Device and method for determining sleep profiles

• US 6,272,378 B1
• Filed: 05/20/1999
• Issued: 08/07/2001
• Est. Priority Date: 11/21/1996
• Status: Expired due to Fees

• Current Assignee: 2RCW GmbH
• Original Assignee: 2RCW GmbH
• Inventors: Baumgart-Schmitt, Rudolf
• Primary Examiner(s): Nasser, Robert L.
• Application Number:
  US09/308,514
• Time in Patent Office: 810 Days
• Field of Search: 600/300, 600/544, 600/545
• 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 5/4812 Detecting sleep stages or cycles
• CPC Class Codes: A61B 5/7232 involving compression of the physiological signal, e.g. to extend the signal recording period
• 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: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

The invention relates to a device and a method for determining sleep profiles. The aim of the invention is to develop a device and a method which automatically generate a sleep stage classification with a grading of approximately 85% (measured according to the crosscorrelation function between automatically and manually generated sleep profiles) with negligible discomfort to the sleeper caused by additional technical equipment in his or her ordinary environment. The inventive device is characterized in that an electrode strip with a preamplifier (active electrode) working on the basis of a single frontal EEG channel is placed symmetrical to the nose root and is connected to a measuring and analysis unit controlled by a microprocessor and working autonomously. The method is characterized in that the EEG signal is compressed according to characteristics, stored and transmitted to a computer after this preprocessing, and classification according to sleep stages occurs in the computer by means of a population of neuronal networks.

Claims

1. A method of determining sleep profiles, comprising the steps of:

filtering an analog signal electroencephalographically detected and tapped off on a patient in the region of the forehead symmetrically to the nose root by using three sensors;

amplifying and digitizing said signal for further processing, and subsequently compressed according to characteristics;

storing and transmitting said signal to a computer for classifying said signal according to sleep stages by use of populations of neural networks topologically optimized by genetic and evolutionary algorithms;

visualizing said signals and profiles, to enable interactive processing.

2. The method of determining sleep profiles according to claim 1, wherein for the step of digitizing a sample rate of 128 values per second has been chosen at a resolution minimum of at least 10 bits.

3. The method of determining sleep profiles according to claim 1, wherein the method comprises Fourier transformation of the EEG-signal analogously amplified and filtered and the method further comprising the step of extracting the following characteristics:

Accumulated power density in the range of 1 to 4 Hz based on the overall power density (characteristic m1);

Accumulated power density in the range of 5 to 7 Hz based on the overall power density (characteristic m2);

Accumulated power density in the range of 8 to 11 Hz based on the overall power density (characteristic m3);

Accumulated power density in the range of 12 to 14 Hz based on the overall power density (characteristic M4);

Accumulated power density in the range of 15 to 30 Hz based on the overall power density (characteristic m5);

Accumulated power density in the range of 31 to 63 Hz based on the overall power density (characteristic m6);

Frequency at 25% of the overall power density (characteristic m7);

Frequency at 50% of the overall power density (characteristic m8);

Frequency at 75% of the overall power density (characteristic m9);

Frequency of the maximal power density value in the range of 1 to 4 Hz (characteristic m10);

Frequency of the maximal power density value in the range of 8 to 14 Hz (characteristic m11);

Frequency of the maximal power density value in the range of 21 to 30 Hz (characteristic m12);

Accumulated power density in the range of 50 to 60 Hz based on the overall power density (characteristic m13); and

Number of the epoch based on the total number of epochs (characteristic m14), which consists of populations of 8 to 30 instructed neural networks and rules including the context of the epochs in the classification.

4. The method of determining sleep profiles according to claim 3, wherein for enhancing generalization capabilities of the networks, a standardization is carried out, whereby the range of each characteristic is transformed to be represented by the range between −

2.0 and +2.0 and offered to the network as input data.

5. The method of determining sleep profiles according to claim 3, wherein instructed networks perform the stage allocations to each epoch.

6. The method of determining sleep profiles according to claim 1, wherein three populations of n=8 networks each are simultaneously employed, whereby one network in each case was instructed with the data of one night.

7. The method of determining sleep profiles according to claim 6, wherein said classification is first carried out separately with each network population, whereby the network populations differ from each other in that they were instructed on the basis of 12 (characteristics m1 to m12), 13 (characteristics m1 to m13) and 14 characteristics respectively, so that the networks of population 1 have 12 input units, the networks of population 2 have 13 input units and the networks of population 3 have 14 input units, wherein the number of output units remains fixed at 7 for all networks in accordance with the number of classes.

8. The method of determining sleep profiles according to claim 1, wherein an output unit with maximal excitation is determined for each network.

9. The method of determining sleep profiles according to claim 1, wherein reproduction of the sleep stages takes place in 7 natural numbers and that the median over all network-specific decisions supplies the respective classes to which the respective epoch is classified.

10. The method for determining sleep profiles according to claim 9, wherein a robust classifier is produced from synergy effects of three populations of said neural networks in that the median of the three population decisions is formed.

11. The method for determining sleep profiles according to claim 1, wherein context rules are employed, such context rules enhancing the sleep profiles via smoothing algorithms.

12. An arrangement for determining sleep profiles said arrangement adapted to be placed on a patient'"'"'s forehead, comprising:

an electrode strip (1) having a pre-amplifier placed on the forehead of the patient, said electrode strip operating based on one single frontal EEG-channel;

an autonomously operating microprocessor-controlled measuring and analysis unit connected to said electrode strip and supplied with energy via a battery, said measuring and analysis unit comprising;

a filter bank;

a final amplifier;

,

an A/D-converter;

a microcontroller with internal system software memory; and

an operating keyboard and display; and

a potential-free serial PC-interface arranged on said measuring and analysis unit, via which interface said measuring and analysis unit is connected to a commercially available PC with special software for adaptive classification by populations of neural networks for visualizing the electric potentials of the brain as well as sleep and wake profiles.

13. The arrangement for determining sleep profiles according to claim 12, wherein when said electrode strip (1) with said preamplifier is arranged on the forehead symmetrically with the root of the nose, solely said information tapped off as potential difference in the region of the forehead of the patient is processed further for determining sleep and wake profiles.

14. The arrangement for determining sleep profiles according to claim 12, further comprising a separate unit for storing original electric potentials of the brain is connected to said electrode strip (1) in an on-line mode.

15. The arrangement for determining sleep profiles according to claim 12, wherein said preamplifier is arranged on the center electrode of the electrode strip, so that the EEG-signal as an analog signal is already subjected near its lead site to amplification and the signal source is so subjected mainly to an impedance conversion.

16. The arrangement for determining sleep profiles according to claim 12, wherein said filter bank consists of a 50-Hz filter, a low-pass filter and a high-pass filter, said filter switched together supplying a bandpass filter with f_u=0.5 Hz and f_o=64 Hz.

17. The arrangement for determining sleep profiles according to claim 12, further comprising at least a 10-bit A/D-converter employed in association with an 8-bit microcontroller for further processing the signal.