CN119700114A - Depression screening system based on sleep physiological signals - Google Patents

Depression screening system based on sleep physiological signals Download PDF

Info

Publication number
CN119700114A
CN119700114A CN202411814883.XA CN202411814883A CN119700114A CN 119700114 A CN119700114 A CN 119700114A CN 202411814883 A CN202411814883 A CN 202411814883A CN 119700114 A CN119700114 A CN 119700114A
Authority
CN
China
Prior art keywords
sleep
depression
data
layer
convolution
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202411814883.XA
Other languages
Chinese (zh)
Inventor
潘家辉
杨家豪
尧韶聪
张嘉慧
邱谦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China Normal University
Original Assignee
South China Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by South China Normal University filed Critical South China Normal University
Priority to CN202411814883.XA priority Critical patent/CN119700114A/en
Publication of CN119700114A publication Critical patent/CN119700114A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A61B5/372—Analysis of electroencephalograms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4806—Sleep evaluation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4806—Sleep evaluation
    • A61B5/4812—Detecting sleep stages or cycles
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4806—Sleep evaluation
    • A61B5/4815—Sleep quality
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A61B5/7257—Details of waveform analysis characterised by using transforms using Fourier transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • Theoretical Computer Science (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Pathology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Data Mining & Analysis (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Physiology (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Psychology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Social Psychology (AREA)
  • Hospice & Palliative Care (AREA)
  • Educational Technology (AREA)
  • Developmental Disabilities (AREA)
  • Child & Adolescent Psychology (AREA)
  • Fuzzy Systems (AREA)
  • Databases & Information Systems (AREA)

Abstract

本发明公开了基于睡眠生理信号的抑郁症筛查系统，包括数据采集模块，所述数据采集模块用于通过脑机接口设备实时采集用户的睡眠脑电数据；数据处理模块，所述数据处理模块对采集到的脑电数据进行预处理；本发明的有益效果是：本发明结合了睡眠分期技术，采用CNN‑BiLSTM混合架构算法对睡眠脑电数据进行处理与分析，该算法融合了CNN和BiLSTM的优势，能够实现对睡眠阶段的高精度划分；引入了DepNet2D模型，通过构建针对频谱信息的二维网络，深入剖析不同频段间的关联与影响，从而精准提取出抑郁症患者的特征性脑电频谱信息，基于睡眠分期的抑郁症诊断方法不仅提高了诊断准确率，还增强了其客观性和说服力，为抑郁症的精准治疗提供了有力支持。

The invention discloses a depression screening system based on sleep physiological signals, comprising a data acquisition module, wherein the data acquisition module is used to collect sleep EEG data of a user in real time through a brain-computer interface device; and a data processing module, wherein the data processing module pre-processes the collected EEG data. The beneficial effects of the invention are as follows: the invention combines the sleep staging technology, and adopts a CNN-BiLSTM hybrid architecture algorithm to process and analyze the sleep EEG data, wherein the algorithm combines the advantages of CNN and BiLSTM, and can achieve high-precision division of sleep stages; the DepNet2D model is introduced, and by constructing a two-dimensional network for spectrum information, the correlation and influence between different frequency bands are deeply analyzed, thereby accurately extracting characteristic EEG spectrum information of patients with depression, and the depression diagnosis method based on sleep staging not only improves the diagnostic accuracy, but also enhances its objectivity and persuasiveness, and provides strong support for the accurate treatment of depression.

Description

Depression screening system based on sleeping physiological signal
Technical Field
The invention belongs to the technical field of electronic information, and particularly relates to a depression screening system based on sleep physiological signals.
Background
In recent years, with the rapid development of deep learning technology, the application of the deep learning technology in the medical field is gradually wide, and particularly in the aspects of sleep stage and depression diagnosis based on brain electrical signals, the deep learning technology has great potential; in the aspect of the current research situation of the classification of the depression based on the electroencephalogram signals, researchers have remarkably progressed in the field of depression identification by carefully designing a deep learning model, acharya and the like have conducted classification research on the electroencephalogram data of depression patients and normal people by adopting a CNN model, and successfully verify the effectiveness of deep learning in depression diagnosis, however, the research is limited by smaller sample size, the generalization capability of the model is still suspected, li and the like are further on the basis, the important role of different layers of information in the electroencephalogram signals in depression identification is deeply explored by combining the CNN and a migration learning technology, a firmer theoretical basis is provided for the identification of the depression by the electroencephalogram signals, the model classification performance is not very good, kwon and the like have provided an innovative deep learning model, the preliminary screening of the depression patients is realized on the basis of an asymmetric image of the front lobe electroencephalogram, but the research is insufficient in the aspect of evaluating modeling capability, the research can only provide preliminary reference electrical signals Lu Guanming and the like (2021) has the advantages of accurately combining the aspect of a dynamic analysis with a dynamic analysis model (35) by a dynamic analysis method of a dynamic analysis, a combination of the dynamic analysis is realized by combining the development algorithm (35) and the dynamic analysis is realized by the method of the model 35, the method has the advantages that the high accuracy of 98.00% is achieved, but the cost is too high, joao Carreira et al (2018) adopts a 3D convolutional network which uses effective connection in depressive disorder (MDD), the new functions of PDC and DMN connection and the like are proposed, but the accuracy of the realization is still lack of more evaluation indexes, the accuracy is 80.90%, qayyum et al (2022) shows the combination method of the convolutional neural network and GRU, the characteristic of the convolutional neural network is superior to LSTM in short sequence learning, the accuracy is 90.45%, the method is still insufficient in time window segmentation length processing, the research of Narazaki et al (2019) is based on a convolutional neural network of 22 layers ResNet, the classifier is convenient to develop and the weight is distributed by using an algorithm, the accuracy is only 78.94%, and the research generally faces the problems of insufficient sample size, high cost, high model accuracy, high generalization capability and the like in spite of the research in deep learning application of depression diagnosis.
In order to further accurately diagnose depression, many scholars analyze the association of a plurality of brain electrical signals and depression, wherein the sleep signals are regarded as stable brain electrical signals, researchers find that the association is close by deeply exploring the relation between the sleep brain electrical signals and the depression, steigerA et al research shows that the long-term administration of antidepressant drugs can have a significant influence on the sleep brain electrical signal characteristics of depression patients, the discovery provides a solid theoretical basis for a depression diagnosis method based on the sleep brain electrical signals, ARMITAGE R et al concentrate on young female depression patients, find that the Delta wave sleep brain electrical signals of the scholars have significant differences with normal people, provide valuable clues for exploring the sleep brain electrical specificity of the depression, LEISTEDT S et al disclose that the parameter values of the depression patients in specific sleep stages are abnormal, the study on the sleep and the depression is strongly confirmed, the analysis of numerical analysis can be realized, pillai V et al (2011) show that the genetic characteristics of the sleep brain electrical information of the depression patients are analyzed, and the genetic characteristics of the depression patients are considered to be a transient depression, and the sleep phase is a rapid depression (the genetic characteristics of the depression is considered to be a rapid depression).
In conclusion, the method for diagnosing the depression based on the brain electrical signals has been researched and developed to a certain extent, however, the field still has a plurality of problems such as low accuracy, low method interpretation and the like.
Disclosure of Invention
The invention aims to provide a depression screening system based on sleep physiological signals, which improves the accuracy and objectivity of diagnosis.
The invention provides a technical scheme for realizing the aim, which comprises a depression screening system based on sleep physiological signals, comprising
The data acquisition module is used for acquiring sleeping brain electricity data of a user in real time through brain-computer interface equipment;
The data processing module is used for preprocessing the acquired electroencephalogram data, and comprises the steps of limiting the frequency range by applying a zero-phase filter and analyzing and separating independent components in the mixed signal by adopting the independent components so as to reduce noise and artifacts;
The sleep stage module comprises a CNN-BiLSTM-based model, wherein the model firstly converts preprocessed electroencephalogram data into time-frequency signals through short-time Fourier transformation and constructs a feature matrix; then, feature extraction is carried out in a time-frequency domain by using CNN, and dependence and correlation of the front and the back of the electroencephalogram signals are explored from a time-domain dimension by BiLSTM, so that automatic staging of a sleep stage is realized;
the depression screening module is used for processing the sleep electroencephalogram data of the rapid eye movement stage based on a DepNet D model, extracting frequency spectrum characteristics through a two-dimensional convolutional neural network, then carrying out high-level characteristic learning through a full-connection layer, and carrying out classification judgment by using a Softmax activation function so as to screen whether a user suffers from depression;
And the user interaction module is used for receiving the user instruction and displaying the sleep report.
As a preferable technical scheme of the invention, a user instruction is received, a sleep report is displayed, the sleep report comprises a sleep stage result and a depression screening report, and the depression screening report comprises a depression symptom degree radar chart and a depression symptom occurrence tendency pie chart of a base Yu Beike-Fan Sen mania self-evaluation scale.
As a preferable technical scheme of the invention, in the CNN-BiLSTM model, the CNN part performs feature extraction through convolution operation, batch normalization, activation function, maximum pooling and dropout layer, and the BiLSTM part processes time series data through LSTM networks in two directions so as to capture long-term time dependence and relevance.
As a preferable technical scheme of the invention, the DepNet D model comprises three convolution layers, each convolution layer is connected with a ReLU activation function and a maximum pooling layer, and the convolution layers are used for extracting and retaining significant features and carrying out high-level feature learning and classification judgment through a full connection layer.
As a preferable technical scheme of the invention, when data preprocessing is carried out, the following steps are adopted:
dividing data samples, namely dividing sleep electroencephalogram data of continuous 30s into one sleep data sample by taking frames as basic units for interpretation;
Setting sleep stage labels, namely dividing sleep periods into a waking period, an N1 period, an N2 period, an N3 period and a REM period according to an AASM sleep stage standard, and setting the corresponding sleep stage labels as 0,1,1,1,2 which correspond to the three stages of the waking period, the deep sleep period and the REM period;
And (3) filtering sleep electroencephalogram data, namely filtering electroencephalogram signals of 2 channels by using Butterworth band-pass filters with cut-off frequencies of 0.1 Hz and 50 Hz, and forming a group of data samples after filtering every 5 sections and discrimination tags to construct a feature matrix.
As a preferred technical scheme of the invention, the system further comprises a CNN-BiLSTM network structure which consists of four layers, namely a convolution layer, a pooling layer, an LSTM-based BiLSTM hidden layer and a multi-task Softmax layer, wherein the system structure has four main characteristics that the convolution layer simultaneously accommodates convolution kernels with 3 different sizes, the pooling strategy of the pooling layer is more suitable for capturing the shift invariance of time signals, the LSTM unit is proved to be capable of extracting long-term time dependence used in a plurality of sequence processing applications, the accuracy of sleep stage can be improved through the LSTM unit at BiLSTM, the extracted generalized sequence representation of EEG frame modes obtained from the first two layers is processed together with a bidirectional LSTM network, the used multi-task Softmax layer is suitable for joint classification and prediction, so that the characteristic extraction of sleep brain data and the training of a model can have more reliable results, and the probability of various sleep stage results can be finally output through a Softmax layer classifier.
As a preferable technical scheme of the invention, the DepNet D model builds a two-dimensional network model of the spectrum information by fully utilizing the spectrum information of REM sleep electroencephalogram data, and analyzes the relevance and the mutual influence between different frequency bands.
As a preferable technical scheme of the invention, the method also comprises the steps of constructing DepNet D characteristic extraction model, and the corresponding method is as follows:
The method comprises the steps of obtaining spectrum information of an original sleep electroencephalogram in an REM period, taking two channels, inputting the spectrum information into a model, using a plurality of stacked convolution layers to conduct feature extraction, respectively passing through a ReLU activation function and a maximum pooling layer after passing through the convolution layers, reducing the size of a feature map through downsampling operation and retaining the most remarkable features, conducting batch normalization, connecting a Dropout layer to each pooling layer, connecting a flat layer, flattening the features extracted by the convolution layers and the pooling layers into a one-dimensional vector, and finally passing through the full connection layer twice.
As a preferable technical scheme of the invention, a plurality of convolution layers are stacked for feature extraction, and the method has the following corresponding content that three two-dimensional convolutions are used, wherein the number of convolution kernels of the first convolution layer is 16, the number of convolution kernels of the second convolution layer is 36, the number of convolution kernels of the third convolution layer is 48, and the convolution kernels of the three convolution layers are all (3, 3).
Compared with the prior art, the invention has the beneficial effects that:
The invention combines the sleep stage technology, adopts a CNN-BiLSTM mixed architecture algorithm to process and analyze sleep electroencephalogram data, combines the advantages of CNN and BiLSTM, can realize high-precision division of sleep stages, introduces a DepNet2D model, deeply analyzes the association and influence among different frequency bands by constructing a two-dimensional network aiming at spectrum information, thereby accurately extracting the characteristic electroencephalogram spectrum information of a depression patient, and improves the diagnosis accuracy, objectivity and persuasion of the sleep stage-based depression diagnosis method and provides powerful support for the accurate treatment of depression;
The invention adopts CNN-BiLSTM mixed architecture algorithm, can realize high-precision division of sleep stage, pay attention to the electroencephalogram signal of the rapid eye movement sleep stage (REM), fully research the pathophysiology mechanism of depression, diagnose depression by analyzing sleep electroencephalogram data, has objectivity, is not influenced by subjective consciousness of a patient, and can provide more objective diagnosis basis, simultaneously, builds a two-dimensional network aiming at frequency spectrum information by introducing DepNet D model, further analyzes association and influence among different frequency bands, learns and identifies the mode from a large amount of data, thereby improving the accuracy of diagnosis, further improves the accuracy and objectivity of diagnosis, greatly reduces subjective factors in the diagnosis process by utilizing the sleep electroencephalogram data and combining with CNN-BiLSTM network and DepNet D model, provides more objective and accurate diagnosis basis, and can avoid or reduce the influence of doctor judgment and patient self-report, thereby improving the accuracy and objectivity of depression diagnosis;
By utilizing sleep electroencephalogram data, the invention combines a CNN-BiLSTM network and a DepNet D model, can provide more objective and accurate diagnosis basis, reduce the influence of subjective factors, can make more accurate results on depression symptoms earlier, and realize early intervention; the depression diagnosis system of the invention utilizes sleep brain electrical data to carry out early screening and diagnosis, has higher objectivity and accuracy, can effectively improve the treatment effect and prognosis, lighten the pain and burden of patients and improve the life quality by early finding the depression symptoms of the patients and carrying out intervention treatment;
Through analysis of the sleep electroencephalogram data of the patient, the information of the sleep quality, the sleep structure, the characteristics of brain activities during sleep and the like of the patient can be known, so that the individual difference of the patient can be better understood;
The invention is based on the analysis result of the REM sleep electroencephalogram data and the frequency spectrum information of the sleep electroencephalogram of the patient, has the advantages of higher accuracy and screening rate, is scientific and referential, enables doctors to formulate a more personalized and targeted treatment scheme, can better regulate the physiological and psychological states of the patient through the treatment scheme formulated for the specific condition of the patient, relieves the symptoms of depression, and improves the treatment effect and prognosis.
Drawings
FIG. 1 is a technical flow chart of the present invention;
FIG. 2 is a schematic diagram of a sliding time window cut sleep brain electrical data according to the present invention;
FIG. 3 is a model building diagram of the present invention;
fig. 4 is a system block diagram of a depression screening system of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
Referring to fig. 1-4, in a first embodiment of the present invention, a sleep physiological signal-based depression screening system is provided, which has the characteristics of high accuracy, high efficiency, high user friendliness, and the like, and faces the diagnosis difficulty of 3.22 million depression patients worldwide, and the system can not only realize the identification of depression, but also monitor and optimize the sleep quality of the patients in real time by analyzing sleep brain signals, so as to improve the diagnosis accuracy to realize the screening function of depression;
Due to the non-invasiveness of sleep brain electrical signals and the high time resolution, sleep classification is carried out by using a DepNet D (Depression Net Dimension) model, feature extraction and classification of REM phase brain electrical data are focused, depNet D learns the time-space dependence of sleep brain electrical data through a two-dimensional convolutional neural network, feature modes of brain activities are effectively captured, the accuracy of identifying depression spectrum features is improved, the model has high generalization capability due to the reduction of parameter quantity, the accuracy in depression screening tasks reaches 88.82%, and compared with a traditional method, the model has higher accuracy and objectivity.
The invention uses different public data sets to test the generalization performance of the model, continuously adjusts and optimizes parameters, and integrates codes at the front end and the rear end;
sleep staging based on CNN-BiLSTM
The application adopts sleep brain electrical signals as the basis for screening depression patients, sleep disorder is typical symptoms of a plurality of depression patients, wherein the latency period of REM (rapid eye movement) sleep is shortened, the duration time is prolonged and the REM density is increased are regarded as biological markers of depression;
preprocessing of sleep electroencephalogram data
The Sleep brain electrical data adopted by the application comes from a Sleep-EDF public data set which is a resource accepted by people and is widely used in the field of Sleep research, wherein the Sleep brain electrical data comprises two groups of data of different groups, one group records the Sleep brain electrical signals of normal people and the other group records the Sleep brain electrical signals of depression patients, the data are obtained by arranging electrodes in the Sleep process to record brain activities, and the application adopts the following steps when the data preprocessing is carried out:
dividing data samples, namely dividing sleep electroencephalogram data of continuous 30s into one sleep data sample by taking a frame (epoch) as a basic unit of interpretation;
Setting sleep stage labels, namely dividing sleep periods into a waking period, an N1 period, an N2 period, an N3 period and a REM period according to AASM sleep stage standards set by the American sleep medical society (AMERICAN ACADEMY of SLEEPMEDICINE, AASM), setting the corresponding sleep stage labels as '0,1,1,1,2', and corresponding to the three stages of the waking period, the deep sleep period and the REM period;
filtering sleep electroencephalogram data, namely filtering electroencephalogram signals of 2 channels by using Butterworth band-pass filters with cut-off frequencies of 0.1 Hz and 50 Hz, and forming a group of data samples after filtering every 5 segments and a discrimination label to construct a feature matrix;
Sleep staging based on CNN-BiLSTM network model
The CNN-BiLSTM network model is a framework combining a Convolutional Neural Network (CNN) and a bidirectional long-short-term memory neural network (BiLSTM) and is used for sleep automatic stage separation, the combination can effectively extract the characteristics of sleep electroencephalogram data and learn time sequence information in the sleep electroencephalogram data so as to realize accurate classification of sleep stage separation, the CNN can be well applied to the characteristic Extraction of Electroencephalogram (EEG) signals through the effective processing capacity of the CNN to non-stationary and non-linear data, and the combination optimization of the characteristics from different modes and BiLSTM layers is carried out in order to better explore the time variation among different electroencephalogram modes so as to better learn the sleep stage conversion rule and output accurate sleep stage separation results, and the combination of the CNN-BiLSTM network model can effectively extract the characteristics of the sleep data and learn the time sequence information in the sleep stage separation so as to realize more accurate sleep stage separation classification;
The CNN-BiLSTM network structure used by the system comprises four layers, namely a convolution layer, a pooling layer, an LSTM-based BiLSTM hidden layer and a multi-task Softmax layer, wherein the system structure is characterized in that the convolution layer accommodates 3 convolution kernels with different sizes at the same time, so that the characteristics of sleep brain data on different resolutions can be learned, the complexity of reconstructing the data in a data characteristic extraction process is reduced, the construction of an automatic sleep stage model is facilitated, secondly, compared with the pooling of common sub-sampling, the pooling strategy of the pooling layer is more suitable for capturing the displacement invariance of time signals, because specific characteristics of sleep brain signals possibly occur at any time position, then, the correlation and the relevance before and after a sleep period are considered, the LSTM unit is proved to extract long-term time dependence used in a plurality of sequence processing application, and therefore, the system can learn the characteristics of sleep stage by the LSTM unit in a BiLSTM module, the extraction generalized sequence of a frame mode obtained from the front two layers can be processed together with the LSTM network, and the characteristics of the two-stage EEG network can be extracted by the LSTM unit, the two-stage brain data can be further predicted by the aid of the LSTM unit, the characteristics can be more accurately stacked at the two layers, the opposite to the end of the sleep stage can be more accurate, the characteristics can be further predicted by considering the characteristics of the layers, and the soft max layer can be more suitable for the sleep stage of the sleep stage has the correlation and the correlation, and the reliability can be further predicted by the characteristics of the sleep stage mode, and the characteristics can be more predicted by the fact expected to have the characteristics and the characteristics of the reliability of the sleep stage and the sleep stage.
As shown in formula (1), for an input vector z= { Z 1,Z2,......,ZK }, wherein Z i is an activation value representing the ith moment;
As shown in formula (1), the probability of various sleep stage results is output through a Softmax classifier, wherein Z is an input vector and comprises the activation values of various classes, Z i is the activation value of the ith element in the input vector Z, e is the bottom of the natural logarithm, k is the length of the input vector Z, namely the total number of classes, Representing an exponential summation over all elements Z i in the input vector Z;
Meanwhile, to solve the problem of sleep data class imbalance, the model uses the sum of cross entropy errors of the subtasks as the classification and prediction errors of training, as shown in formula (2), wherein θ and The network is trained, so that the multi-task cross entropy error can be reduced to the greatest extent;
As shown in equation (3), where log (y i) represents the log probability of the predicted result of the network on sample i, The weight of the category c i is adjusted, so that the samples of different categories can be focused to different degrees, and the problem of unbalanced categories is solved;
As shown in formula (3), through adjusting the weight value, the samples of different categories can be focused to different degrees, so that the problem of unbalanced categories is solved, wherein L represents the value of a loss function; Representing the weight of category c i, log (y i) representing the log probability of the network for the predicted outcome of sample i, and y i representing the predicted probability of sample i.
DepNet 2D-based depression screening
In clinic, the application provides a DepNet D model to realize the function of screening depression, wherein the DepNet D model is used for constructing a two-dimensional network model of spectrum information by fully utilizing the spectrum information of REM sleep brain data, analyzing the relevance and the mutual influence of different frequency bands, comparing the brain spectrum information of normal people, and effectively extracting the brain spectrum characteristics of the depression patient to realize the function of screening depression.
Preprocessing of sleep electroencephalogram data
Firstly, we take out the original SLEEP brain electrical data in REM period to conduct data analysis and data conversion, then use a zero-phase Filter (FIR) to limit the frequency to 0.5Hz to 45Hz, which can help to reduce unnecessary noise without causing phase distortion, then use Independent Component Analysis (ICA) to separate the mixed signals into independent components, effectively extract potential brain electrical activity sources, reduce the influence of artifacts, improve the quality and accuracy of the signals, make the neural activity modes of different brain regions appear more obvious during SLEEP, then cut out and draw the filtered REM-SLEEP-EEG into time domain diagrams with a time window of 30s, and further extract the frequency spectrum information of brain electrical signals.
Construction DepNet of 2D feature extraction model
Firstly, the extracted spectrum information of the original sleep brain electricity in REM period is taken, two channels (EEG Fpz-Cz and EEG Pz-Oz) are taken, the spectrum information is input into a model, the size of an input layer is 150,150,3, and a plurality of convolution layers are stacked for feature extraction, so that a model with stronger characterization capability can be constructed; after passing through the convolution layers, the model can learn higher level feature representations by means of a combination of full-connection layers (Dense) two times and mapping these feature representations to the probability distribution of the output class, by reducing the size of the feature map and preserving the most significant features by means of a downsampling operation, then performing a batch normalization, after connecting Dropout layers (ratio=0.1) to each pooling layer in order to reduce the neural network overfit and improve the performance of the model, then connecting the layers, flattening the features extracted by the convolution layers and pooling layers into one-dimensional vectors, finally, by means of a combination of full-connection layers (Dense) two times, the model can learn the non-linear feature representations of the input data using the ReLU activation function, the second full-connection layer (Dense=tf. Soft) has 2 neurons with the expression capability when learning the feature representations, using the f function, converting the output of the neural network into a distribution representing the probability of each category;
The model parameter optimizing method has the advantages that as shown in a formula (4), for Binary classification, a Binary Cross entropy function (Binary Cross-Entropy Loss) is used, the loss function can measure the difference between probability distribution output by the model and a real label so as to guide the optimizing process of the model parameter, by using an Adam optimizer, we can dynamically adjust the learning rate according to the first moment estimation and the second moment estimation of the gradient of each parameter, compared with a traditional gradient descent algorithm, adam has faster convergence speed and better performance, and the accuracy (accuracy) is used as an evaluation index, so that the classifying performance of the model can be intuitively reflected;
For N data, where t i is the true label with a value of 0 or 1, p i is the Softmax probability for the i-th class;
For N data, wherein L represents the value of a loss function, t i is the real label of the ith sample, the value is 0 or 1, p i is the prediction probability that the ith sample belongs to the positive class, log (p i) is the logarithmic value of the prediction probability that the ith sample belongs to the positive class, log (1-p i) is the logarithmic value of the prediction probability that the ith sample belongs to the negative class, and N is the total number of samples; representing the summation of all N samples.
Example 2
Referring to fig. 1-4, a second embodiment of the present invention is shown, which is based on the previous embodiment, except that:
Because of the noise problem of the noninvasive electroencephalogram signals, in order to ensure the reliability of experimental study, the invention needs to rely on high-precision equipment of the south China brain control intelligent science and technology Co., ltd, and the invention uses an LT3 type electroencephalogram cap;
The invention utilizes the carefully selected SleepEDF public data set with actual credibility and the data collected by interaction with brain-computer interaction of the south China university and the mixed intelligent team, the sample data has actual credibility, the reliability of experimental equipment and the scientificity of the collection process can be ensured, and a more accurate depression screening model can be obtained under the training of a CNN-BiLSTM network model and a DepNet model; in addition, the invention improves the quality and reliability of the subsequent treatment by strictly screening the data, can realize the diversity of the acquired data and the rigor of the screening process, and ensures the generalization capability of the model and the credibility of the experimental result;
sleep stage module based on CNN-BiLSTM
Forming a feature matrix
Analyzing the brain electrical data obtained after preprocessing by using a short-time Fourier transform method in time-frequency analysis, converting a signal with amplitude changing along with time into a signal with time changing along with frequency, setting window size and step length through a sliding window mechanism, enabling the window to slide on a time domain signal, respectively calculating Fourier transform of each window, splicing frequency signals corresponding to different formed time windows to form a time-frequency signal, obtaining a corresponding Mel spectrogram through calculation, and classifying the brain electrical data of each 5 sections of continuous 30s into a group to form a feature matrix;
Feature extraction
① Structure of CNN-BiLSTM model
The application adopts a CNN-BiLSTM network model to extract the characteristics of the matrix, namely the network model consists of a Convolutional Neural Network (CNN) and a bidirectional long-short-time memory neural network (BiLSTM), and takes the characteristics of non-stability and nonlinearity of electroencephalogram multi-modal PSG (including EEG, EOG, EMG and the like) sleep signal data into consideration, the traditional method is difficult to capture the long-term change of a sleep stage, and the CNN model is suitable for processing network structure data, and the core idea is that local perception fields, weight sharing and downsampling are pooling layers.
The system inputs the vector obtained based on the CNN model into a bidirectional long-short time memory network BiLSTM based on bidirectional time structure modeling, processes sleep information in front and back directions through two completely independent long-short time memory networks LSTM, and extracts information features in different dimensions respectively so as to better mine the dependence and relevance of electroencephalogram signals before and after each time point;
② Feature extraction in time-frequency domain dimension by combining CNN convolutional neural network
After the feature matrix group formed by the Mel spectrogram is input into the model, the convolution operation layer carries out local association and window sliding treatment on input data, then carries out batch normalization treatment, then carries out maximum pooling on the data by applying an activation function and connecting the data to the back of each pooling layer by utilizing a dropout layer to make the data more robust, and then connects the Flatten layer to realize the transition of the data from the convolution layer to the full connection layer. Then, through a Dense one-dimensional full-connection layer, normalization and ReLU activation, and splicing of data blocks among the last different channels, transition from a convolution layer to a long-short-term memory neural network is realized;
③ Connection BiLSTM network model explores dependence and association of electroencephalogram signals from time domain dimension
In BiLSTM, the model uses a bidirectional long-short memory time network to transform the data in ② to obtain a matrix with a brand new size dimension to be used as input of the classifier;
The application dynamically adjusts the learning rate of each parameter by using a multi-classified logarithmic loss function corresponding to a Softmax classifier and optimizing the loss function by Adam, inputs the data in ③ above to a Softmax layer for three classification tasks, and respectively obtains corresponding sleep states (W, LS, SWS, REM);
DepNet 2D-based depression screening module
Preprocessing of sleep electroencephalogram data
The application of DepNet D model in depression screening depends on the comprehensive preprocessing of REM sleep brain data, by extracting data from original brain signals and applying a zero-phase Filter (FIR) to limit the frequency range (0.5 Hz to 45 Hz), the method effectively reduces noise and avoids phase distortion, then, the Independent Component Analysis (ICA) is adopted to separate independent components in the mixed signal, so that the definition of the signal is enhanced and the artifact is reduced, thereby reflecting the neural activity modes of different brain areas more accurately;
feature extraction of depression:
In the feature extraction stage, depNet D models process the extracted REM-phase electroencephalogram spectrum information by constructing a two-dimensional Convolutional Neural Network (CNN). The size of spectrum information received by an input layer is 150,150,3, the model comprises three convolution layers which respectively comprise 16, 36 and 48 convolution kernels, the convolution kernels are 3 and 3, each convolution layer is connected with a ReLU activation function and a maximum pooling layer with the size of 2 and 2 so as to extract and retain remarkable characteristics, the robustness and the overfitting prevention of the model are enhanced through batch normalization and Dropout layers, the characteristics extracted by the convolution layers and the pooling layers are flattened into one-dimensional vectors, high-level characteristic learning is carried out through two fully connected layers, the first fully connected layer comprises 64 neurons, nonlinear characteristics of input data are learned by using the ReLU activation function, and the second fully connected layer converts network output into distribution representing probability of each category through a Softmax activation function for two classification tasks, namely whether depression is judged;
Model training and evaluation
In the model training process, a binary cross entropy function is used as a loss function to measure the difference between probability distribution output by the model and a real label, the learning rate is dynamically adjusted through an Adam optimizer to achieve faster convergence speed and better performance, and the classification performance of the model directly reflects the effect of DepNet D in depression screening through accuracy evaluation. The whole set of process combines the efficient data processing and the deep learning method, so that the DepNet D model can effectively extract the brain electrical spectrum characteristics of a patient suffering from depression and provide an accurate depression screening function;
Depression screening system
When a user wears an LT3 type brain-computer cap as brain-computer interface equipment, the rear end of the connection system is connected for acquiring sleep brain-computer data, and the real-time monitoring of sleep brain-computer signals can be realized;
When a user clicks a 'finish sleeping' button, the system generates a sleep report, and when the sleep report is clicked, the system presents specific data of the sleeping situation of the user, such as the average sleeping time length of a month, the deep sleeping time length of monthly and the like, and makes a sleeping target for the user, and meanwhile, the system summarizes the sleeping record of the user, the sleeping situation of the week and the sleeping depth distribution in a chart form, so that the user is helped to intuitively know the sleeping situation of the user.
In addition, after sleep stage is carried out through a CNN-BiLSTM network, a system extracts REM sleep electroencephalogram, after electroencephalogram spectrum analysis is carried out through DepNet D, the system can learn sleep data characteristics of a user and carry out depression screening classification, the system generates a depression symptom degree radar chart according to a Bei Kela Fan Sen mania self-evaluation chart (BRMS) in clinical medicine, obtains depression tendency degree pie charts of the user based on an algorithm, and generates a depression report of the user.
While embodiments of the present invention have been shown and described in detail with reference to the foregoing detailed description, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations may be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (9)

1. Depression screening system based on sleep physiological signals is characterized by comprising
The data acquisition module is used for acquiring sleeping brain electricity data of a user in real time through brain-computer interface equipment;
The data processing module is used for preprocessing the acquired electroencephalogram data, and comprises the steps of limiting the frequency range by applying a zero-phase filter and analyzing and separating independent components in the mixed signal by adopting the independent components so as to reduce noise and artifacts;
The sleep stage module comprises a CNN-BiLSTM-based model, wherein the model firstly converts preprocessed electroencephalogram data into time-frequency signals through short-time Fourier transformation and constructs a feature matrix; then, feature extraction is carried out in a time-frequency domain by using CNN, and dependence and correlation of the front and the back of the electroencephalogram signals are explored from a time-domain dimension by BiLSTM, so that automatic staging of a sleep stage is realized;
the depression screening module is used for processing the sleep electroencephalogram data of the rapid eye movement stage based on a DepNet D model, extracting frequency spectrum characteristics through a two-dimensional convolutional neural network, then carrying out high-level characteristic learning through a full-connection layer, and carrying out classification judgment by using a Softmax activation function so as to screen whether a user suffers from depression;
And the user interaction module is used for receiving the user instruction and displaying the sleep report.
2. The system for screening depression based on sleep physiological signals according to claim 1, wherein the system for screening depression comprises a step of receiving a user instruction, displaying a sleep report including a sleep staging result and a depression screening report, wherein the depression screening report comprises a radar map of the degree of depression symptoms and a pie chart of the tendency of occurrence of depression symptoms based on the Fan Sen mania self-evaluation scale of Yu Beike.
3. The sleep physiological signal based depression screening system according to claim 1, wherein in the CNN-BiLSTM model, the CNN part performs feature extraction by convolution operation, batch normalization, activation function, max pooling and dropout layer, and the BiLSTM part processes time series data by LSTM network in two directions to capture long term time dependence and correlation.
4. The sleep physiological signal based depression screening system according to claim 1, wherein DepNet D model comprises three convolution layers, each convolution layer is followed by a ReLU activation function and a max pooling layer for extracting and preserving salient features, and high-level feature learning and classification judgment is performed through a full connection layer.
5. The sleep physiological signal based depression screening system according to claim 1, wherein the data preprocessing comprises the steps of:
dividing data samples, namely dividing sleep electroencephalogram data of continuous 30s into one sleep data sample by taking frames as basic units for interpretation;
Setting sleep stage labels, namely dividing sleep periods into a waking period, an N1 period, an N2 period, an N3 period and a REM period according to an AASM sleep stage standard, and setting the corresponding sleep stage labels as 0,1,1,1,2 which correspond to the three stages of the waking period, the deep sleep period and the REM period;
And (3) filtering sleep electroencephalogram data, namely filtering electroencephalogram signals of 2 channels by using Butterworth band-pass filters with cut-off frequencies of 0.1 Hz and 50 Hz, and forming a group of data samples after filtering every 5 sections and discrimination tags to construct a feature matrix.
6. The sleep physiological signal based depression screening system according to claim 1, further comprising a CNN-BiLSTM network structure consisting of four layers, a convolution layer, a pooling layer, an LSTM-based BiLSTM hidden layer and a multi-task Softmax layer, the architecture having four main features, the convolution layer accommodating 3 convolution kernels of different sizes simultaneously, the pooling strategy of the pooling layer being more suitable for capturing shift invariance of time signals, the LSTM unit being proven to extract long-term time dependencies used in many sequential processing applications, the accuracy of sleep stage being improved by the LSTM unit at BiLSTM module, the generalized order of extraction of EEG frame patterns obtained from the first two layers being indicative of processing with a bi-directional LSTM network, the multi-task Softmax layer being used being adapted for joint classification and prediction, enabling feature extraction of sleep electroencephalographic data and training of models with more reliable results, and probabilities of various sleep stage results being finally output by the Softmax layer classifier.
7. The sleep physiological signal-based depression screening system according to claim 1, wherein DepNet D model is used for constructing a two-dimensional network model of spectrum information by fully utilizing spectrum information of REM sleep electroencephalogram data, and analyzing relevance and mutual influence among different frequency bands.
8. The sleep physiological signal based depression screening system according to claim 7, further comprising constructing DepNet D feature extraction model, corresponding to the method as follows:
The method comprises the steps of obtaining spectrum information of an original sleep electroencephalogram in an REM period, taking two channels, inputting the spectrum information into a model, using a plurality of stacked convolution layers to conduct feature extraction, respectively passing through a ReLU activation function and a maximum pooling layer after passing through the convolution layers, reducing the size of a feature map through downsampling operation and retaining the most remarkable features, conducting batch normalization, connecting a Dropout layer to each pooling layer, connecting a flat layer, flattening the features extracted by the convolution layers and the pooling layers into a one-dimensional vector, and finally passing through the full connection layer twice.
9. The sleep physiological signal based depression screening system according to claim 8, wherein the feature extraction is performed by stacking a plurality of convolution layers, corresponding to three two-dimensional convolutions, the number of convolution kernels of the first convolution layer being 16, the number of convolution kernels of the second convolution layer being 36, the number of convolution kernels of the third convolution layer being 48, and the convolution kernels of all three convolution layers being (3, 3).
CN202411814883.XA 2024-12-11 2024-12-11 Depression screening system based on sleep physiological signals Pending CN119700114A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411814883.XA CN119700114A (en) 2024-12-11 2024-12-11 Depression screening system based on sleep physiological signals

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411814883.XA CN119700114A (en) 2024-12-11 2024-12-11 Depression screening system based on sleep physiological signals

Publications (1)

Publication Number Publication Date
CN119700114A true CN119700114A (en) 2025-03-28

Family

ID=95086188

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411814883.XA Pending CN119700114A (en) 2024-12-11 2024-12-11 Depression screening system based on sleep physiological signals

Country Status (1)

Country Link
CN (1) CN119700114A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121092891A (en) * 2025-08-19 2025-12-09 北京数通魔方科技有限公司 Artificial intelligence-based petroleum industry personnel behavior safety video monitoring system
CN121549823A (en) * 2026-01-21 2026-02-24 湖南工商大学 Depression detection device

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017016086A1 (en) * 2015-07-30 2017-02-02 华南理工大学 Depression evaluating system and method based on physiological information
CN109363670A (en) * 2018-11-13 2019-02-22 杭州电子科技大学 An intelligent detection method for depression based on sleep monitoring
CN113080966A (en) * 2021-03-22 2021-07-09 华南师范大学 Automatic depression detection method based on sleep stages
WO2024249524A1 (en) * 2023-05-31 2024-12-05 Neurovigil, Inc. Rapid eye movement sleep disorder detection to facilitate selective screening for parkinson's disorder

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017016086A1 (en) * 2015-07-30 2017-02-02 华南理工大学 Depression evaluating system and method based on physiological information
CN109363670A (en) * 2018-11-13 2019-02-22 杭州电子科技大学 An intelligent detection method for depression based on sleep monitoring
CN113080966A (en) * 2021-03-22 2021-07-09 华南师范大学 Automatic depression detection method based on sleep stages
WO2024249524A1 (en) * 2023-05-31 2024-12-05 Neurovigil, Inc. Rapid eye movement sleep disorder detection to facilitate selective screening for parkinson's disorder

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
杨家豪，等: "《计算机系统应用》", "基于睡眠脑电信号的抑郁症诊断算法", vol. 33, no. 11, 30 November 2024 (2024-11-30), pages 167 - 176 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN121092891A (en) * 2025-08-19 2025-12-09 北京数通魔方科技有限公司 Artificial intelligence-based petroleum industry personnel behavior safety video monitoring system
CN121549823A (en) * 2026-01-21 2026-02-24 湖南工商大学 Depression detection device

Similar Documents

Publication Publication Date Title
Zhao et al. SleepContextNet: A temporal context network for automatic sleep staging based single-channel EEG
CN114366124B (en) A method for epileptic EEG recognition based on semi-supervised deep convolutional channel attention single-classification network
CN113768519B (en) Method for analyzing consciousness level of patient based on deep learning and resting state electroencephalogram data
CN113951900B (en) A motor imagery intent recognition method based on multimodal signals
CN119700114A (en) Depression screening system based on sleep physiological signals
CN116058800B (en) Automatic sleep staging system based on deep neural network and brain-computer interface
CN112641451B (en) Multi-scale residual error network sleep staging method and system based on single-channel electroencephalogram signal
Peng et al. A deep learning framework for ECG denoising and classification
CN111166327A (en) Epilepsy diagnosis device based on single-channel EEG signal and convolutional neural network
Wei et al. MS-Net: Sleep apnea detection in PPG using multi-scale block and shadow module one-dimensional convolutional neural network
CN114569096B (en) A non-contact continuous blood pressure measurement method and system based on video stream
CN118902458B (en) Multimodal emotion recognition and consciousness detection method based on electroencephalogram and micro-expression
WO2021147292A1 (en) Product design scheme decision-making method combining electroencephalogram and eye movement with user similarity
CN119128803A (en) Optoelectronic multimodal decision fusion method and system for mental disorder identification
CN115399735A (en) Multi-head attention mechanism sleep staging method based on time-frequency double-current enhancement
CN113116300A (en) Physiological signal classification method based on model fusion
CN118153626A (en) Lightweight network model and construction method for surface electromyography signal gesture recognition
CN120130931A (en) Epileptic seizure detection method and system based on self-attention mechanism and GRU-LSTM fusion
Ying et al. HybridDomainSleepNet: A hybrid common-private domain deep learning network for automatic sleep staging
Li et al. Efficient ECG classification method for arrhythmia using MODWPT and adaptive incremental broad learning
CN112788200A (en) Method and device for determining frequency spectrum information, storage medium and electronic device
CN113598791B (en) Consciousness disturbance classification method based on time-space convolution neural network used by resting state electroencephalogram
CN115146680A (en) Multi-modal physiological signal identification method and system
CN120873742A (en) Atrial fibrillation prediction method and system based on fusion of local and global features
CN120478843A (en) Electromagnetic personalized treatment parameter generation method and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination