Method of diagnosing breast cancer

• US 6,056,690 A
• Filed: 12/26/1997
• Issued: 05/02/2000
• Est. Priority Date: 12/27/1996
• Status: Expired due to Fees

• Current Assignee: Linda M. Roberts
• Original Assignee: Linda M. Roberts
• Inventors: Roberts, Linda M.
• Primary Examiner(s): O'Connor, Cary
• Assistant Examiner(s): ASTORINO, MICHAEL C
• Application Number:
  US08/998,494
• Time in Patent Office: 858 Days
• Field of Search: 600/368, 600/300-301, 128/903, 128/920-925
• US Class Current: 600/300
• CPC Class Codes: G06F 18/256 of results relating to different input data, e.g. multimodal recognition
• CPC Class Codes: G16H 15/00 ICT specially adapted for medical reports, e.g. generation or transmission thereof
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 70/60 relating to pathologies
• CPC Class Codes: Y10S 128/92 Computer assisted medical diagnostics

Abstract

A method of diagnosing breast cancer by generating an inference system having a predetermined number of nodes organized into a tree of cliques. After the inference system is initialized, an evidence vector is created and initialized. The evidence vector is then customized or updated by receiving patient-specific evidence. This updated evidence vector is then propagated through the inference system to determine a breast cancer diagnosis. While determine the breast cancer diagnosis, the system can estimate the influence of each piece of patient-specific evidence on the diagnosis. Thus, the system is capable of not only determining and reporting a breast cancer diagnosis based upon patient-specific evidence, but also determining and reporting the estimated influence of each piece of patient-specific evidence considered.

Claims

1. A method of diagnosing breast cancer comprising the steps of

generating an inference engine;

calculating probability data using conditional and nonconditional values;

receiving patient-specific evidence;

propagating said patient-specific evidence through said inference engine to update said probability data and thereby determine a breast cancer diagnosis; and

reporting said breast cancer diagnosis.

2. The method of claim 1, wherein said generating an inference engine step further comprises generating an inference engine that uses Bayesian network theory and techniques.

3. The method of claim 2, wherein said receiving patient-specific evidence step further comprises acquiring evidence in a single-patient mode.

4. The method of claim 2, wherein said receiving patient-specific evidence step further comprises acquiring evidence in a multiple-patient mode.

5. The method of claim 3 or 4, wherein said receiving patient-specific evidence step further comprises entering evidence through a user interface selected from the group consisting of graphical user interfaces, hypertext documents, database forms, and batch files.

6. A method of diagnosing breast cancer comprising the steps of

generating a Bayesian network having a first predetermined number of nodes and a second predetermined number of cliques;

initializing said Bayesian network;

creating an evidence vector;

initializing said evidence vector;

receiving patient-specific evidence;

updating said evidence vector using said patient-specific evidence;

propagating said updated evidence vector through said Bayesian network to determine a breast cancer diagnosis;

estimating an influence of each piece of patient-specific evidence on said breast cancer diagnosis;

reporting said breast cancer diagnosis; and

reporting said estimated influence.

7. A method of diagnosing breast cancer using an inference engine, comprising the steps of

(a) generating a Bayesian network having a predetermined number of member components;

(b) initializing said Bayesian network;

(c) creating an evidence vector;

(d) initializing said evidence vector;

(e) computing a posterior probability value of breast cancer by propagating said evidence vector through said network;

(f) customizing said evidence vector by specifying observed evidence values for a specific patient;

(g) verifying that said customized evidence vector is correct;

(h) updating said posterior probability value of breast cancer by propagating said customized evidence vector through said network; and

(i) outputting said posterior probability value.

8. The method of claim 7, wherein said generating a Bayesian network step further comprises creating twenty-six nodes as said member components, and organizing said nodes into a tree of twenty-one cliques comprising groups of said nodes.

9. The method of claim 8, wherein said initializing said Bayesian network step further comprises assigning initial values to each of said twenty-six nodes and said twenty-one cliques.

10. The method of claim 8, wherein said twenty-six nodes comprise five nodes without parents and twenty-one nodes with parents, and wherein said initializing said Bayesian network step further comprises

(a) assigning initial prior probability vectors to all nodes without parents;

(b) assigning initial conditional probability vectors to all nodes with parents;

(c) calculating initial joint probability vectors for each of said nodes;

(d) storing said initial joint probability vectors;

(e) calculating new prior probability vectors for each of said nodes with parents using said initial joint probability vectors; and

(f) storing said new prior probability vectors.

11. The method of claim 8, wherein said initializing said Bayesian network step further comprises

(a) initializing each of said twenty-six nodes with the following information:

number of possible states, textual name of node for reports, node number for identification, number of parent nodes, number of children nodes, size of an associated joint vector, children nodes, parent nodes, clique node is a member of, prior probability vector, conditional probability vector; and

(b) initializing each of said twenty-one cliques with the following information;

number of node members in said clique, clique number for identification, root clique indicator, number of links to other cliques, list of links to other cliques, number of common nodes, member nodes referenced by node number, common nodes referenced by node number, and base node.

12. The method of claim 8, wherein said creating an evidence vector step further comprises creating an evidence vector containing an entry corresponding to each of said twenty-six nodes.

13. The method of claim 12, wherein said twenty-six nodes comprise five nodes without parents and twenty-one nodes with parents, and wherein said updating said posterior probability value of breast cancer step further comprises

(a) selecting an observed evidence value from said evidence vector;

(b) calculating new joint probability vectors for each of said cliques comprising said node corresponding to said selected observed evidence value;

(c) storing said new joint probability vectors;

(d) calculating new prior probability vectors for each of said nodes with a parent using said new joint probability vectors; and

(e) storing said new prior probability vectors.

14. The method of claim 8, wherein, in said step (h) of claim 7, said propagating said customized evidence vector further comprises marking one of said cliques as a root clique, and traversing said clique tree starting at said root clique.

15. The method of claim 13, further comprising the steps of

(a) storing said posterior probability value following step (e) of claim 5;

(b) storing said posterior probability value following step (h) of claim 5;

(c) storing said selected observed evidence value;

(d) calculating a difference between said posterior probability value stored in step (a) of this claim and said posterior probability value stored in step (b) of this claim;

(e) determining a reason for said difference calculated in step (d) of this claim using said stored selected observed evidence value; and

(f) outputting said reason.

16. A method of diagnosing breast cancer comprising the steps of

generating a Bayesian network having twenty-six nodes and twenty-one cliques;

initializing said Bayesian network;

creating an evidence vector;

initializing said evidence vector;

receiving patient-specific evidence;

updating said evidence vector using said patient-specific evidence;

propagating said updated evidence vector through said Bayesian network to determine a breast cancer diagnosis;

estimating an influence of each piece of patient-specific evidence on said breast cancer diagnosis;

reporting said breast cancer diagnosis; and

reporting said estimated influence.

17. The method of claim 16, wherein said receiving patient-specific evidence step further comprises receiving patient-specific evidence concerning at least one of the following factors:

age in years, age of menarche, age of first live birth, number of first order relatives with cancer, previous biopsy at site, breast pain, nipple discharge, palpable mass, developing density in breast, asymmetry in breast, dilated duct in breast, architectural distortion, halo sign, tumor location by quadrant, mass size, mass margin, mass density, calcification shape, calcification density, calcification number, calcification size, calcification cluster shape, and calcification arrangement.

18. The method of claim 16, wherein said receiving patient-specific evidence step further comprises the following:

receiving demographic factors;

receiving physical examination findings; and

receiving mammographic findings.

19. The method of claim 18, wherein said receiving mammographic findings step comprises receiving direct mammographic findings and receiving indirect mammographic findings.

20. The method of claim 18, wherein said receiving demographic factors step comprises receiving information concerning at least one of the following factors:

age in years, age of menarche, age of first live birth, number of first order relatives with cancer, and previous biopsy at site.

21. The method of claim 20, wherein said receiving physical examination findings step comprises receiving information concerning at least one of the following factors:

breast pain, nipple discharge, and palpable mass.

22. The method of claim 21, wherein said receiving mammographic findings step comprises receiving information concerning at least one of the following factors:

mass margin, mass density, mass size, halo sign, tumor location by quadrant, calcification cluster shape, calcification number, calcification shape, calcification density, calcification arrangement, calcification size, architectural distortion, asymmetry in breast, developing density in breast, and dilated duct in breast.

23. The method of claim 21, wherein said receiving mammographic findings step comprises receiving direct mammographic findings and receiving indirect mammographic findings.

24. The method of claim 23, wherein said receiving direct mammographic findings step comprises receiving information concerning at least one of the following factors:

mass margin, mass density, mass size, halo sign, tumor location by quadrant, calcification cluster shape, calcification number, calcification shape, calcification density, calcification arrangement, and calcification size; and

wherein said receiving indirect mammographic findings step comprises receiving information concerning at least one of the following factors;

architectural distortion, asymmetry in breast, developing density in breast, and dilated duct in breast.