System and method for processing patient polysomnograph data utilizing multiple neural network processing

• US 6,805,668 B1
• Filed: 06/26/2002
• Issued: 10/19/2004
• Est. Priority Date: 06/26/2001
• Status: Active Grant

• Current Assignee: Cadwell Industries Incorporated
• Original Assignee: Cadwell Industries Incorporated
• Inventors: Cadwell, John A.
• Primary Examiner(s): Jones, Mary Beth
• Assistant Examiner(s): ASTORINO, MICHAEL C
• Application Number:
  US10/184,100
• Time in Patent Office: 846 Days
• Field of Search: 600/300-301, 600/544-545, 600/529, 128/920, 128/925, 706/924
• US Class Current: 600/300
• CPC Class Codes: A61B 5/16 Devices for psychotechnics using teaching or educational appliances G09B1/00 - G09B7/00; Testing reaction times ; Devices for evaluating the psychological state
• CPC Class Codes: A61B 5/4812 Detecting sleep stages or cycles
• CPC Class Codes: A61B 5/7257 using Fourier transforms
• CPC Class Codes: A61B 5/7267 involving training the classification device
• 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: Y10S 128/925 Neural network
• CPC Class Codes: Y10S 706/924 Medical

Abstract

A system and method for processing patient polysomnograph data are provided. An abstractor obtains raw patient polysomnograph data and generates a subset of the data to include selected factors, including data clusters. The subset of the patient polysomnograph data is transferred to two or more neural networks that process the data and generate sleep classification data. An integrator obtains the sleep classification data from the two or more neural networks by integrating the sleep classification data from each neural network. A cumulative sleep stage score is generated including confidence values and accuracy estimations for review.

Claims

1. A system for processing a set of patient polysomnograph data, the system comprising:

two or more distinct neural networks, each neural network obtaining the set of patient polysomnograph data and generating a sleep classification data therefrom; and

an interpreter obtaining the sleep classification data generated by the two or more neural networks and generating a cumulative sleep stage score.

2. The system as recited in claim 1 further comprising an abstractor operable to obtain the patient polysomnograph data and generate a subset of the patient polysomnograph data.

3. The system as recited in claim 2, wherein the subset of the patient polysomnograph data includes at least 24 factors.

4. The system as recited in claim 3, wherein the factors include a selection from the group consisting of a fast fourier transform in a delta band, a fast fourier transform power in a beta band, fast fourier transform power in an alpha band, a ratio of alpha band power to delta band power, a total power in the delta, alpha and beta band, a number of alpha wave segments, a total area of alpha wave segments, a percentage of pages comprised of K-complexes in the alpha band, a number of K-complexes per page, a percentage of pages comprised of K-complexes in the delta band, a number of spindles per page, a total area of spindle segments, a percentage of pages comprising spindles, a number of position changes per page, a number of eye movements both in and out of phase per page, a percentage of pages with eye movements, a number of REM movements per page, a ratio of eye movements in a first half versus a second half of page, a measure of in-phase versus out-of-phase activity, a number of chin movements per page, a total area of chin activity, a percentage of page with chin activity, a ratio of chin activity in a first half versus a second half per page, and a total area of chin/EEG noise.

5. The system as recited in claim 2, wherein the abstractor is further operable to generate at least one patient data cluster, the patient data cluster including a grouping of two or more patient data factors.

6. The system as recited in claim 5, wherein the patient data cluster includes a ratio of eye movements in a first half versus a second half of a page.

7. The method recited in claim 5, wherein the patient data cluster includes a ratio of chin activity in a first half versus a second half of a page.

8. The system as recited in claim 1, wherein the cumulative sleep stage score is generated by integrating sleep classification data from a first neural network with sleep classification data obtained from a second neural network.

9. The system as recited in claim 1, wherein the interpreter is further operable to generate a confidence value for the cumulative sleep stage score.

10. The system as recited in claim 9, wherein the confidence value is embodied as the quotient of a numerical value corresponding to a selected sleep stage score divided by the sum of numerical values for every possible sleep stage score.

11. The system as recited in claim 1, wherein the interpreter is further operable to generate a ranked list of patient data factors relied upon to generate the cumulative sleep stage score.

12. A method for processing a set of patient polysomnograph data, the method comprising:

obtaining the patient polysomnograph data;

obtaining sleep classification data corresponding to the processing of the set of patient polysomnograph data from a first neural network;

obtaining sleep classification data corresponding to the processing of the set of patient polysomnograph data from a second neural network, wherein the second neural network is distinct from said first neural network;

integrating the sleep classification data from the first and second neural network to generate a cumulative sleep stage score.

13. The method as recited in claim 12 further comprising abstracting the patient polysomnograph data to generate a subset of the patient polysomnograph data prior to obtaining the sleep stage score from the first and second neural networks.

14. The method as recited in claim 13, wherein the subset of the patient polysomnograph data includes at least 24 factors.

15. The method as recited in claim 12, wherein the factors include a selection from the group consisting of a fast fourier transform in a delta band, a fast fourier transform power in a beta band, fast fourier transform power in an alpha band, a ratio of alpha band power to delta band power, a total power in the delta, alpha and beta band, a number of alpha wave segments, a total area of alpha wave segments, a percentage of pages comprised of K-complexes in the alpha band, a number of K-complexes per page, a percentage of pages comprised of K-complexes in the delta band, a number of spindles per page, a total area of spindle segments, a percentage of pages comprising spindles, a number of position changes per page, a number of eye movements both in and out of phase per page, a percentage of pages with eye movements, a number of REM movements per page, a ratio of eye movements in a first half versus a second half of page, a measure of in-phase versus out-of-phase activity, a number of chin movements per page, a total area of chin activity, a percentage of page with chin activity, a ratio of chin activity in a first half versus a second half per page, and a total area of chin/EEG noise.

16. The method as recited in claim 12 further comprising generating at least one patient data cluster, the patient data cluster including a grouping of two or more patient data factors.

17. The method as recited in claim 16, wherein the patient data clusters includes a ratio of eye movements in a first half versus a second half of a page.

18. The method as recited in claim 16, wherein the patient data cluster includes a ratio of chin activity in a first half versus a second half of a page.

19. The method as recited in claim 12 further comprising generating a confidence value for the cumulative sleep stage score.

20. Method as recited in claim 19, wherein the confidence value is embodied as the quotient of a numerical value corresponding to a sleep stage score divided by the sum of numerical values for each possible sleep stage score.

21. The method as recited in claim 12 further comprising generating a ranked list of patient data factors relied upon to generate the cumulative sleep stage score.

22. A computer-readable medium having computer-implementable instructions for performing the method recited in any one of claims 12-21.

23. A computer system having a processor, an operating environment, and a memory, the computer system operable to perform the method recited in any of claims 12-21.

24. A system for processing patient polysomnograph data, the system comprising:

an abstractor operable to obtain obtaining patient polysomnograph data and generate generating a subset of patient polysomnograph data;

two or more distinct neural networks, each neural network operable to obtain obtaining the set of patient polysomnograph data and generate generating a sleep classification data therefrom; and

an interpreter operable to obtain obtaining the sleep classification data generated by the two or more neural networks and to generate generating a cumulative sleep stage score including confidence and accuracy values.

25. The system as recited in claim 24, wherein the subset of the patient polysomnograph data includes at least 24 factors.

26. The system as recited in claim 24, wherein the factors include a selection from the group consisting of a fast fourier transform in a delta band, a fast fourier transform power in a beta band, fast fourier transform power in an alpha band, a ratio of alpha band power to delta band power, a total power in the delta, alpha and beta band, a number of alpha wave segments, a total area of alpha wave segments, a percentage of pages comprised of K-complexes in the alpha band, a number of K-complexes per page, a percentage of pages comprised of K-complexes in the delta band, a number of spindles per page, a total area of spindle segments, a percentage of pages comprising spindles, a number of position changes per page, a number of eye movements both in and out of phase per page, a percentage of pages with eye movements, a number of REM movements per page, a ratio of eye movements in a first half versus a second half of page, a measure of in-phase versus out-of-phase activity, a number of chin movements per page, a total area of chin activity, a percentage of page with chin activity, a ratio of chin activity in a first half versus a second half per page, and a total area of chin/EEG noise.

27. The system as recited in claim 24, wherein the abstractor is further operable to generate at least one patient data cluster, the patient data cluster including a grouping of two or more patient data factors from the subset of patient polysomnograph data.

28. The system as recited in claim 27, wherein the patient data cluster includes a ratio of eye movements in a first half versus the second half of a page.

29. The system as recited in claim 27, wherein the cumulative sleep stage scores generated by obtaining a sleep stage score from a first neural network and adjusting the sleep stage score by a sleep stage score obtained from a second neural network.

30. The system as recited in claim 24, wherein the confidence value is embodied as a quotient of a numerical value corresponding to a selected sleep stage score divided by the sum of numerical values for every possible sleep stage score.