Automatic detection of seizures using electroencephalographic signals

• US 5,311,876 A
• Filed: 11/18/1992
• Issued: 05/17/1994
• Est. Priority Date: 11/18/1992
• Status: Expired due to Fees

• Current Assignee: Johns Hopkins University
• Original Assignee: Johns Hopkins University
• Inventors: Olsen, Dale E., Webber, W. Robert S., Lesser, Ronald P., Harris, John C., Cristion, John A.
• Primary Examiner(s): Kamm, William E.
• Application Number:
  US07/978,220
• Time in Patent Office: 545 Days
• Field of Search: 128/731
• US Class Current: 600/544
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/372 Analysis of electroencephalograms
• CPC Class Codes: A61B 5/4094 Diagnosing or monitoring seizure diseases, e.g. epilepsy
• CPC Class Codes: Y10S 128/925 Neural network

Abstract

A patient monitoring system and a method of operating same, including automatically detecting a seizure in the patient, by detecting an electrical discharge in the patient'"'"'s brain; converting the detected electrical discharge into a digital signal; inputting the digital signal into a microprocessor; detecting a seizure by dividing the digital signals into time segments, preprocessing each time segment including standardizing the signal, reducing the signal in each time segment to a feature, the feature providing information about whether a seizure is occurring using the feature from each time segment; and indicating that a seizure is occurring.

Claims

1. A method of operating a patient monitoring system wherein a seizure in the patient is automatically detected comprising the steps of:

detecting an electrical discharge in the patient'"'"'s brain;

converting the detected electrical discharge into a digital signal;

inputting the digital signal into a microprocessor;

detecting automatically a seizure using the microprocessor and the digital signal comprising the steps of;

dividing the digital signal into a plurality of time segments;

standardizing the digital signal in each time segment;

processing the standardized digital signal in each time segment to form at least one additional signal in each time segment;

reducing each signal in each time segment to a feature, the feature providing information about whether a seizure is occurring ;

standardizing each feature; and

determining whether a seizure is occurring using the standardized features from each time segment; and

indicating that a seizure is occurring.

2. The method as recited in claim 1, wherein each time segment is three seconds in length.

3. The method of operating a patient monitoring system as recited in claim 1, the electrical discharge detecting step comprising the step of attaching a plurality of electrodes to the patient.

4. The method as recited in claim 3, wherein the determining step comprises the step of converting the standardized features from each time segment into a probability that a seizure is occurring.

5. The method as recited in claim 4, wherein the standardized features converting step comprises the step of using logistic regression to convert the standardized features into a probability that a seizure is occurring.

6. The method as recited in claim 5, wherein the determining step further comprises the step of converting the probabilities for adjacent time segments into a determination that a seizure is occurring.

7. The method as recited in claim 6, wherein the determining step further comprises the step of deciding whether the probabilities for adjacent time segments from an electrode exceed a threshold value and a probability for a time segment from a second electrode also exceeds the threshold value during either of the two adjacent time segments.

8. The method of operating a patient monitoring system as recited in claim 4, the determining step further comprising the step of comparing the probability that a seizure is occurring to a threshold value.

9. The method as recited in claim 1, the processing step comprising the step of filtering the digital signal.

10. The method as recited in claim 9, wherein the processing step further comprises:

subtracting the filtered digital signal from the standardized digital signal to obtain a residual digital signal; and

differencing successive standardized digital signal points to obtain a first difference digital signal.

11. The method as recited in claim 1, wherein the reducing step comprises the step of counting in a signal of the time segment, the signal containing a plurality of extrema, the number of the extrema in the signal.

12. The method as recited in claim 1, wherein the reducing step comprises the step of calculating a percentile of an absolute value of a signal in the time segment.

13. The method as recited in claim 1, wherein the reducing step comprises the step of computing an average of the absolute value of a signal in the time segment.

14. The method as recited in claim 1, wherein the reducing step comprises the step of calculating a coefficient for an autoregressive time series model of a signal in the time segment.

15. The method as recited in claim 1, wherein the reducing step comprises the step of computing a moment for a signal in the time segment.

16. The method as recited in claim 1, wherein the reducing step comprises the step of summing an absolute change in an amplitude of a signal in the time segment.

17. The method of operating a patient monitoring system as recited in claim 1, the detected electrical discharge converting step further comprising the step of amplifying the digital signal.

18. The method of operating a patient monitoring system as recited in claim 1, wherein the determining step comprises the step of using a neural network to convert the standardized features into a determination that a seizure is occurring.

19. The method of operating a patient monitoring system as recited in claim 1, the indicating step comprising the step of displaying the signals from each time segment.

20. The method of operating a patient monitoring system as recited in claim 1, the indicating step comprising the step of sounding an alarm when a seizure is determined to be occurring.

21. The method of operating a patient monitoring system as recited in claim 1, further comprising the step of providing treatment to the patient when a seizure is determined to be occurring.

22. The method of operating a patient monitoring system as recited in claim 21, the treatment providing step comprising the step of electrically stimulating the patient to control the seizure.

23. The method of operating a patient monitoring system as recited in claim 21, the treatment providing step comprising the step of dispensing a drug to the patient to control the seizure.

24. A patient monitoring system for automatically detecting a seizure, the monitoring system comprising:

means for detecting an electrical discharge in the patient'"'"'s brain;

means for converting the detected electrical discharge into a digital signal;

a microprocessor for receiving the digital signal;

means for automatically detecting a seizure using the microprocessor and the digital signal, the automatic detection means comprising;

means for dividing the digital signal into a plurality of time segments;

means for standardizing the digital signal in each time segment;

means for processing the standardized digital signal in each time segment to form at least one additional signal in each time segment;

means for reducing each signal in each time segment to a feature, the feature providing information about whether a seizure is occurring;

means for standardizing each feature; and

means for determining whether a seizure is occurring using the standardized features from each time segment; and

means for indicating that a seizure is occurring, the indicating means being connected to the automatic detection means.

25. The patient monitoring system as recited in claim 24, the electrical discharge detecting means comprising a plurality of electrodes.

26. The patient monitoring system as recited in claim 25, the means for determining whether a seizure is occurring comprising a means for converting the standardized features from each time segment into a probability that a seizure is occurring.

27. The patient monitoring system as recited in claim 26, the means for determining whether a seizure is occurring further comprising a means for comparing the probability that a seizure is occurring to a threshold value.

28. The patient monitoring system as recited in claim 26, the feature converting means comprising means for using logistic regression to convert the standardized features into a probability that a seizure is occurring.

29. The patient monitoring system as recited in claim 28, the means for determining whether a seizure is occurring further comprising a means for converting the probabilities for adjacent time segments into a determination that a seizure is occurring.

30. The patient monitoring system as recited in claim 29, the means for determining whether a seizure is occurring further comprising a means for deciding whether the probabilities for adjacent time segments from an electrode exceed a threshold value and a probability for a time segment from a second electrode also exceeds the threshold value during either of the two adjacent time segments.

31. The patient monitoring system as recited in claim 24, the detected electrical discharge converting means further comprising an amplifier for amplifying the digital signal.

32. The patient monitoring system as recited in claim 24, the digital signal processing means comprising a filter.

33. The patient monitoring system as recited in claim 32, the digital signal processing means further comprising:

means for subtracting the filtered digital signal from the standardized digital signal to obtain a residual digital signal; and

means for differencing successive standardized digital signal points to obtain a first difference digital signal.

34. The patient monitoring system as recited in claim 24, the determining means comprising a neural network means to convert the standardized features into a determination that a seizure is occurring.

35. The patient monitoring system as recited in claim 24, the reducing means comprising a means for counting in a signal of the time segment, the signal containing a plurality of extrema, the number of the extrema in the signal.

36. The patient monitoring system as recited in claim 24, the reducing means comprising a means for calculating a percentile of an absolute value of a signal in the time segment.

37. The patient monitoring system as recited in claim 24, the reducing means comprising a means for computing an average of the absolute value of a signal in the time segment.

38. The patient monitoring system as recited in claim 24, the reducing means comprising a means for calculating a coefficient for an autoregressive time series model of a signal in the time segment.

39. The patient monitoring system as recited in claim 24, the reducing means comprising a means for computing a moment for a signal in the time segment.

40. The patient monitoring system as recited in claim 24, the reducing means comprising a means for summing an absolute change in an amplitude of a signal in the time segment.

41. The patient monitoring system as recited in claim 24, the indicating means comprising means for displaying the signals from each time segment.

42. The patient monitoring system as recited in claim 24, the indicating means comprising alarm means.

43. The patient monitoring system as recited in claim 24, further comprising a therapeutic means connected to the automatic detection means for providing treatment to the patient.

44. The patient monitoring system as recited in claim 43, the therapeutic means comprising a means for electrically stimulating the patient to control the seizure.

45. The patient monitoring system as recited in claim 43, the therapeutic means comprising a means for dispensing a drug to the patient to control the seizure.