CN-122004791-A - Sleep detection method and device based on brain electricity, program product and electronic equipment

CN122004791ACN 122004791 ACN122004791 ACN 122004791ACN-122004791-A

Abstract

The disclosure provides a sleep detection method, device, program product and electronic equipment based on brain electricity, and relates to the technical field of computers. The method comprises the steps of obtaining electroencephalogram data of a tested object, determining at least two initial feature maps according to the electroencephalogram data, processing the at least two initial feature maps by utilizing a visual model to obtain first intermediate features, performing cascade expansion convolution processing on the first intermediate features to obtain second intermediate features, and determining sleep detection results corresponding to the electroencephalogram data according to the second intermediate features. The automatic sleep detection device and method achieve automatic sleep detection, improve detection efficiency, and improve accuracy and reliability of detection results.

Inventors

LIN NAN
GAO WEIFANG
HE HAIBO
HU PENG
LI LIAN
LU QIANG
LIANG ZI
CUI LIYING
ZHANG SHAOBO
SUN HEYANG
DONG YISU

Assignees

杭州网之易创新科技有限公司
中国医学科学院北京协和医院

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (14)

1. An electroencephalogram-based sleep detection method, comprising: Acquiring brain electricity data of a tested object; Determining at least two initial feature maps according to the electroencephalogram data; processing the at least two initial feature maps by using a visual model to obtain a first intermediate feature; performing cascade expansion convolution processing on the first intermediate feature to obtain a second intermediate feature; and determining a sleep detection result corresponding to the electroencephalogram data according to the second intermediate characteristic.
2. The method of claim 1, wherein the electroencephalogram data comprises brain electrode data and ear electrode data, and wherein the preprocessing is performed after acquiring the electroencephalogram data of the subject by: determining first reference electrode data, determining unipolar lead data from the brain electrode data and the first reference electrode data; in the case of changing reference electrodes, determining second reference electrode data from the ear electrode data, determining bipolar lead data from the brain electrode data and the second reference electrode data; Combining the unipolar lead data and the bipolar lead data to obtain preprocessed electroencephalogram data; the determining at least two initial feature maps according to the electroencephalogram data comprises: And determining at least two initial feature maps according to the preprocessed electroencephalogram data.
3. The method of claim 2, wherein the brain electrode data comprises left brain electrode data, midline electrode data, right brain electrode data, the ear electrode data comprises left ear electrode data and right ear electrode data, wherein the determining a second reference electrode data from the ear electrode data in the case of changing a reference electrode, determining bipolar lead data from the brain electrode data and the second reference electrode data comprises: And for left half brain electrode data and central line electrode data, taking right ear electrode data as second reference electrode data, respectively calculating difference values of the left half brain electrode data and the central line electrode data and the right ear electrode data, for right half brain electrode data, taking left ear electrode data as second reference electrode data, and calculating difference values of the right half brain electrode data and the left ear electrode data to obtain bipolar lead data.
4. The method of claim 1, wherein the at least two initial feature maps comprise at least two of a first initial feature map, a second initial feature map, a third initial feature map, and a fourth initial feature map, wherein the determining at least two initial feature maps from the electroencephalographic data comprises at least two of: converting the electroencephalogram data into a first initial feature map by taking the number of channels of the electroencephalogram data as the height and the dimension in each channel as the width; performing dimension conversion on the first initial feature map to obtain a second initial feature map, wherein the difference between the height and the width of the second initial feature map is smaller than the difference between the height and the width of the first initial feature map; Converting the electroencephalogram data into time-frequency characteristics, and converting the time-frequency characteristics into a third initial characteristic diagram by taking the time dimension and the frequency dimension of the time-frequency characteristics as the height and the width; And acquiring an electroencephalogram corresponding to the electroencephalogram data, and determining a fourth initial feature map according to the electroencephalogram.
5. The method of claim 1, wherein processing the at least two initial feature maps using a visual model to obtain a first intermediate feature comprises: processing each initial feature map by using a visual model to obtain visual features corresponding to each initial feature map; And splicing the visual features corresponding to each initial feature map to form the first intermediate features.
6. The method of claim 1, wherein performing a cascade dilation convolution on the first intermediate feature to obtain a second intermediate feature comprises: acquiring one or more expansion parameter sequences, each expansion parameter sequence comprising an expansion rate or a plurality of progressively increasing expansion rates; Performing serial expansion convolution processing on the first intermediate feature by adopting each expansion parameter sequence to obtain a corresponding convolution result; and fusing convolution results corresponding to each expansion parameter sequence to obtain the second intermediate feature.
7. The method of claim 6, wherein the first intermediate features include first intermediate features corresponding to each initial feature map, and the fusing convolution results corresponding to each expansion parameter sequence to obtain the second intermediate features includes: And carrying out primary fusion on convolution results corresponding to different initial feature graphs and the same expansion parameter sequences, and carrying out secondary fusion on primary fusion results corresponding to different expansion parameter sequences to obtain the second intermediate feature.
8. The method according to claim 1, wherein determining a sleep detection result corresponding to the electroencephalogram data according to the second intermediate feature includes: converting the second intermediate feature into a one-dimensional feature, and adding position embedded information to form a third intermediate feature; Inputting the third intermediate feature into a transducer, and determining a fourth intermediate feature according to information output by the transducer; and processing the fourth intermediate feature by using a classification model to obtain the sleep detection result.
9. The method of claim 8, wherein converting the second intermediate feature to a one-dimensional feature and adding location embedded information to form a third intermediate feature comprises: converting the second intermediate feature into a one-dimensional feature, and adding position embedded information and a classification mark to form a third intermediate feature; The determining a fourth intermediate feature according to the information output by the transducer includes: and extracting information corresponding to the classification mark from the information output by the transducer as a fourth intermediate feature.
10. The method of claim 8, wherein the method further comprises: Acquiring sample electroencephalogram data and sleep tags corresponding to the sample electroencephalogram data; determining at least two initial feature graphs of the sample according to the electroencephalogram data of the sample, and inputting the visual model to obtain intermediate features of the first sample; Performing cascade expansion convolution processing on the first sample intermediate features to obtain second sample intermediate features; converting the second sample intermediate feature into a one-dimensional feature, and adding position embedded information to form a third sample intermediate feature; inputting the third sample intermediate feature into the converter, and determining a fourth sample intermediate feature according to information output by the converter; Processing the intermediate characteristics of the fourth sample by using the classification model to obtain a sample sleep detection result; determining a loss function value according to the sample sleep detection result and the sleep label; Parameters of at least one of the visual model, the transducer, and the classification model are updated based on the loss function values.
11. The method of claim 10, wherein the determining a loss function value based on the sample sleep detection result and the sleep label comprises: Constructing a classification loss function based on focus loss and L2 regular loss according to the sample sleep detection result and the deviation of the sleep label; Constructing a sleep state transfer loss function according to a plurality of sample sleep detection results corresponding to a plurality of sample electroencephalogram data with a time sequence relation; A loss function value is determined based on the classification loss function and the sleep state transition loss function.
12. An electroencephalogram-based sleep detection apparatus, the apparatus comprising: the electroencephalogram data acquisition module is configured to acquire electroencephalogram data of a tested object; The initial feature map determining module is configured to determine at least two initial feature maps according to the electroencephalogram data; The visual processing module is configured to process the at least two initial feature images by utilizing a visual model to obtain a first intermediate feature; the expansion convolution module is configured to perform cascade expansion convolution processing on the first intermediate feature to obtain a second intermediate feature; And the detection result determining module is configured to determine a sleep detection result corresponding to the electroencephalogram data according to the second intermediate feature.
13. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 11.
14. An electronic device, comprising: Processor, and A memory for storing executable instructions of the processor; wherein the processor is configured to perform the method of any one of claims 1 to 11 via execution of the executable instructions.

Description

Sleep detection method and device based on brain electricity, program product and electronic equipment Technical Field The present disclosure relates to the field of computer technology, and more particularly, to a sleep detection method, apparatus, program product, and electronic device based on electroencephalogram. Background The sleep detection refers to the detection of the sleep stage or the sleep state of a tested object, is a core link for evaluating the sleep quality and diagnosing sleep disorder, and the sleep detection result can be used as important health information of the tested object, so as to provide reference for diagnosis and treatment of related diseases. With the increasing attention of society to sleep health and the continuous development of medical detection technology, the demands for sleep detection and related research are increasing. Currently, sleep detection is still mainly implemented by manually analyzing electroencephalograms. Disclosure of Invention However, the sleep detection method by manually analyzing the electroencephalogram has the problems that a great amount of electroencephalogram data is generated in the sleep detection process, manual identification and analysis are time-consuming and laborious, the efficiency is low, and the requirements of large-scale screening or long-term monitoring are difficult to meet. Moreover, the manual analysis result has subjective factors, the accuracy of the manual analysis result is difficult to be ensured, and particularly, the analysis and detection result by different experts may have the problem of lower consistency, so that the reliability and comparability of the detection result are affected. In view of the above, the present disclosure provides an electroencephalogram-based sleep detection method, apparatus, program product, and electronic device. According to a first aspect of the disclosure, an electroencephalogram-based sleep detection method is provided, and the method comprises the steps of obtaining electroencephalogram data of a detected object, determining at least two initial feature graphs according to the electroencephalogram data, processing the at least two initial feature graphs by utilizing a visual model to obtain first intermediate features, performing cascade expansion convolution processing on the first intermediate features to obtain second intermediate features, and determining sleep detection results corresponding to the electroencephalogram data according to the second intermediate features. According to a second aspect of the disclosure, an electroencephalogram-based sleep detection device is provided, and the device comprises an electroencephalogram data acquisition module, an initial feature map determining module, a vision processing module, an expansion convolution module and a detection result determining module, wherein the electroencephalogram data acquisition module is configured to acquire electroencephalogram data of a detected object, the initial feature map determining module is configured to determine at least two initial feature maps according to the electroencephalogram data, the vision processing module is configured to process the at least two initial feature maps by utilizing a vision model to obtain first intermediate features, the expansion convolution module is configured to perform cascade expansion convolution processing on the first intermediate features to obtain second intermediate features, and the detection result determining module is configured to determine sleep detection results corresponding to the electroencephalogram data according to the second intermediate features. According to a third aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of the first aspect described above and possible implementations thereof. According to a fourth aspect of the present disclosure, there is provided an electronic device comprising a processor and a memory for storing executable instructions of the processor, wherein the processor is configured to perform the method of the first aspect and possible implementations thereof via execution of the executable instructions. The embodiments of the present disclosure have the following technical effects: In a first aspect, an automatic sleep detection scheme is provided, so that the problems of time and labor consumption and low efficiency of manual electroencephalogram analysis are solved, the sleep detection efficiency is improved, large-scale electroencephalogram data can be rapidly processed, the sleep detection time is shortened, and the requirements of large-scale sleep disorder screening, long-term sleep monitoring and the like are met. According to the method, the information comprehensiveness of initial features is improved, standardized feature initialization is achieved through a visual model, m