Analysis of eeg signals to detect hypoglycaemia

• US 20090287107A1
• Filed: 06/07/2007
• Published: 11/19/2009
• Est. Priority Date: 06/15/2006
• Status: Active Grant

• Current Assignee: Hypo-Safe A/S
• Original Assignee: Hypo-Safe A/S
• Inventors: Madsen, Rasmus Elsborg, Beck-Nielsen, Henning
• Granted Patent: US 8,298,140 B2
• Time in Patent Office: Days
• US Class Current: 600/544
• CPC Class Codes: A61B 5/14532 for measuring glucose, e.g. by tissue impedance measurement
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/37 Intracranial electroencephalography [IC-EEG], e.g. electrocorticography [ECoG]
• CPC Class Codes: A61B 5/372 Analysis of electroencephalograms
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: A61B 5/7267 involving training the classification device

Abstract

Features indicative of hypoglycaemia in EEG signals are detected by: —dividing EEG signals into a sequence of time segments, —for each time segment determining whether a pattern of EEG signals is present which is indicative of hypoglycaemia and, where a pattern of EEG signals indicative of hypoglycaemia is determined to be present in a time segment, recording this as an event, —integrating the number of events recorded during a selected number of preceding time segments which together constitute a selected time period, optionally in a time weighted manner, and—determining that the EEG signals are indicative that hypoglycaemia is present based on said integration when the said integrated number of events exceeds a preset threshold number and/or when there exists a threshold level of matching between a curve of said integration over time and a previously established ideal model of said curve indicative of hypoglycaemia.

Claims

1. A method for the analysis of EEG signals to detect features therein which are indicative of hypoglycaemia comprising:

dividing EEG signals into a sequence of time segments,

for each time segment determining whether a pattern of EEG signals is present which is indicative of hypoglycaemia and, where a pattern of EEG signals indicative of hypoglycaemia is determined to be present in a time segment, recording this as an event,

integrating the number of events recorded during a selected number of preceding time segments which together constitute a selected time period, and

determining that the EEG signals are indicative that hypoglycaemia is present based on said integration.

2. A method as claimed in claim 1, wherein it is determined that the EEG signals are indicative that hypoglycaemia is present when the said integrated number of events exceeds a preset threshold number and/or when there exists a threshold level of matching between a curve of said integration over time and a previously established ideal model of said curve indicative of hypoglycaemia.

3. A method as claimed in claim 1, wherein said integration is performed as a weighted integration in which events detected in time segments further back in time are given a lesser weighting than events detected in more recent time segments.

4. A method as claimed in claim 3, wherein said weighted integration is performed using a linear weighting function or a sine-curve weighting function.

5. A method as claimed in claim 1, wherein determination of whether a pattern of EEG signals is present in a time segment which is indicative of hypoglycaemia is performed by applying to said signals a previously trained Bayes classifier, a support vector machine (SVM), a relevance vector machine (RVM), a Gaussian process classifier, a classifier based on Fisher'"'"'s discriminant, a boosted classifier, a naï

ve Bayes classifier, a K-nearest neighbour classifier, a Binary Decision Tree, a Parzen Window classifier, or a neural network.

6. A method as claimed in claim 5, wherein said determination is carried out using a Bayes gaussian classifier.

7. A method as claimed in claim 6, wherein during a period of monitoring EEG signals there is carried out a re-estimation of the mean and covariance parameters of the Gaussian model over time.

8. A method as claimed in claim 7, wherein said classifier is a quadratic Bayes Gaussian classifier.

9. A method as claimed in claim 5, wherein a previously trained classifier is applied to said signals which is a classifier trained by supervised learning.

10. A method as claimed in claim 1, wherein raw EEG signals are subjected to spectrum analysis to extract therefrom power averages in at least four frequency bands upon which is based the said determination of whether a pattern of EEG signals is present which is indicative of hypoglycaemia in a time segment of said signals.

11. A method as claimed in claim 10, wherein said at least five frequency bands cover frequencies from 2 Hz to 32 Hz.

12. A method as claimed in claim 11, wherein said frequency bands include bands of approximately (2 Hz-5 Hz), (5 Hz-8 Hz), (8 Hz-11 Hz), (11 Hz-14 Hz), and (14 Hz-32 Hz), wherein all of the boundaries of said frequency bands are variable by up to 20%.

13. A method as claimed in claim 10, wherein prior to said spectrum analysis, artefact or noise elements in said EEG signals are identified and eliminated.

14. A method as claimed in claim 1, wherein said EEG signals have been obtained from one or more electrodes positioned within the area bounded by the FC3-FC1-C1-CP1-P1-PO3-P5-CP5-C5 standard EEG electrode positions and/or in an equivalent area around C4.

15. A method as claimed in claim 14, wherein the electrodes are positioned within the area bounded by the FC3-FC1-C1-CP1-CP3-C5 positions and/or in the equivalent area around C4.

16. A method as claimed in claim 14, wherein the electrodes are positioned within the area bounded by the FC3-C1-CP3-C5 positions and/or in the equivalent area around C4.

17. A method as claimed in claim 14, wherein said EEG signals have been obtained from one or more electrodes positioned in at least approximately the C3 and/or C4 or P3 and/or P4 standard EEG electrode positions.

18. A method as claimed in claim 14, wherein said signals are measured with reference to signals obtained from a reference electrode positioned approximately on the frontal-occipital mid-line.

19. A method as claimed in claim 18, wherein said reference electrode is in approximately the Cz or Pz standard EEG electrode position.

20. A method as claimed in claim 1, wherein the time period is adjusted to optimise the sensitivity and specificity of the determination that the obtained signals are indicative of hypoglycaemia for an individual from whom the signals originate.

21. A method as claimed in claim 1, wherein EEG signals are determined to be indicative of hypoglycaemia by establishing that a threshold level of matching between the curve of said integration over time and a previously established ideal model of said curve indicative of hypoglycaemia exists, wherein said matching is carried out by a mathematical convolution of the measured and ideal curves.

22. A method as claimed in claim 1, further comprising activating an alarm upon determination that the EEG signals are indicative of hypoglycaemia.

23. Apparatus for use in a method as claimed in claim 1 and comprising pre-programmed computation means comprising inputs for receiving EEG signals from EEG electrodes, and means for

dividing said EEG signals into a sequence of time segments,

for each time segment determining whether a pattern of EEG signals is present which is indicative of hypoglycaemia and, where a pattern of EEG signals indicative of hypoglycaemia is determined to be present in a time segment, recording this as an event,

integrating the number of events recorded during a selected number of preceding time segments which together constitute a selected time period,

determining that the EEG signals are indicative that hypoglycaemia is present when the said integrated number of events exceeds a preset threshold number, and—

providing an output indicating that said threshold number has been exceeded.

24. Apparatus as claimed in claim 23, wherein said output takes the form of an alarm.

25. Apparatus as claimed in claim 23, further comprising one or more electrodes for positioning on a user to receive EEG signals.