Computer assisted methods for diagnosing diseases

• US 6,306,087 B1
• Filed: 01/10/2000
• Issued: 10/23/2001
• Est. Priority Date: 10/13/1994
• Status: Expired due to Term

• Current Assignee: Horus Therapeutics, Inc.
• Original Assignee: Horus Therapeutics, Inc.
• Inventors: Zhang, Zhen, Barnhill, Stephen D.
• Primary Examiner(s): Jaworski, Francis J.
• Application Number:
  US09/479,487
• Time in Patent Office: 652 Days
• Field of Search: 127/922-924, 600/300, 706/15, 706/20, 706/25, 382/156-158
• US Class Current: 600/300
• CPC Class Codes: G01N 33/50 Chemical analysis of biological material, e.g. blood, urine; Testing involving biospecific ligand binding methods; Immunological testing measuring or testing processes involving enzymes or microorganisms, compositions or test papers therefor; processes for forming such compositions, condition responsive control in microbiological or enzymological processes C12Q
• CPC Class Codes: G01N 33/53 Immunoassay; Biospecific binding assay; Materials therefor
• CPC Class Codes: G01N 33/573 for enzymes or isoenzymes
• CPC Class Codes: G01N 33/57484 involving compounds serving as markers for tumor, cancer, neoplasia, e.g. cellular determinants, receptors, heat shock/stress proteins, A-protein, oligosaccharides, metabolites
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• 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 50/30 for calculating health indices; for individual health risk assessment
• CPC Class Codes: G16H 50/50 for simulation or modelling of medical disorders
• CPC Class Codes: G16H 70/60 relating to pathologies
• CPC Class Codes: Y10S 128/924 using artificial intelligence

Abstract

The simultaneous multi access reasoning technology system of the present invention utilizes both existing knowledge and implicit information that can be numerically extracted from training data to provide a method and apparatus for diagnosing disease and treating a patient. This technology further comprises a system for receiving patient data from another location, analyzing the data in a trained neural network, producing a diagnostic value, and optionally transmitting the diagnostic value to another location.

Claims

1. A method for constructing a computer-based neural network based classifier for the diagnosis and prognosis of a disease for use in a trained neural network, comprising:

initially selecting primary biomarker inputs for the neural network that are relevant to the disease, wherein selecting the relevant biomarkers are dependent upon current biomedical science;

testing for discriminating power of the selected primary biomarker inputs and removing any biomarker input that does not exhibit discriminating power;

grouping the primary biomarker inputs having like properties into subsets;

preprocessing the primary biomarker inputs by combining at least two primary biomarker inputs to create secondary biomarker inputs;

testing the discriminating power of the primary and secondary biomarker inputs;

selecting the primary and secondary biomarker inputs with the highest discriminating power;

creating neural network-based classifiers by combining the selected primary and secondary biomarker inputs;

evaluating an individual neural network-based classifier against test data to rank the contribution of the individual primary and secondary biomarker inputs; and

selecting the best trained neural network classifier.

2. The method of claim 1, wherein the step of selecting the best trained neural network based classifiers comprises an iterative process having a predefined number of iterations, the iterative process comprising:

(a) determining whether the number of iterations exceeds the predefined number of iterations;

if the number of iterations exceeds the predefined number of iterations, then setting the neural network-based classifier as the best trained neural network-based classifier; and

if the number of iterations exceeds the predefined number of iterations, performing the sequence, comprising;

(b) evaluating the neural network-based classifier against test data to determine its effectiveness to diagnose the disease, comprising;

(c) determining whether the neural network-based classifier produces a result within a predefined value based on a comparison with existing methods of diagnosing diseases and the cost associated with each biomarker input associated with the neural network-based classifier;

if the determination is made that the results are within a predefined value, then

(d) creating a new neural network-based classifier by removing at least one input from the neural network-based classifier with the least contribution and adding at a least one biomarker input selected from the remaining primary and secondary biomarker inputs that has the highest discriminating power; and

(e) returning to step (a);

if the determination is made that the results are not within a predefined value, then

(f) creating a new neural network-based classifier by adding at least one biomarker input selected from the remaining primary and secondary biomarker inputs that has the highest discriminating power; and

(g) returning to step (a).

3. The method of claim 1, wherein the step of evaluating an individual neural network-based classifier against test data to rank the contribution of the individual primary and secondary biomarker inputs, comprises inspecting a neural network connection strength initiated from each primary and secondary biomarker input.

4. The method of claim 1, wherein the step of evaluating an individual neural network-based classifier against test data to rank the contribution of the individual primary and secondary biomarker inputs comprises performing a sensitivity analysis that compares a relative change in each neural network output with a change to a single biomarker input value.

5. The method of claim 1, wherein the step if evaluating an individual neural network-based classifier against test data to rank the contribution of the individual primary and secondary inputs, comprises using Monte Carlo sampling methods to construct a sensitivity surface with respect to simultaneous changes in multiple biomarker inputs.