CN111493836B - Postoperative acute pain prediction system based on brain-computer interface and deep learning and application - Google Patents

Postoperative acute pain prediction system based on brain-computer interface and deep learning and application Download PDF

Info

Publication number
CN111493836B
CN111493836B CN202010481486.0A CN202010481486A CN111493836B CN 111493836 B CN111493836 B CN 111493836B CN 202010481486 A CN202010481486 A CN 202010481486A CN 111493836 B CN111493836 B CN 111493836B
Authority
CN
China
Prior art keywords
deep learning
brain
pain
electroencephalogram
postoperative
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.)
Active
Application number
CN202010481486.0A
Other languages
Chinese (zh)
Other versions
CN111493836A (en
Inventor
高忠科
曲志勇
侯林华
马文庆
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.)
Tianjin Furuilong Metal Products Co ltd
Tianjin University
Original Assignee
Tianjin Furuilong Metal Products Co ltd
Tianjin 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 Tianjin Furuilong Metal Products Co ltd, Tianjin University filed Critical Tianjin Furuilong Metal Products Co ltd
Priority to CN202010481486.0A priority Critical patent/CN111493836B/en
Publication of CN111493836A publication Critical patent/CN111493836A/en
Application granted granted Critical
Publication of CN111493836B publication Critical patent/CN111493836B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • A61B5/48—Other medical applications
    • A61B5/4824—Touch or pain perception evaluation
    • 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
    • A61B5/25—Bioelectric electrodes therefor
    • A61B5/279—Bioelectric electrodes therefor specially adapted for particular uses
    • A61B5/291—Bioelectric electrodes therefor specially adapted for particular uses for 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
    • 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
    • A61B5/40—Detecting, measuring or recording for evaluating the nervous system
    • A61B5/4005—Detecting, measuring or recording for evaluating the nervous system for evaluating the sensory system
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
    • A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
    • A61B5/6802—Sensor mounted on worn items
    • A61B5/6803—Head-worn items, e.g. helmets, masks, headphones or goggles
    • 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)
  • Animal Behavior & Ethology (AREA)
  • Veterinary Medicine (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Psychiatry (AREA)
  • Artificial Intelligence (AREA)
  • Physiology (AREA)
  • Hospice & Palliative Care (AREA)
  • Pain & Pain Management (AREA)
  • Neurology (AREA)
  • Neurosurgery (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Signal Processing (AREA)
  • Psychology (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

A postoperative acute pain prediction system based on brain-computer interface and deep learning and application thereof comprise: acquiring resting state electroencephalogram signals of a patient before operation for 0-3 days by portable electroencephalogram acquisition equipment; performing data segmentation on the preoperative resting state electroencephalogram signal through a sliding window to obtain a series of resting state electroencephalogram signal segments; a series of resting state electroencephalogram signal segments are correspondingly converted into a series of two-dimensional images by utilizing a preprocessing method for subsequent analysis; constructing a data set, adding labels and dividing a training set and a testing set; and constructing a prediction model based on deep learning, determining a model structure and model parameters to be optimized, and obtaining the prediction model based on deep learning, which can predict postoperative acute pain, through training and testing. The portable electroencephalogram acquisition equipment can conveniently and efficiently acquire electroencephalogram signals; the postoperative pain category is predicted under the condition of no damage to the patient through the analysis of the brain electrical signals of the patient in the preoperative resting state.

Description

Postoperative acute pain prediction system based on brain-computer interface and deep learning and application
Technical Field
The invention relates to prediction of postoperative acute pain. In particular to a postoperative acute pain prediction system based on brain-computer interface and deep learning and application thereof.
Background
Pain is a common condition defined by the international association for pain research as "an unpleasant subjective feeling and emotional experience associated with tissue damage or potential tissue damage". Pain can be classified into acute pain and chronic pain according to the duration and nature of the pain. Acute pain refers to pain that exists for a short period of time, usually after a noxious stimulus, and may progress to chronic pain if not completely controlled in the initial phase. Chronic pain refers to pain which lasts for more than three months due to untimely treatment or improper medication, and long-term chronic pain can cause abnormal changes of the cerebral nervous system and seriously affect the life quality of patients. The scalp electroencephalogram signals are an important way for objectively evaluating pain diseases, research and analysis on the electroencephalogram signals of the patient before operation in a resting state can provide important basis for classification of acute pain and chronic pain, and effective prediction of postoperative acute pain has important guiding significance for a medical operation process.
The brain electrical activity is derived from the free discharge activity of the brain neuron population, and is the general reflection of the electrophysiological activity of the brain nerve cells on the surface of the cerebral cortex or scalp. The electroencephalogram signals contain a large amount of physiological and disease information, and research and analysis on the electroencephalogram signals can not only provide diagnosis basis for certain brain diseases, but also provide effective treatment means for certain brain diseases. People try to effectively extract and analyze electroencephalogram signals through a brain-computer interface (BCI), thereby achieving a certain control purpose. Because the electroencephalogram signal is a non-stationary random signal without ergodicity and the noise influence is great, the analysis and the processing of the electroencephalogram signal are important research contents. The electroencephalogram signal analysis method comprises the traditional methods of Fourier transform, frequency domain analysis, time domain analysis and the like, and then wavelet analysis, matching tracking method, neural network analysis, chaos analysis and the like appear, so that deep learning provides a more efficient analysis method for electroencephalogram analysis, and the development of the electroencephalogram signal analysis method is powerfully promoted.
Deep learning, which is a data-driven algorithm, can automatically learn abstract representation features of original data, and is widely used in a feature extraction process. The convolutional neural network is one of the most effective deep learning algorithms and is mainly applied to the fields of image recognition, face detection, character recognition and the like. The convolutional neural network uses convolutional layers and pooling layers, and simultaneously introduces a mechanism of local receptive field and weight sharing, so that the number of parameters to be trained is greatly reduced, the best convolutional kernels aiming at a specific classification task and the combination mode of the convolutional kernels are automatically learned, and the best feature expression of an input image for the classification task is calculated. In recent years, the use of the convolutional neural network in time sequence analysis is gradually increased, the convolutional neural network is used for fully extracting the characteristics existing in the time sequence, and the analysis of the time sequence, particularly the effective analysis of electroencephalogram signals can be effectively realized.
Disclosure of Invention
The invention aims to solve the technical problem of providing a postoperative acute pain prediction system based on brain-computer interface and deep learning and capable of obtaining a relatively accurate prediction result and application thereof.
The technical scheme adopted by the invention is as follows: a postoperative acute pain prediction system based on brain-computer interface and deep learning comprises the following steps:
1) obtaining resting state electroencephalogram signals of 0-3 days before operation of a patient through portable electroencephalogram acquisition equipment
Figure BDA0002517561920000021
Wherein g denotes the g-th channel, LeIndicating the data length of each signal;
2) for resting state electroencephalogram signals before operation
Figure BDA0002517561920000022
A passage length of leThe sliding window is used for carrying out data segmentation, the sliding step length of the sliding window is s, a series of resting state electroencephalogram signal segments are obtained, and the p-th resting state electroencephalogram signal segment is expressed as
Figure BDA0002517561920000023
3) A series of resting state electroencephalogram signal segments are correspondingly converted into a series of two-dimensional images by utilizing a preprocessing method for subsequent analysis;
4) constructing a data set, adding labels and dividing a training set and a testing set;
5) and constructing a prediction model based on deep learning, determining a model structure and model parameters to be optimized, and obtaining the prediction model based on deep learning, which can predict postoperative acute pain, through training and testing.
Use of a brain-computer interface and deep learning based post-operative acute pain prediction system according to claim 1, comprising:
1) based on it is portableThe electroencephalogram acquisition equipment acquires the preoperative resting state electroencephalogram signals of patients in the actual medical process
Figure BDA0002517561920000024
Uploading, storing and data partitioning;
2) the data signals obtained by segmentation are subjected to the prediction system of the postoperative acute pain based on the brain-computer interface and the deep learning
Figure BDA0002517561920000025
Carrying out pretreatment;
3) inputting the preprocessed two-dimensional image into a prediction model capable of predicting postoperative acute pain and based on deep learning to predict postoperative acute pain category of the patient, and displaying the postoperative acute pain category on a display in real time.
According to the postoperative acute pain prediction system based on the brain-computer interface and the deep learning and the application thereof, the portable electroencephalogram acquisition equipment is low in manufacturing cost, open in interface, small in size, convenient to carry and wear and capable of conveniently and efficiently acquiring electroencephalogram signals; the implicit characteristics in the electroencephalogram signals can be fully extracted by using an intelligent algorithm based on deep learning, and the result of accurately predicting postoperative acute pain can be obtained; the intelligent algorithm based on deep learning can change the model structure by adjusting the number of different layers, so that the method is suitable for the calculation and analysis of data with different data volumes and complexity and has good expansibility; the postoperative pain category is predicted under the condition of no damage to the patient through the analysis of the preoperative resting state electroencephalogram signal of the patient, and the method has important guiding significance for the operation medical treatment.
Drawings
FIG. 1 is a flow chart of the post-operative acute pain prediction system and application based on brain-computer interface and deep learning of the present invention;
FIG. 2 is a system block diagram of a portable electroencephalogram acquisition device in the present invention;
FIG. 3 is a diagram illustrating the distribution and names of the 32-lead EEGs according to the present invention;
FIG. 4 is a schematic diagram of a method for preprocessing brain electrical signals according to the present invention;
FIG. 5 is a schematic diagram of the deep learning-based prediction model of the present invention.
Detailed Description
The following describes the brain-computer interface and deep learning based postoperative acute pain prediction system and application of the present invention in detail with reference to the accompanying drawings.
As shown in fig. 1, the system for predicting acute postoperative pain based on brain-computer interface and deep learning of the present invention comprises the following steps:
1) obtaining resting state electroencephalogram signals of 0-3 days before operation of a patient through portable electroencephalogram acquisition equipment
Figure BDA0002517561920000031
Wherein g denotes the g-th channel, LeIndicating the data length of each signal;
as shown in fig. 2, the portable electroencephalogram acquisition device comprises: a brain electrode cap and a patch cord 1 thereof which are connected in sequence and used for collecting brain electrical signals, a bioelectric signal collecting module 2 used for amplifying and converting the brain electrical signals, an FPGA processor 3 used for controlling the collection of the brain electrical signals and outputting the brain electrical signals through a USB communication circuit 4, and a system power supply circuit 5 respectively connected with the bioelectric signal collecting module 2 and the FPGA processor 3, wherein,
the brain electrode cap and the brain electrode cap in the patch cord 1 thereof are used for collecting brain electrical signals of different brain areas, are connected with the bioelectric signal collecting module 2 through the patch cord and a DSUB37 interface and are used for collecting and transmitting bioelectric signals;
the bioelectrical signal acquisition module 2 consists of a plurality of bioelectrical signal acquisition chips which are integrated with a high common mode rejection ratio analog input module for receiving the electroencephalogram voltage signals acquired by the electroencephalogram cap, a low-noise programmable gain amplifier for amplifying the electroencephalogram voltage signals and a high-resolution synchronous sampling analog-to-digital converter for converting the analog signals into digital signals;
the FPGA processor 3 is used for adjusting the acquisition mode and parameters of the bioelectricity signal acquisition module 2 and controlling the USB communication circuit 4 to output electroencephalogram signal data;
the USB communication circuit 4 works in an asynchronous FIFO mode, has the highest transmission rate of 8 MB/s, and periodically outputs the acquired electroencephalogram signals in the form of data packets under the control of the FPGA processor 3;
the system power supply circuit 5 has an input voltage of 5V, is powered by the USB interface, and provides working voltages of different chips of the system through the voltage conversion module.
As shown in fig. 3, the portable electroencephalogram acquisition equipment acquires electroencephalogram signals of twenty-one electrodes of a subject corresponding to FP1, FP2, F3, F7, T7, P7, O1, C3, P3, F4, C4, P4, F8, O2, Fz, Cz, Pz, FT9, FT10, T8 and P8 of an electroencephalogram cap; the electrode distribution of the brain electrode cap conforms to 10/20 international standard leads.
2) For resting state electroencephalogram signals before operation
Figure BDA0002517561920000032
A passage length of leThe sliding window is used for carrying out data segmentation, the sliding step length of the sliding window is s, a series of resting state electroencephalogram signal segments are obtained, and the p-th resting state electroencephalogram signal segment is expressed as
Figure BDA0002517561920000033
3) A series of resting state electroencephalogram signal segments are correspondingly converted into a series of two-dimensional images by utilizing a preprocessing method for subsequent analysis; as shown in fig. 3, includes:
randomly obtaining a length L from each resting state electroencephalogram signal segment2The value of s (i) at each time point in the signal segment, i ═ 1,2, … L2Where i represents a time point and L is a length, the pixel intensity of the converted two-dimensional image is
Figure BDA0002517561920000034
j is 1,2, … N, k is 1,2, … N, round (-) is a rounding function, max (-) is a maximum function, min (-) is a minimum function, the integer pixel value is normalized to the range of 0-255, i.e. the pixel intensity of the gray scale image;
wherein, P is the pixel intensity, S is the signal value, j and k are respectively the abscissa and ordinate corresponding to the pixel point, and N is the maximum coordinate value.
4) Constructing a data set, adding labels and dividing a training set and a testing set;
the data set is a series of preprocessed two-dimensional images, the labels are two kinds of labels of postoperative acute pain and postoperative chronic pain, the postoperative acute pain is to continuously track and investigate a patient after operation, pain digital scoring is carried out on the patient by using a European five-dimensional health scale and a simple pain assessment scale, and the postoperative acute pain is defined if the pain digital scoring is more than 3 on the third day after operation; the postoperative chronic pain is obtained by performing follow-up visits on patients for 1,2, 4, 6 and 8 weeks after operation, performing pain numerical scoring on the patients by using a European five-dimensional health scale, a concise pain assessment scale and a pain catastrophe scale, and defining the postoperative chronic pain as the postoperative chronic pain if the pain numerical scoring is greater than or equal to 2 at 8 weeks after operation.
5) And constructing a prediction model based on deep learning, determining a model structure and model parameters to be optimized, and obtaining the prediction model based on deep learning, which can predict postoperative acute pain, through training and testing. Wherein the content of the first and second substances,
(A) as shown in fig. 5, the prediction model based on deep learning includes: n is a radical of an alkyl radical1A convolution layer, n2Individual pooling layer, n3A full connection layer, n4Batch normalization layer, n5A random loss layer, 1 branch fusion structure; wherein n is1、n2、n3、n4And n5Is set artificially according to the data quantity and the data characteristics, n1、n2And n4Are all positive integers greater than 5, n3And n5All values are positive integers not less than 2;
each convolutional layer extracts valid features from the input of the model by convolution calculation, which is as follows:
Xl＝f(∑Xl-1*wl+bl)
wherein XlAnd Xl-1Feature maps, w, representing the current and previous layers, respectivelylRepresents a weight, blRepresenting bias, f represents an activation function, and a Relu activation function is selected;
each pooling layer is used for enlarging the receptive field, a matrix window is used for scanning on the characteristic diagram, the number of elements in each matrix is reduced through a pooling method, the spatial position relation of the characteristics is kept, and the pooling method is a maximum pooling method, an average pooling method or a spatial pyramid pooling method;
each full connection layer is used for flattening multi-dimensional data and converting the multi-dimensional data into one-dimensional vectors;
each batch normalization layer is used for distributing the neurons output by the previous layer in a standard normal distribution with the mean value of 0 and the variance of 1, and the formula is
Figure BDA0002517561920000041
Wherein, yoDenotes output after batch normalization, BN denotes batch normalization, x0Is a neuron before the batch standardization,
Figure BDA0002517561920000042
is the neuron after the conversion, and the neuron is the neuron,
Figure BDA0002517561920000043
is the mean of the smallest batch, m is the smallest batch,
Figure BDA0002517561920000044
is the variance under the current batch, γ and β are the parameters being learned, ε is a constant;
each random loss layer is used for randomly deleting a part of neurons in the network according to a set quantity so as to reduce the overfitting phenomenon;
the branch fusion structure comprises more than 2 convolutional neural networks, and features extracted by the convolutional neural networks are fused together through vector addition or vector splicing.
(B) The model training process comprises the following steps:
(1) setting an initial learning rate by taking a training set as the input of a prediction model based on deep learning, and optimizing model parameters by using an optimization algorithm for training, wherein the optimization algorithm is Adam or SGD;
(2) taking a test set as the input of a prediction model based on deep learning after training, and adjusting the structure and parameters of the prediction model based on deep learning through a generated accuracy rate change curve and a loss condition, wherein a cross entropy loss function is used for evaluating the difference condition of probability distribution and real distribution obtained by current training;
(3) and (3) repeating the steps (1) to (2) until a deep learning-based prediction model with the accuracy rate of predicting the postoperative acute pain larger than 90% is obtained as the deep learning-based prediction model capable of predicting the postoperative acute pain.
The invention discloses an application of a postoperative acute pain prediction system based on brain-computer interface and deep learning, which comprises the following steps:
1) method for acquiring preoperative resting state electroencephalogram signals of patients in actual medical process based on portable electroencephalogram acquisition equipment
Figure BDA0002517561920000051
Uploading, storing and data partitioning;
2) the data signals obtained by segmentation are subjected to the prediction system of the postoperative acute pain based on the brain-computer interface and the deep learning
Figure BDA0002517561920000052
Carrying out pretreatment;
3) inputting the preprocessed two-dimensional image into a prediction model capable of predicting postoperative acute pain and based on deep learning to predict postoperative acute pain category of the patient, and displaying the postoperative acute pain category on a display in real time.
As a preferred embodiment, the invention analyzes the brain electrical signals of the preoperative resting state of 26 patients and predicts postoperative acute pain. The branch fusion structure extracts multi-scale features by using convolution kernels with different sizes, wherein convolution calculation in one branch structure uses a convolution kernel of 3 x 3, convolution calculation in the other branch structure uses a convolution kernel of 5 x 5, the features extracted by the two branch structures are fused in series, then a series of operations such as convolution, pooling, batch standardization and full connection are carried out again, a Relu activation function is selected for activation after random loss, finally an output result is obtained through a full connection layer with the dimensionality of 2 and the activation function of softmax, an initial learning rate is set to be 0.0005 in the training process, model parameters are optimized by using an Adam algorithm, a data set comprises 355 training set samples and 40 testing set samples, and through training of 500 epochs, the testing accuracy rate of over 94 percent is obtained, and the effectiveness of the invention is verified.
The above description of the present invention and the embodiments is not limited thereto, and the description of the embodiments is only one of the implementation manners of the present invention, and any structure or embodiment similar to the technical solution without inventive design is within the protection scope of the present invention without departing from the inventive spirit of the present invention.

Claims (6)

1. A postoperative acute pain prediction system based on brain-computer interface and deep learning is characterized in that the operation of the system specifically comprises the following steps:
1) obtaining resting state electroencephalogram signals of 0-3 days before operation of a patient through portable electroencephalogram acquisition equipment
Figure FDA0003596142030000011
Wherein g denotes the g-th channel, LeIndicating the data length of each signal;
2) for resting state electroencephalogram signals before operation
Figure FDA0003596142030000012
A passage length of leThe sliding window is used for carrying out data segmentation, the sliding step length of the sliding window is s, a series of resting state electroencephalogram signal segments are obtained, and the p-th resting state electroencephalogram signal segment is expressed as
Figure FDA0003596142030000013
3) A series of resting state electroencephalogram signal segments are correspondingly converted into a series of two-dimensional images by utilizing a preprocessing method for subsequent analysis; the method comprises the following steps: randomly obtaining a length L from each resting state electroencephalogram signal segment2The value of s (i) at each time point in the signal segment, i ═ 1,2, … L2Where i represents a time point and L is a length, the pixel intensity of the converted two-dimensional image is
Figure FDA0003596142030000014
The round (-) function is a rounding function, max (-) is a maximum function, min (-) is a minimum function, and the integer pixel value is normalized to the range of 0-255, i.e. the pixel intensity of the gray scale image;
wherein, P is pixel intensity, S is a signal value, j and k are respectively a horizontal coordinate and a vertical coordinate corresponding to the pixel point, and N is a maximum coordinate value;
4) constructing a data set, adding labels and dividing a training set and a testing set;
5) and constructing a prediction model based on deep learning, determining a model structure and model parameters to be optimized, and obtaining the prediction model based on deep learning, which can predict postoperative acute pain, through training and testing.
2. The system for predicting postoperative acute pain based on brain-computer interface and deep learning of claim 1, wherein the portable brain electricity collecting device of step 1) comprises: a brain electrode cap and a patch cord (1) thereof which are connected in sequence and used for collecting brain electrical signals, a bioelectric signal collecting module (2) used for amplifying and converting the brain electrical signals, an FPGA processor (3) used for controlling the collection of the brain electrical signals and outputting the brain electrical signals through a USB communication circuit (4), and a system power supply circuit (5) respectively connected with the bioelectric signal collecting module (2) and the FPGA processor (3), wherein,
the brain electrode cap and the brain electrode cap in the patch cord (1) thereof are used for collecting brain electrical signals of different brain areas, are connected with the bioelectricity signal collecting module (2) through the patch cord and a DSUB37 interface and are used for collecting and transmitting bioelectricity signals;
the bioelectrical signal acquisition module (2) is composed of a plurality of bioelectrical signal acquisition chips which are integrated with a high common mode rejection ratio analog input module for receiving electroencephalogram voltage signals acquired by an electroencephalogram cap, a low-noise programmable gain amplifier for amplifying the electroencephalogram voltage signals and a high-resolution synchronous sampling analog-to-digital converter for converting the analog signals into digital signals;
the FPGA processor (3) is used for adjusting the acquisition mode and parameters of the bioelectricity signal acquisition module (2) and controlling the USB communication circuit (4) to output electroencephalogram signal data;
the USB communication circuit (4) works in an asynchronous FIFO mode, the highest transmission rate is 8 MB/s, and the acquired electroencephalogram signals are periodically output in a data packet mode under the control of the FPGA processor (3);
the input voltage of the system power supply circuit (5) is 5V, the USB interface supplies power, and the working voltage of different chips of the system is provided through the voltage conversion module.
3. The brain-computer interface and deep learning based postoperative acute pain prediction system of claim 2, wherein the portable electroencephalograph acquisition device acquires electroencephalograms of twenty-one electrodes in total of the subject corresponding to FP1, FP2, F3, F7, T7, P7, O1, C3, P3, F4, C4, P4, F8, O2, Fz, Cz, Pz, FT9, FT10, T8 and P8 of the electroencephalogram cap; the electrode distribution of the brain electrode cap conforms to 10/20 international standard leads.
4. The brain-computer interface and deep learning based postoperative acute pain prediction system according to claim 1, wherein the data set in step 4) is a pre-processed series of two-dimensional images, the labels are two types of labels of postoperative acute pain and postoperative chronic pain, the postoperative acute pain is to be continuously followed and investigated for patients after operation, the patients are scored with a European five-dimensional health scale and a concise pain assessment scale, and the postoperative acute pain is defined if the score of the pain score is greater than 3 on the third day after operation; the postoperative chronic pain is obtained by performing follow-up visits on patients for 1,2, 4, 6 and 8 weeks after operation, performing pain numerical scoring on the patients by using a European five-dimensional health scale, a concise pain assessment scale and a pain catastrophe scale, and defining the postoperative chronic pain as the postoperative chronic pain if the pain numerical scoring is greater than or equal to 2 at 8 weeks after operation.
5. The brain-computer interface and deep learning based prediction system for acute postoperative pain according to claim 1, wherein the deep learning based prediction model in step 5) comprises: n is1A convolution layer, n2Individual pooling layer, n3A full connection layer, n4Batch normalization layer, n5A random loss layer, 1 branch fusion structure; wherein n is1、n2、n3、n4And n5Is set artificially according to the data quantity and the data characteristics, n1、n2And n4Are all positive integers greater than 5, n3And n5All values are positive integers not less than 2;
each convolutional layer extracts valid features from the input of the model by convolution calculation, which is as follows:
Xl＝f(∑Xl-1*wl+bl)
wherein XlAnd Xl-1Feature maps, w, representing the current and previous layers, respectivelylRepresents a weight, blRepresenting bias, f represents an activation function, and a Relu activation function is selected;
each pooling layer is used for enlarging the receptive field, a matrix window is used for scanning on the characteristic diagram, the number of elements in each matrix is reduced through a pooling method, the spatial position relation of the characteristics is kept, and the pooling method is a maximum pooling method, an average pooling method or a spatial pyramid pooling method;
each full connection layer is used for flattening multi-dimensional data and converting the multi-dimensional data into one-dimensional vectors;
each batch normalization layer is used for distributing the neurons output by the previous layer in a standard normal distribution with the mean value of 0 and the variance of 1, and the formula is
Figure FDA0003596142030000021
Wherein, yoDenotes output after batch normalization, BN denotes batch normalization, x0Is a neuron before the batch standardization,
Figure FDA0003596142030000022
is the neuron after the conversion, and the neuron is the neuron,
Figure FDA0003596142030000023
is the mean of the smallest batch, m is the smallest batch,
Figure FDA0003596142030000024
is the variance under the current batch, γ and β are the parameters being learned, ε is a constant;
each random loss layer is used for randomly deleting a part of neurons in the network according to a set quantity so as to reduce the overfitting phenomenon;
the branch fusion structure comprises more than 2 convolutional neural networks, and features extracted by the convolutional neural networks are fused together through vector addition or vector splicing.
6. The brain-computer interface and deep learning based acute pain prediction system of claim 1, wherein the model training process in step 5) comprises:
(1) setting an initial learning rate by taking a training set as the input of a prediction model based on deep learning, and optimizing model parameters by using an optimization algorithm for training, wherein the optimization algorithm is Adam or SGD;
(2) taking a test set as the input of a prediction model based on deep learning after training, and adjusting the structure and parameters of the prediction model based on deep learning through a generated accuracy rate change curve and a loss condition, wherein a cross entropy loss function is used for evaluating the difference condition of probability distribution and real distribution obtained by current training;
(3) and (3) repeating the steps (1) to (2) until a deep learning-based prediction model with the accuracy rate of predicting the postoperative acute pain larger than 90% is obtained as the deep learning-based prediction model capable of predicting the postoperative acute pain.
CN202010481486.0A 2020-05-31 2020-05-31 Postoperative acute pain prediction system based on brain-computer interface and deep learning and application Active CN111493836B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010481486.0A CN111493836B (en) 2020-05-31 2020-05-31 Postoperative acute pain prediction system based on brain-computer interface and deep learning and application

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010481486.0A CN111493836B (en) 2020-05-31 2020-05-31 Postoperative acute pain prediction system based on brain-computer interface and deep learning and application

Publications (2)

Publication Number Publication Date
CN111493836A CN111493836A (en) 2020-08-07
CN111493836B true CN111493836B (en) 2022-06-03

Family

ID=71872280

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010481486.0A Active CN111493836B (en) 2020-05-31 2020-05-31 Postoperative acute pain prediction system based on brain-computer interface and deep learning and application

Country Status (1)

Country Link
CN (1) CN111493836B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112263253B (en) * 2020-11-18 2021-09-17 山东大学 Depression recognition system, medium and equipment based on deep learning and electrocardiosignal
CN113925509B (en) * 2021-09-09 2024-01-23 杭州回车电子科技有限公司 Attention value calculation method and device based on electroencephalogram signals and electronic device
CN116831595B (en) * 2023-06-21 2024-04-12 中国人民解放军西部战区总医院 Pain grade assessment method

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110087125A1 (en) * 2009-10-09 2011-04-14 Elvir Causevic System and method for pain monitoring at the point-of-care
CN106503799B (en) * 2016-10-11 2018-11-30 天津大学 Deep learning model based on multiple dimensioned network and the application in brain status monitoring
CN107292887B (en) * 2017-06-20 2020-07-03 电子科技大学 Retinal vessel segmentation method based on deep learning adaptive weight
CN108446020B (en) * 2018-02-28 2021-01-08 天津大学 Motor imagery idea control method fusing visual effect and deep learning and application
CN108433722A (en) * 2018-02-28 2018-08-24 天津大学 Portable brain electric collecting device and its application in SSVEP and Mental imagery
CN110599413B (en) * 2019-08-15 2023-05-09 江苏大学 Laser facula image denoising method based on deep learning convolutional neural network
CN111079837B (en) * 2019-12-16 2022-06-28 桂林电子科技大学 Method for detecting, identifying and classifying two-dimensional gray level images

Also Published As

Publication number Publication date
CN111493836A (en) 2020-08-07

Similar Documents

Publication Publication Date Title
CN111493836B (en) Postoperative acute pain prediction system based on brain-computer interface and deep learning and application
CN109389059B (en) P300 detection method based on CNN-LSTM network
CN111616681B (en) Anesthesia state monitoring system based on portable electroencephalogram acquisition equipment and deep learning
CN109784023B (en) Steady-state vision-evoked electroencephalogram identity recognition method and system based on deep learning
CN110969108B (en) Limb action recognition method based on autonomic motor imagery electroencephalogram
CH716863A2 (en) Depression detection system based on channel selection of multi-channel electroencephalography made using training sets.
CN111616682B (en) Epileptic seizure early warning system based on portable electroencephalogram acquisition equipment and application
CN111513735B (en) Major depressive disorder identification system based on brain-computer interface and deep learning and application
Li et al. EEG signal classification method based on feature priority analysis and CNN
KR20190111570A (en) A system of detecting epileptic seizure waveform based on coefficient in multi-frequency bands from electroencephalogram signals, using feature extraction method with probabilistic model and machine learning
CN111584029A (en) Electroencephalogram self-adaptive model based on discriminant confrontation network and application of electroencephalogram self-adaptive model in rehabilitation
CN113855053A (en) Wearable muscle threshold monitoring system based on myoelectricity
CN114564990A (en) Electroencephalogram signal classification method based on multi-channel feedback capsule network
CN113723557A (en) Depression electroencephalogram classification system based on multiband time-space convolution network
Varnosfaderani et al. A two-layer lstm deep learning model for epileptic seizure prediction
CN110338760B (en) Schizophrenia three-classification method based on electroencephalogram frequency domain data
CN114595725A (en) Electroencephalogram signal classification method based on addition network and supervised contrast learning
Dzitac et al. Identification of ERD using fuzzy inference systems for brain-computer interface
Coyle et al. Extracting features for a brain-computer interface by self-organising fuzzy neural network-based time series prediction
Alazrai et al. A deep learning approach for decoding visually imagined digits and letters using time–frequency–spatial representation of EEG signals
CN116831874A (en) Lower limb rehabilitation device control method based on electromyographic signals
CN116746947A (en) Cross-subject electroencephalogram signal classification method based on online test time domain adaptation
CN114145745B (en) Graph-based multitasking self-supervision emotion recognition method
CN116421200A (en) Brain electricity emotion analysis method of multi-task mixed model based on parallel training
CN114548165B (en) Myoelectricity mode classification method capable of crossing users

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
GR01 Patent grant
GR01 Patent grant