CN113397562A - Sleep spindle wave detection method based on deep learning - Google Patents

Sleep spindle wave detection method based on deep learning Download PDF

Info

Publication number
CN113397562A
CN113397562A CN202110817798.9A CN202110817798A CN113397562A CN 113397562 A CN113397562 A CN 113397562A CN 202110817798 A CN202110817798 A CN 202110817798A CN 113397562 A CN113397562 A CN 113397562A
Authority
CN
China
Prior art keywords
spindle wave
sleep
features
data set
deep learning
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
CN202110817798.9A
Other languages
Chinese (zh)
Inventor
刘铁军
王林
郜东瑞
王钰潇
李静
曹文鹏
秦云
汪曼青
陈卓
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.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
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 University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN202110817798.9A priority Critical patent/CN113397562A/en
Publication of CN113397562A publication Critical patent/CN113397562A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • A61B5/48—Other medical applications
    • A61B5/4806—Sleep evaluation
    • A61B5/4809—Sleep detection, i.e. determining whether a subject is asleep or not
    • 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
    • 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

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Molecular Biology (AREA)
  • Medical Informatics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Psychiatry (AREA)
  • Physiology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Anesthesiology (AREA)
  • Psychology (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention discloses a sleep spindle wave detection method based on deep learning, which is applied to the field of sleep detection and aims at solving the problem that the existing spindle wave detection cannot adapt to real-time spindle wave detection; then, carrying out special extraction on the manufactured data set; secondly, establishing a spindle wave network, and training the spindle wave network by adopting the extracted characteristics; and finally, performing sleep spindle wave real-time detection according to the trained spindle wave network.

Description

Sleep spindle wave detection method based on deep learning
Technical Field
The invention belongs to the field of sleep detection, and particularly relates to a sleep spindle wave detection technology.
Background
Sleep spindle waves are transient neurooscillations produced by the interaction of the thalamocortical nuclei with other thalamocortical nuclei during NREM sleep (stages N2 and N3), in the frequency range of 9-16 or 11-16Hz, and for a duration of 0.5-3 s. During NREM sleep, spindle waves can be observed in a wide variety of thalamic and neocortical structures and temporally couple with slow neocortical oscillations (SOs) and hippocampal Sharp Wave Ripples (SWRs). Mednick, Johnson et al, 2013, found that spindle waves are thought to contribute to many neurological processes, such as somatosensory development, thalamocortical sensory gating, and synaptic plasticity. The online monitoring and training of sleep spindles can bring more benefits and can provide other benefits, and Hennies et al in 2016 show that the influence of priori knowledge on memory consolidation can be predicted by spindle wave density due to experimental evidence. Sleep disorders or disorders are important symptoms in many neurological or neuropsychiatric diseases. Astori, Mander, gorgonin et al in 2016 suggest that characterization of sleep spindle waves (e.g., oscillation frequency, spindle density, duration) can be used as important biomarkers related to brain health, for early detection of neurodegenerative diseases such as Mild Cognitive Impairment (MCI) and alzheimer's disease, for assessment of cognitive development in children, and for prediction of stress and schizophrenia.
Spindle is an important hallmark of stage N2 sleep. In the human sleep laboratory, spindle detection requires manual annotation by sleep experts, a time-consuming and labor-intensive task, and is affected by inter-rater variability. Most of the work of automated spindle wave detection to date has been directed to off-line applications for specific populations, which may rely on various unsupervised or unsupervised techniques, including constant or adaptive thresholding, matching pursuits, time-frequency transforms, decision trees, and low-level optimization. For real-time processing, the spindle wave detection method requires dynamic processing of neural data, which is faster in computation speed or comparable to the data flow rate. Many of the available spindle wave detection algorithms cannot be used directly for online applications because they cannot process data in sequence or because they do not meet speed requirements.
Disclosure of Invention
In order to solve the technical problems, the invention provides a sleep spindle wave detection method based on deep learning.
The technical scheme adopted by the invention is as follows: a sleep spindle wave detection method based on deep learning comprises the following steps:
s1, making the acquired electroencephalogram physiological data including the sleep state into a data set; and processing the data set;
s2, extracting traditional features and abstract features of the data set processed in the step S1;
s3, constructing a spindle wave network;
s4, training the spindle wave network according to the traditional characteristics and the abstract characteristics of the data set;
and S5, detecting the sleep spindle waves by adopting the spindle wave network obtained by training.
Step S1 is to process the data set, and specifically includes:
s11, respectively carrying out zero-phase digital filtering;
s12, removing extremely low frequency base line, power frequency noise and high frequency noise in the signal;
s13, resampling the signal to 200 Hz;
and S14, segmenting the data processed in the step S13.
Step S3 the spindle wave network includes: two sub-networks, and a fully connected layer connected to the two sub-network outputs.
The two sub-networks described in step S3 have the same structure, and are both CNN + LSTM, where the CNN includes five convolutional layers for extracting time features, and then the extracted time features are input to the LSTM to obtain a time pattern in the time features.
The inputs to both sub-networks are conventional characteristics of the data set.
And splicing the output of the two sub-networks with the traditional characteristics, and inputting the spliced data into the full-connection layer.
The legacy features include time domain features and frequency domain features.
The abstract features are extracted through an unsupervised clustering algorithm K-means algorithm.
The invention has the beneficial effects that: firstly, processing acquired sleep data, reasonably dividing the data, aligning the data with a label, and making a data set; then, carrying out special extraction on the manufactured data set; secondly, establishing a spindle wave network, and training the spindle wave network by adopting the extracted characteristics; and finally, performing sleep spindle wave real-time detection according to the trained spindle wave network.
Drawings
FIG. 1 is a flow chart of a protocol of the present invention;
FIG. 2 is a schematic diagram of a data set production process according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a data set generation process provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of extracted spectral power signatures provided by an embodiment of the present invention;
FIG. 5 is a schematic diagram of K-Means extraction filters according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a slinglenet model according to an embodiment of the present invention;
fig. 7 is a schematic diagram of a CNN single layer provided in an embodiment of the present invention;
FIG. 8 is a schematic diagram of a CNN-LSTM provided by an embodiment of the present invention;
fig. 9 is a schematic diagram of identifying spindle waves according to an embodiment of the present invention.
Detailed Description
In order to facilitate the understanding of the technical contents of the present invention by those skilled in the art, the present invention will be further explained with reference to the accompanying drawings.
As shown in figure 1, the overall thought of the invention is gradually refined according to three processes of data set making, feature engineering and model layering, and a spindle wave spindle detection algorithm model of electroencephalogram in human sleep based on deep learning is constructed.
As shown in fig. 2, the data set generation flow of step S1 is divided into three parts, namely, data acquisition, data preprocessing and sleep data segmentation.
S11, data acquisition
First, physiological data acquired during human sleep is made into a data set containing electroencephalogram (EEG) physiological signals. C ═ C1,C2,…,CNC represents the data set of all people and N represents the number of people.
During human sleep, the physiological signal data of human bodies are continuously monitored, recorded and stored by a polysomnography monitor with a physiological signal acquisition function according to the American society for sleep medicine standard.
S12, preprocessing data
The original signals of the physiological data such as the brain electricity and the like are digitized after being sampled, and then zero-phase digital filtering is firstly respectively carried out to prevent the physiological signals with non-stationary property from phase distortion. Then, the extremely low frequency baseline, power frequency noise and high frequency noise in the signal are removed. And finally, resampling the signal to 200Hz, and finishing signal preprocessing.
S13 sleep data segmentation
Extracting electroencephalogram and brain signals, and performing original data set C (C) by using electroencephalogram physiological signals1,C2,…,CNAnd (6) making. The sleep electroencephalogram data from N persons are divided into subsets according to persons.
Visualizing the structure of sleep data, as in FIG. 3, dividing each person's overnight sleep data into M segments, F, in chronological orderi(X; filters) (i ═ 1,2, … …, K), filters represent abstract features. Grouping all the tested overnight sleep data into an individual data set, wherein the data set comprises N multiplied by M data size sample sections. A data set sample segment contains L signal sample points. The data set sample size for C is N × M × L.
The duration time of the sleep spindle wave is 0.5-3 seconds, so that the reasonable setting of the size of the segmented sleep data segment is important for the detection of the subsequent sleep spindle wave. The present invention divides the sleep data into one data segment per 50 (i.e., 250ms) sample points. One data segment corresponds to one category label. There are two class labels, including a positive class (containing sleep spindles) and a negative class (not containing sleep spindles). For each training sample data segment, the present invention preprocesses the sample by de-trending and scale normalization to calibrate. Where the spindle wave sample size in the data set for C is N × P × L. P represents the number of spindle waves in the sleep data section of each person.
And aligning the label with the data, and checking the segment data set to prepare and store the segment data set to a file.
S2, feature extraction, comprising the following sub-steps:
and S21, performing characteristic engineering on the data set sample size of the C. The method comprises the following steps of extracting abstract features: K-Means extracts filters, and traditional characteristics are extracted: time domain features and frequency domain features. The abstract features filters extracted by K-Means are used as initial weight parameters of a Convolutional Neural Network (CNN). The extracted traditional features and frequency domain features are taken as input.
And S22, extracting traditional features. The power signature is calculated from the sleep brain electrical signal using a fourier transform (FFT). The frequency domain characteristic quantity comprises a power spectral density characteristic quantity (PSD) and is calculated by the formula
Figure BDA0003170811530000041
Where x is the original EEG segment, f (x) is the amplitude value from the EEG data after fourier (FFT) transformation, and fmax and fmin are the maximum and minimum frequencies of 9-16Hz at the frequency end of the sleep spindle, respectively. The present invention also calculates the spectral power ratio, as shown in fig. 4, the formula is the spectral power ratio ═ sleep spindle wave/(θ + δ), (where the sleep spindle wave refers to the 9-16Hz band power, θ refers to the Theta band power, and δ refers to the Delta band power).
And S23, extracting abstract features. Abstract feature extraction was performed by the unsupervised clustering algorithm K-Means (K-Means) algorithm, as shown in FIG. 5. By using the unsupervised learning characteristic of the algorithm, representative features can be obtained from input data. Without adding extra human assistance, the input signal data is divided into K classes set and the centroids (i.e. filters) with high characterization effect are learned from the K classes, namely K-Means. The output dimension is determined by the size and number of filters according to CNN in the subsequent model. In order to accelerate the convergence speed of the K-Means when finding the mass center and reduce the calculated amount, the invention adopts a Mini Batch K-Means algorithm, and the Mini Batch only extracts a part of samples in the category to replace the respective types for calculation each time. By reducing the sample size, more points are avoided from being repeatedly calculated, while the calculation amount is also reduced, and the result does not differ much in accuracy from the K-Means.
S3, model construction, comprising the following steps:
and S31, performing characteristic engineering on the data set of the C. The abstract features extracted are the three-dimensional tensors K × 1 × 7. The extracted spectral power ratio is characterized by a three-dimensional tensor N × M × H. M represents a data set sample segment and H represents the number of features.
S32, through data set making process and characteristic engineering, the data set structure reaches the design idea of the invention.
And S33, dividing the data set of the C into a final training data set, a verification set and a test data set.
S34, inputting the data of the training data set into SpingleNet (spindle wave network) for training. And reasonably setting the structure, the training mode and the initial parameters of the network model, and starting the training of the model.
In order to improve the detection delay and accuracy of the sleep spindle wave, the invention provides a Deep Neural Network (DNN) named as SpindeNet to know the complex nonlinear characteristics and the spectral temporal structure of the sleep spindle wave.
The model of the invention is shown in fig. 6, the network structure of the spindlenet is 2 networks 1 and 2 including CNN + LSTM (Long Short-Term Memory network, a variation of RNN), and FC (full connection layer); the present embodiment is explained from three parts, namely a network input part, a network processing part and data splicing:
the network input part is specifically as follows: the network 1 input consists of the original EEG time series (1 x 50) and the network 2 input consists of the envelope of the same length of the band pass filtered signal. The power input is the spectral power ratio extracted by the characteristic engineering, and the dimension is NxMxH.
The network processing part is specifically as follows: as shown in fig. 8, in which the structure of CNN is shown in fig. 7, CNN is composed of five convolutional layers, each layer is composed of K one-dimensional (1D) temporal filters with size of 1 × 7, which are derived from the abstract features extracted in step S2, i.e., K centroids, and the filters are followed by Exponential Linear Units (ELUs). After each layer of convolutional layers, merging was performed using Max Pool of 1 × 5. After the CNN extracts the temporal features, the temporal features are input into the LSTM, and temporal patterns in the features are discovered through h hidden units of the LSTM. The data output by the second part is three-dimensional data N × M × h.
The data splicing specifically comprises the following steps: and splicing the data output by the second part and the spectral power ratio characteristics extracted in the characteristic engineering to form three-dimensional vector data NxMxU, wherein U is the sum of H and H. Then, the data is input into an FC (full connection layer) of the algorithm model of the invention, and the output data of the FC is used for detecting whether the sleep spindle wave exists through a Softmax loss function, as shown in figure 9.
And S35, training SpindeLeNet by using the training set generated by the data set to obtain the predicted value of whether spindle waves exist or not corresponding to the data sample. And comparing the loss function with the label of the expert manual classification result, calculating a loss function, and then performing back propagation. The optimal parameters (maximum average accuracy) are trained by iteration. And recording evaluation indexes such as accuracy, recall rate and F1 score of the data set, and completing the construction of the network model.
And inputting the test data set into the trained SpindeLeNet to obtain the detection result of the sleep spindle waves. Performance of SpindleNet and the two other sleep spindle wave methods were compared on a MASS dataset (n ═ 19 subjects).
Sleep spindle notes with two human expert notes were selected in both data sets (n 15 MASS subjects, containing notes from both experts). The present embodiment runs these detection methods and compares various performance indexes. On the MASS dataset (n ═ 15 subjects), the specificity of SpindleNet of the invention reached 97.06% ± 0.67%, and the specificity of the existing McSleep was: 94.49% ± 0.77%; the specificity of the existing spindlers is as follows: 98.37% + -0.37%; the FDR (false discovery rate) of the present invention is 19.03% ± 3.18%, and the FDR of the conventional McSleep is: 35.71% ± 4.6%, FDR of existing spindlers is: 14.42% ± 2.6%; the F1 score of the present invention was: 0.83 ± 0.02, and the existing McSleep has an F1 score of: 0.72 ± 0.04, and the existing Spindler has a F1 score of: 0.75 plus or minus 0.02; the accuracy of the invention is: 96.08% ± 0.44%, the accuracy of the existing McSleep is: 93.75% + -0.57%; the accuracy of the existing Spindler is: 95.37% ± 0.36%; it is clear that the method of the invention is more efficient and more accurate than the prior art.
It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims (8)

1. A sleep spindle wave detection method based on deep learning is characterized by comprising the following steps:
s1, making the acquired electroencephalogram physiological data including the sleep state into a data set; and processing the data set;
s2, extracting traditional features and abstract features of the data set processed in the step S1;
s3, constructing a spindle wave network;
s4, training the spindle wave network according to the traditional characteristics and the abstract characteristics of the data set;
and S5, detecting the sleep spindle waves by adopting the spindle wave network obtained by training.
2. The method for detecting sleep spindle waves based on deep learning of claim 1, wherein the step S1 is to process the data set, and specifically includes:
s11, respectively carrying out zero-phase digital filtering;
s12, removing extremely low frequency base line, power frequency noise and high frequency noise in the signal;
s13, resampling the signal to 200 Hz;
and S14, segmenting the data processed in the step S13.
3. The deep learning-based sleep spindle wave detection method according to claim 1, wherein the spindle wave network of step S3 includes: two sub-networks, and a fully connected layer connected to the two sub-network outputs.
4. The method as claimed in claim 3, wherein the two sub-networks in step S3 have the same structure, and are CNN + LSTM, and CNN includes five convolutional layers for extracting temporal features, and then inputting the extracted temporal features into LSTM to obtain temporal patterns in the temporal features.
5. The deep learning-based sleep spindle wave detection method according to claim 4, wherein the input of two sub-networks is a conventional feature of the data set.
6. The deep learning-based sleep spindle wave detection method according to claim 5, further comprising splicing outputs of two sub-networks with conventional features, and inputting the spliced data into a full connection layer.
7. The deep learning-based sleep spindle wave detection method according to claim 6, wherein the conventional features include time domain features and frequency domain features.
8. The deep learning-based sleep spindle wave detection method according to claim 6, wherein the abstract features are extracted by an unsupervised clustering algorithm K-means algorithm.
CN202110817798.9A 2021-07-20 2021-07-20 Sleep spindle wave detection method based on deep learning Pending CN113397562A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110817798.9A CN113397562A (en) 2021-07-20 2021-07-20 Sleep spindle wave detection method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110817798.9A CN113397562A (en) 2021-07-20 2021-07-20 Sleep spindle wave detection method based on deep learning

Publications (1)

Publication Number Publication Date
CN113397562A true CN113397562A (en) 2021-09-17

Family

ID=77687042

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110817798.9A Pending CN113397562A (en) 2021-07-20 2021-07-20 Sleep spindle wave detection method based on deep learning

Country Status (1)

Country Link
CN (1) CN113397562A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115844424A (en) * 2022-10-17 2023-03-28 北京大学 Sleep spindle wave grading identification method and system

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5283838A (en) * 1990-04-16 1994-02-01 Sharp Kabushiki Kaisha Neural network apparatus
CN105608444A (en) * 2016-01-27 2016-05-25 大连楼兰科技股份有限公司 Wild animal image identification method used for automatic driving
CN108446738A (en) * 2018-03-23 2018-08-24 联想(北京)有限公司 A kind of clustering method, device and electronic equipment
CN108852348A (en) * 2018-05-14 2018-11-23 中国人民解放军军事科学院军事医学研究院 The collection point sort method and system of scalp brain electricity
CN110378394A (en) * 2019-06-26 2019-10-25 广西大学 More physiological data convergence analysis methods neural network based
CN111631688A (en) * 2020-06-24 2020-09-08 电子科技大学 Algorithm for automatic sleep staging
CN113033898A (en) * 2021-03-26 2021-06-25 国核电力规划设计研究院有限公司 Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5283838A (en) * 1990-04-16 1994-02-01 Sharp Kabushiki Kaisha Neural network apparatus
CN105608444A (en) * 2016-01-27 2016-05-25 大连楼兰科技股份有限公司 Wild animal image identification method used for automatic driving
CN108446738A (en) * 2018-03-23 2018-08-24 联想(北京)有限公司 A kind of clustering method, device and electronic equipment
CN108852348A (en) * 2018-05-14 2018-11-23 中国人民解放军军事科学院军事医学研究院 The collection point sort method and system of scalp brain electricity
CN110378394A (en) * 2019-06-26 2019-10-25 广西大学 More physiological data convergence analysis methods neural network based
CN111631688A (en) * 2020-06-24 2020-09-08 电子科技大学 Algorithm for automatic sleep staging
CN113033898A (en) * 2021-03-26 2021-06-25 国核电力规划设计研究院有限公司 Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
PRATHAMESH M. KULKARNI ET AL.: "A deep learning approach for real-time detection of sleep spindles", 《J NEURAL ENG.》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115844424A (en) * 2022-10-17 2023-03-28 北京大学 Sleep spindle wave grading identification method and system
CN115844424B (en) * 2022-10-17 2023-09-22 北京大学 Sleep spindle wave hierarchical identification method and system

Similar Documents

Publication Publication Date Title
Wen et al. Deep convolution neural network and autoencoders-based unsupervised feature learning of EEG signals
Ranjan et al. A fuzzy neural network approach for automatic K-complex detection in sleep EEG signal
CN110163180A (en) Mental imagery eeg data classification method and system
Wei et al. Time-frequency convolutional neural network for automatic sleep stage classification based on single-channel EEG
CN111291614A (en) Child epilepsy syndrome classification method based on transfer learning multi-model decision fusion
Al-Salman et al. Detection of EEG K-complexes using fractal dimension of time frequency images technique coupled with undirected graph features
CN113158964B (en) Sleep stage method based on residual error learning and multi-granularity feature fusion
CN108567418A (en) A kind of pulse signal inferior health detection method and detecting system based on PCANet
CN109222966A (en) A kind of EEG signals sensibility classification method based on variation self-encoding encoder
Liu et al. Automatic sleep arousals detection from polysomnography using multi-convolution neural network and random forest
Vallabhaneni et al. Deep learning algorithms in eeg signal decoding application: a review
Li et al. Computerized wrist pulse signal diagnosis using gradient boosting decision tree
Sabor et al. Detection of the interictal epileptic discharges based on wavelet bispectrum interaction and recurrent neural network
CN110192864B (en) Cross-domain electrocardiogram biological characteristic identity recognition method
CN116211319A (en) Resting state multichannel electroencephalogram signal identification method based on graph attention network and sparse coding
CN113397562A (en) Sleep spindle wave detection method based on deep learning
Zan et al. Local Pattern Transformation-Based convolutional neural network for sleep stage scoring
Gnana Rajesh Analysis of MFCC features for EEG signal classification
CN114091529A (en) Electroencephalogram emotion recognition method based on generation countermeasure network data enhancement
Ebrahimpour et al. Epileptic seizure detection using a neural network ensemble method and wavelet transform
Qu et al. Epileptogenic region detection based on deep CNN with transfer learning
Hao et al. A novel sleep staging algorithm based on hybrid neural network
CN114757260A (en) Physiological electric signal classification method, device, electronic equipment and storage medium
Ouelli et al. Epilepsy seizure detection using autoregressive modelling and multiple layer perceptron neural network
Elghamry et al. Comparative analysis of various machine learning techniques for epileptic seizures detection and prediction using EEG data

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20210917