objects of interest, said step of associating producing associative information;

display program for displaying said image, at least some of said multi-source information, and for producing and displaying graphical information in context of said objects of interest of said image; and

storage means and program for storing said image, said image information, said graphical information, and said associative information in local or remote repositories.

9. The system as claimed in claim 8, further comprising:

means for automatically searching one or a plurality of said repositories for images that satisfy one or a plurality of data-mining criteria, said data-mining criteria being manually or automatically defined;

means for automatically producing and displaying searching results, said searching results composed of at least a list of found images; and

means for selecting and displaying at least one of said images from said mining results by activating at least one element of said list, wherein said displaying comprises emphasizing said objects of interest of said selected images.

20. The system as claimed in claim 8, wherein said image analysis program uses the method of

computing statistical distribution of noise information in said image;

computing a multiscale analysis level N in accordance to said statistical distribution;

computing a multiscale image of said image up to said level N, and generating at least one type of regionalization of said multiscale image;

identifying objects of interest in said image in correspondence with said multiscale image and said regionalization;

identifying organized structures in said image said organized structures not objects of interest; and

characterizing and classifying said objects of interest.

10. A system for providing object-based image discovery, comprising:

image input means for inputting an image;

image analysis program for automatically identifying and quantifying objects of interest within said image, said program producing image information, said image and said image information stored in at least one repository;

a user input means for inputting a discovery criteria;

a searching program for searching within said repositories for images that satisfy said discovery criteria; and

a display means for displaying searching results and said images.

23. The system as claimed in claim 10, wherein said image analysis program uses the method of

computing statistical distribution of noise information in said image;

computing a multiscale analysis level N in accordance to said statistical distribution;

computing a multiscale image of said image up to said level N, and generating at least one type of regionalization of said multiscale image;

identifying objects of interest in said image in correspondence with said multiscale image and said regionalization;

identifying organized structures in said image said organized structures not objects of interest; and

characterizing and classifying said objects of interest.

11. A method for automatic spot detection in digital images, comprising the steps of:

reading an image;

computing statistical distribution of noise information in said image;

computing a multiscale analysis level N in accordance to said statistical distribution;

computing a multiscale image of said image up to said level N, and generating at least one type of regionalization of said multiscale image;

identifying objects of interest in said image in correspondence with said multiscale image and said regionalization;

identifying organized structures in said image said organized structures not objects of interest; and

characterizing and classifying said objects of interest.

14. The method as claimed in claim 11, wherein said step of characterizing is achieved by

computing a multiscale representation of said image up to a level N, wherein said step of computing providing a multiscale image;

identifying and defining spot object regions on each of said levels of said multiscale image; and

linking said spot object regions identified on each of said levels of said multiscale image, said linking creating a multiscale event tree, said multiscale event tree providing information for characterizing and classifying said spot objects.

15. The method as claimed in claim 11, wherein said step of classifying is achieved by means of an artificial neural network.

16. The method as claimed in claim 11, wherein said organized structures are smear lines.

17. The method as claimed in claim 11, wherein said organized structures are image artifacts, said image artifacts including air bubbles, hair, rips, and scratches.

12. A method for automatically attributing a confidence level to one or a plurality of spot objects in a digital image, comprising the steps of:

reading an image;

automatically identifying spot objects in said image;

computing confidence level of said spot objects; and

displaying confidence level for at least one of said spot objects.

21. The method as claimed in claim 12, wherein said step of automatically identifying is achieved by means of the method of

computing statistical distribution of noise information in said image;

computing a multiscale analysis level N in accordance to said statistical distribution;

computing a multiscale image of said image up to said level N, and generating at least one type of regionalization of said multiscale image;

identifying objects of interest in said image in correspondence with said multiscale image and said regionalization;

identifying organized structures in said image said organized structures not objects of interest; and

characterizing and classifying said objects of interest.

13. A method for characterizing spot objects in an image, comprising:

computing a multiscale representation of said image up to a level N, wherein said step of computing providing a multiscale image;

identifying and defining spot object regions on each of said levels of said multiscale image; and

linking said spot object regions identified on each of said levels of said multiscale image, said linking creating a multiscale event tree, said multiscale event tree providing information for characterizing and classifying said spot objects.

18. The method as claimed in claim 13, wherein said spot object regions are watershed regions.

22. A method for quantifying identified spot objects, comprising the steps of:

computing one or a plurality of 2D diffusion functions;

fitting said diffusions functions to said identified spot objects by varying parameters of said diffusion functions in order to optimize said fitting, said parameters providing the variance, width and height of said diffusion functions;

simulating and calculating cumulative effect of said identified spot objects by means of said diffusion functions; and

quantifying said identified spot objects without said cumulative effect by means of said diffusion functions.