Analysis of EEG signals to detect hypoglycaemia

• US 10,327,656 B2
• Filed: 11/23/2011
• Issued: 06/25/2019
• Est. Priority Date: 11/26/2010
• Status: Active Grant

• Current Assignee: Uneeg Medical A/S
• Original Assignee: Uneeg Medical A/S
• Inventors: Madsen, Rasmus Elsborg, Jensen, Rasmus Stig
• Primary Examiner(s): Natnithithadha, Navin
• Assistant Examiner(s): Portillo, Jairo H
• Application Number:
  US13/989,304
• Publication Number:
  US 20130274580A1
• Time in Patent Office: 2,771 Days
• Field of Search: None
• CPC Class Codes: A61B 5/14532 for measuring glucose, e.g. by tissue impedance measurement
• CPC Class Codes: A61B 5/291 for electroencephalography [EEG]
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/374 Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G06F 2218/00 Aspects of pattern recognition specially adapted for signal processing
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

Apparatus for detecting hypoglycemia or impending hypoglycemia by analysis of an EEG comprises at least one EEG measuring electrode (10) for gathering an EEG signal and a computer (12) for receiving said EEG signals programmed to obtain a plurality of signal components each comprising a different band of frequencies, obtain a measure of the varying intensity of each said component, obtain a long time estimate of the mean of each intensity measure, obtain a long time estimate of the variability of each intensity measure, normalise each intensity measure e.g. by a subtracting from the intensity measure the long time estimate of the mean and dividing the result by the long time estimate of the variability so as to generate from each band a normalised feature, use machine analysis of the normalised features to obtain a varying cost function, classify values of the cost function according to the probability of the cost function being indicative of hypoglycemia, integrate the probabilities obtained during a selected time period, and determine that the EEG signals are indicative of hypoglycemia being present or being impending based on said integration.

Claims

1. A computer based method for detecting hypoglycaemia or impending hypoglycaemia by analysis of an EEG from a wearable EEG monitor, said method comprising:

measuring an EEG signal from a person carrying said EEG monitor;

inputting said EEG signal to a computer;

in said computer, obtaining from said EEG signal a plurality of components thereof, each component comprising a different band of frequencies, and obtaining a measure of the varying intensity of each said component;

obtaining a long term estimate of the mean of each said intensity measure and obtaining a long term estimate of the variability of each said intensity measure;

normalising each said intensity measure by a process arithmetically equivalent to subtracting from the intensity measure the long term estimate of the mean and dividing the result by the long term estimate of the variability so as to generate from each frequency band a normalised feature, using machine analysis of the normalised features to obtain a time-varying hypoglycaemia cost function;

classifying values of said hypoglycaemia cost function according to a probability of the values of said hypoglycaemia cost function being indicative of hypoglycaemia;

integrating the probabilities obtained during a selected time period comprising a plurality of time segments;

determining in said computer that the EEG signals are indicative of hypoglycaemia being present or being impending based on said integration; and

in response to said determining step, providing an output notification of present or impending hypoglycaemia,

wherein the method further comprises

detecting artefact time segments from said plurality of time segments of said EEG, which contain signal contaminating artefacts, by obtaining a sum of a linear or non-linear function of the normalised features using a pre-established set of weighting coefficients to thereby obtain a time-varying artefact detection cost function, and classifying values of each said artefact detection cost function according to the probability of values of said artefact detection cost function being indicative of one of said artefacts; and

excluding said artefact time segments from generating events to be included in said integration.

2. A method as claimed in claim 1, wherein said hypoglycaemia cost function is obtained as a sum of a linear or non-linear function of the normalised features using a pre-established set of weighting coefficients.

3. A method as claimed in claim 1, wherein the values of said hypoglycaemia cost function are classified as being an event indicative of hypoglycaemia or as being not an event indicative of hypoglycaemia, and said integration of probabilities is carried out by integrating the number of events detected during said selected time period.

4. A method as claimed in claim 1, further comprising estimating a long term mean of the hypoglycaemia cost function, estimating a long term variability of the hypoglycaemia cost function, and normalising the hypoglycaemia cost function by a process arithmetically equivalent to subtracting from the hypoglycaemia cost function the long term estimate of the mean of the hypoglycaemia cost function and dividing the result by the long term estimate of the variability of the hypoglycaemia cost function so as to generate the normalised hypoglycaemia cost function and wherein it is said normalised hypoglycaemia cost function which is classified in said classifying step.

5. A method as claimed in claim 1, further comprising estimating a long term mean of the artefact detection cost function, estimating a long term variability of the artefact detection cost function, and normalising the artefact detection cost function by a process arithmetically equivalent to subtracting from the artefact detection cost function the long term estimate of the mean of the artefact cost function and dividing the result by the long term estimate of the variability of the artefact cost function so as to generate the normalised artefact detection cost function and wherein it is said normalised artefact detection cost function which is classified in said classifying step.

6. A method as claimed in claim 1, wherein said measure of the intensities of said components is obtained for each of said plurality of time segments.

7. A method as claimed in claim 6, wherein said values of said hypoglycaemia cost function which are classified are values thereof for each of said plurality of time segments.

8. A method as claimed in claim 7, wherein said integration is carried out over a selected number of preceding time segments which together constitute said selected time period.

9. A method as claimed in claim 1, wherein a previous session estimate of the mean and variability is used at the start of a current session and is progressively updated as the current session progresses, and wherein a device used to implement the method has a nighttime and daytime mode, previous session estimates from a night time session are used to normalise intensity measures when the device is in the night time mode, and previous session estimates from a daytime session are used to normalise intensity measures when the device is in the daytime mode.

10. A method as claimed in claim 9, wherein a first set of pre-established weighting parameters is used when the system is in the night-time mode and a second set of pre-established weighting parameters is used when the system is in the daytime mode.

11. A method as claimed in claim 9, wherein the first set of weighting coefficients is determined by training on data collected during a night-time session and the second set of weighting coefficients is determined by training on data collected during a daytime session.

12. A method as claimed in claim 1, wherein said computer comprises a first processing unit coupled to receive said EEG signal and perform initial signal processing, and a second processing unit receiving processed signals transmitted from said first processing unit and performing further signal processing including said step of determining that the EEG signals are indicative of hypoglycaemia being present or impending.

13. A method as claimed in claim 12, wherein said first processing unit is an implanted unit.

14. A computer programmed to accept an EEG signal as an input and to perform thereon the steps of:

obtaining from said EEG signal a plurality of components thereof each comprising a different band of frequencies;

obtaining a measure of the varying intensity of each said component, obtaining a long term estimate of the mean of each said intensity measure and obtaining a long term estimate of the variability of each said intensity measure;

normalising each said intensity measure by a process arithmetically equivalent to subtracting from the intensity measure the long term estimate of the mean and dividing the result by the long term estimate of the variability so as to generate from each frequency band a normalised feature;

using machine analysis of the normalised features to obtain a time-varying hypoglycaemia cost function;

classifying values of said hypoglycaemia cost function according to a probability of the values of said hypoglycaemia cost function being indicative of hypoglycaemia;

integrating the probabilities obtained during a selected time period comprising a plurality of time segments,

determining in said computer that the EEG signals are indicative of hypoglycaemia being present or being impending based on said integration;

in response to said determining step, providing an output notification of present or impending hypoglycaemia, wherein a previous session estimate of the mean and variability is used at the start of a current session and is progressively updated as the current session progresses;

detecting artefact time segments, from said plurality of time segments of said EEG, which contain signal contaminating artefacts by obtaining a sum of a linear or non-linear function of the normalised features using a pre-established set of weighting coefficients to thereby obtain a time-varying artefact detection cost function, and classifying values of each said artefact detection cost function according to the probability of values of said artefact detection cost function being indicative of one of said artefacts; and

excluding said artefact time segments from generating events to be included in said integration.

15. A computer as claimed in claim 14, wherein said computer comprises a first processing unit coupled to receive said EEG signal and perform initial signal processing, and a second processing unit receiving processed signals transmitted from said first processing unit and performing further signal processing including said step of determining that the EEG signals are indicative of hypoglycaemia being present or impending.

16. A computer as claimed in claim 15, wherein said first processing unit is an implanted unit.

17. A non-transitory computer-readable medium carrying thereon a machine instruction set containing instructions for causing a compatible computer to carry out the step of receiving as an input an EEG signal and to perform thereon the steps of:

obtaining from said EEG signal a plurality of components thereof each comprising a different band of frequencies;

obtaining a measure of the varying intensity of each said component, obtaining a long term estimate of the mean of each said intensity measure and obtaining a long term estimate of the variability of each said intensity measure;

normalising each said intensity measure by a process arithmetically equivalent to subtracting from the intensity measure the long term estimate of the mean and dividing the result by the long term estimate of the variability so as to generate from each frequency band a normalised feature;

using machine analysis of the normalised features to obtain a time-varying hypoglycaemia cost function;

classifying values of said hypoglycaemia cost function according to a probability of the values of said hypoglycaemia cost function being indicative of hypoglycaemia;

integrating the probabilities obtained during a selected time period comprising a plurality of time segments, and determining in said computer that the EEG signals are indicative of hypoglycaemia being present or being impending based on said integration;

in response to said determining step, providing an output notification of present or impending hypoglycaemia, wherein a previous session estimate of the mean and variability is used at the start of a current session and is progressively updated as the current session progresses;

detecting artefact time segments, from said plurality of time segments of said EEG, which contain signal contaminating artefacts by obtaining a sum of a linear or non-linear function of the normalised features using a pre-established set of weighting coefficients to thereby obtain a time-varying artefact detection cost function, and classifying values of each said artefact detection cost function according to the probability of values of said artefact detection cost function being indicative of one of said artefacts; and

excluding said artefact time segments from generating events to be included in said integration.

18. A wearable apparatus for detecting hypoglycaemia or impending hypoglycaemia by analysis of an EEG, said apparatus comprising one or more EEG measuring electrodes for gathering an EEG signal, a computer for receiving said EEG signals, said computer being programmed so as to carry out the steps of:

obtaining from said EEG signal a plurality of components thereof, each component comprising a different band of frequencies;

obtaining a measure of the varying intensity of each said component, obtaining a long term estimate of the mean of each said intensity measure and obtaining a long term estimate of the variability of each said intensity measure;

normalising each said intensity measure by a process arithmetically equivalent to subtracting from the intensity measure the long term estimate of the mean and dividing the result by the long term estimate of the variability so as to generate from each frequency band a normalised feature;

using machine analysis of the normalised features to obtain a time-varying hypoglycaemia cost function;

classifying values of said hypoglycaemia cost function according to a probability of the values of said hypoglycaemia cost function being indicative of hypoglycaemia;

integrating the probabilities obtained during a selected time period comprising a plurality of time segments,

determining in said computer that the EEG signals are indicative of hypoglycaemia being present or being impending based on said integration;

in response to said determining step, providing an output notification of present or impending hypoglycaemia, wherein a previous session estimate of the mean and variability is used at the start of a current session and is progressively updated as the current session progresses;

detecting artefact time segments, from said pluraliy of time segments of said EEG, which contain signal contaminating artefacts confusable with hypoglycaemia patterns and excluding of a linear or non-linear function of the normalised features using a pre-established set of weighting coefficients to thereby obtain a time-varying artefact detection cost function, and classifying values of each said artefact detection cost function according to the probability of values of said artefact detection cost function being indicative of one of said artefacts; and

excluding said artefact time segments from generating events to be included in said integration.

19. An apparatus as claimed in claim 18, wherein said computer comprises a first processing unit coupled to receive said EEG signals and perform initial signal processing, and a second processing unit receiving processed signals transmitted from said first processing unit and performing further signal processing including said step of determining that the EEG signals are indicative of hypoglycaemia being present or impending.

20. An apparatus as claimed in claim 19, wherein said first processing unit is an implanted unit.