Electroencephalogram signal analysis device, method, electronic device, and storage medium

Abstract

The application is suitable for the technical field of computers, and provides an electroencephalogram analysis device, method, electronic device and storage medium. The electroencephalogram analysis device comprises a collection module, a processing module and an analysis module. The collection module is used for collecting an electroencephalogram of a target to be analyzed and transmitting the electroencephalogram to the processing module. The processing module is used for receiving the electroencephalogram and pre-processing the electroencephalogram to obtain a plurality of electroencephalogram input data, and then transmitting the plurality of electroencephalogram input data to the analysis module. The analysis module is internally provided with a trained electroencephalogram analysis model. The analysis module is used for receiving the plurality of electroencephalogram input data, inputting the plurality of electroencephalogram input data into the electroencephalogram analysis model, and obtaining an analysis result output by the electroencephalogram analysis model. The analysis result represents the possibility of the target to be analyzed suffering from Parkinson's disease. The application can timely and accurately find out whether the target to be analyzed has the possibility of suffering from Parkinson's disease.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computers, and particularly relates to an electroencephalogram analysis device and method, an electronic device, and a storage medium. BACKGROUND

[0002] Parkinson's disease (PD) is a common neurodegenerative disease, which often occurs in people around 60 years old. However, in recent years, there has been a trend of younger people developing Parkinson's disease. In the related art, when a subject has abnormal movement symptoms such as walking, speaking, and writing, the subject may have reached the late stage of Parkinson's disease. In fact, non-motor symptoms such as insomnia, anxiety, and convulsions of the subject may have existed in the earlier stage of the appearance of motor symptoms, but non-motor symptoms are usually not easily detected or are often considered to be symptoms of other diseases, so it is difficult to diagnose Parkinson's disease in a timely manner.

[0003] Therefore, it is insufficient to determine whether a subject has Parkinson's disease only according to the external symptoms of the subject, and there is an urgent need for a device that can more timely and accurately determine whether a subject has the possibility of developing Parkinson's disease. SUMMARY

[0004] Embodiments of the present application provide an electroencephalogram analysis device, method, electronic device, and storage medium to solve the problem that it is currently difficult to timely and accurately determine whether a subject has the possibility of developing Parkinson's disease.

[0005] The present application is achieved by the following technical solutions:

[0006] In a first aspect, the embodiments of the present application provide an electroencephalogram analysis device, comprising: a collection module, a processing module, and an analysis module; the collection module is configured to collect an electroencephalogram of a target to be analyzed and transmit the electroencephalogram to the processing module; the processing module is configured to receive the electroencephalogram and pre-process the electroencephalogram to obtain a plurality of electroencephalogram input data, and then transmit the plurality of electroencephalogram input data to the analysis module; the analysis module is provided with a trained electroencephalogram analysis model, and the analysis module is configured to receive the plurality of electroencephalogram input data and input the plurality of electroencephalogram input data into the electroencephalogram analysis model to obtain an analysis result output by the electroencephalogram analysis model, wherein the analysis result represents the possibility that the target to be analyzed has Parkinson's disease.

[0007] With reference to the first aspect, in some embodiments, the EEG analysis model comprises a plurality of convolution modules, a plurality of fully connected layers and a SoftMax layer arranged in sequence; the analysis module is specifically configured to: receive the plurality of EEG input data; input the plurality of EEG input data into the EEG analysis model, gradually extract features of the plurality of EEG input data through the plurality of convolution modules, and finally obtain the analysis result corresponding to the EEG signal of the target to be analyzed by weighting and fusing the features extracted by the last convolution module and performing classification prediction through the plurality of fully connected layers and the SoftMax layer.

[0008] With reference to the first aspect, in some embodiments, each convolution module comprises a convolution layer, a batch normalization layer and a pooling layer, and the output data of each pooling layer is intermediate layer output data; the EEG signal analysis device further comprises a display module; the display module is configured to receive the intermediate layer output data output by at least one specified pooling layer sent by the analysis module, and display a visual graph of the intermediate layer output data output by the at least one specified pooling layer.

[0009] With reference to the first aspect, in some embodiments, the display module is specifically configured to: for each EEG input data, obtain a first heat vector output by each specified pooling layer in the EEG analysis model, the first heat vector being a vector corresponding to the intermediate layer output data output by the specified pooling layer, the first heat vector comprising a convolution channel dimension, a time dimension and a space dimension; and average the first heat vector corresponding to each specified pooling layer in the convolution channel dimension and the time dimension to obtain a second heat vector corresponding to each specified pooling layer; for each specified pooling layer, average a preset number of second heat vectors corresponding to the specified pooling layer in the plurality of EEG input data to obtain a third heat vector corresponding to the specified pooling layer; and draw a visual graph of the output data corresponding to each specified pooling layer according to the third heat vector corresponding to each specified pooling layer.

[0010] With reference to the first aspect, in some embodiments, the acquisition module comprises a plurality of electrodes and a plurality of electrode channels, the plurality of electrodes comprising a reference electrode and a non-reference electrode; the EEG signal is acquired by the plurality of electrodes and transmitted to the processing module through the plurality of electrode channels; the processing module is specifically configured to: receive the EEG signal and screen out available EEG signal, the available EEG signal being EEG data of an undamaged non-reference electrode; perform denoising processing on the available EEG signal to obtain available EEG data; and slice the available EEG data through a sliding window of a preset size to obtain a plurality of EEG input data of the target to be analyzed.

[0011] With reference to the first aspect, in some embodiments, the electroencephalogram signal analysis device further comprises a prediction module configured to output an electroencephalogram signal analysis conclusion of the target to be analyzed according to the analysis result and the visualization graph, the electroencephalogram signal analysis conclusion comprising characteristics of each brain region of the target to be analyzed and a possibility of the target to be analyzed having Parkinson's disease.

[0012] With reference to the second aspect, in some embodiments, the electroencephalogram signal analysis method comprises: obtaining an electroencephalogram signal of a target to be analyzed; preprocessing the electroencephalogram signal to obtain a plurality of electroencephalogram input data; inputting the plurality of electroencephalogram input data into a trained electroencephalogram analysis model to obtain an analysis result output by the electroencephalogram analysis model.

[0013] With reference to the second aspect, in some embodiments, the training process of the electroencephalogram analysis model comprises: obtaining sample electroencephalogram signals of a plurality of sample analysis targets, the sample analysis targets comprising normal sample analysis targets and diseased sample analysis targets; a true label corresponding to the diseased sample analysis target being a first classification value, and a true label corresponding to the normal sample analysis target being a second classification value; preprocessing the sample electroencephalogram signals of each sample analysis target to obtain a plurality of sample input data of each sample analysis target; and inputting the plurality of sample input data into the electroencephalogram analysis model for training.

[0014] With reference to the third aspect, in some embodiments, the electronic device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the electroencephalogram signal analysis method of any one of the second aspect when executing the computer program.

[0015] With reference to the fourth aspect, in some embodiments, the computer readable storage medium stores a computer program, and the computer program is executable on a processor to implement the electroencephalogram signal analysis method of any one of the second aspect.

[0016] With reference to the fifth aspect, in some embodiments, the computer program product, when executed on an electronic device, causes the electronic device to perform the electroencephalogram signal analysis method of any one of the second aspect.

[0017] Compared with the related art, the embodiments of the present application have the following beneficial effects:

[0018] The embodiment of the application provides a brain electrical signal analysis device, which comprises a collection module, a processing module, an analysis module and a display module. The brain electrical signal of an analysis target is collected through the collection module, the brain electrical signal is preprocessed through the processing module, a plurality of brain electrical input data are obtained, then the plurality of brain electrical input data are input into a trained brain electrical analysis model of the analysis module, and an analysis result output by the brain electrical analysis model is obtained. In this way, the processed brain electrical input data can be analyzed through the trained brain electrical analysis model, so that whether the analysis target has a possibility of Parkinson's disease can be found in time and accurately.

[0019] It can be understood that the beneficial effects of the second aspect to the fifth aspect can be referred to the related description in the first aspect, which will not be repeated here.

[0020] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the specification. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed to be used in the embodiments or related technical description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 is an application scenario schematic diagram of the brain electrical signal analysis device provided by an embodiment of the application;

[0023] Figure 2 is a structural schematic diagram of the brain electrical signal analysis device provided by an embodiment of the application;

[0024] Figure 3 is an architecture schematic diagram of the brain electrical analysis model provided by an embodiment of the application;

[0025] Figure 4 is a schematic diagram of the matrix of the brain electrical input data provided by an embodiment of the application;

[0026] Figure 5 is a structural schematic diagram of the brain electrical signal analysis device provided by another embodiment of the application;

[0027] Figure 6 is a schematic diagram of the generation process of the visualized graph provided by an embodiment of the application;

[0028] Figure 7 is an average heat map corresponding to a normal person test sample provided by an embodiment of the application;

[0029] Figure 8 is an average heat map corresponding to a test sample of a Parkinson's disease patient provided by an embodiment of the present application;

[0030] Figure 9 is a flowchart of a brain electrical signal analysis method provided by an embodiment of the present application;

[0031] Figure 10 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0032] In the following description, specific details are set forth, such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, persons skilled in the art will understand that the present application can be practiced without these specific details. In other instances, well-known structures, devices, circuits, and methods have not been described in detail in order to avoid obscuring the description of the present application.

[0033] It should be understood that the term "comprises" as used in the specification and the appended claims indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0034] It should also be understood that the term "and / or" as used in the specification and the appended claims, means any one or more of the associated listed items, as well as all possible combinations of the items.

[0035] As used in the specification and the appended claims, the term "if" can be interpreted as meaning "when" or "upon" or "in response to a determination" or "in response to detecting" depending on the context. Similarly, the phrase "if determined" or "if detected [the described condition or event]" can be interpreted as meaning "upon determining" or "in response to determining" or "upon detecting [the described condition or event]" or "in response to detecting [the described condition or event]" depending on the context.

[0036] In addition, in the description of the specification and the appended claims, the terms "first", "second", "third", etc. are only used for differentiation in description, and cannot be understood as indicating or implying relative importance.

[0037] Reference within the specification of this application to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," and the like in the specification of the application do not necessarily all refer to the same embodiment, but are intended to be understood as meaning that "one or more but not all embodiments of the application." The terms "including," "comprising," "having," and variations thereof are meant to encompass the items listed thereafter and equivalents thereof as well as additional items.

[0038] The diagnosis of Parkinson's disease usually relies on experienced doctors to diagnose whether a subject has Parkinson's disease or the severity of Parkinson's disease by asking the subject to walk, speak, write, etc. However, the diagnosis by doctors has the following problems: first, the diagnosis result is relatively subjective, and the experience level of different doctors may be different or even opposite for the diagnosis result of the same patient; second, the intervention space is limited after the diagnosis, and when the patient is diagnosed by this way, the patient has usually lost more than 60% dopamine neurons, which greatly delays the treatment time of the patient; third, it cannot be diagnosed in time, the above diagnosis method is mainly aimed at the motor symptoms of Parkinson's disease, and the non-motor symptoms may have appeared before the motor symptoms appear, but the concealment of the non-motor symptoms makes the patient not easy to notice or be considered as other diseases, so it cannot be diagnosed in time and loses the best treatment period and sufficient treatment time.

[0039] In the related art, electroencephalogram, CT (Computed Tomography), magnetic resonance imaging and other medical instruments can assist in the diagnosis of Alzheimer's disease, depression and other diseases, and provide an objective and rapid diagnosis method for related diseases. However, the technology of the auxiliary diagnosis instrument for Parkinson's disease is still not mature.

[0040] Based on the above problems, the embodiments of the present application provide an electroencephalogram signal analysis device, which comprises a collection module, a processing module and an analysis module display module, first acquires the electroencephalogram signal of the target to be analyzed through the collection module, then pre-processes the electroencephalogram signal through the processing module to obtain a plurality of electroencephalogram input data, and then inputs the plurality of electroencephalogram input data into the trained electroencephalogram analysis model in the analysis module to obtain the analysis result output by the electroencephalogram analysis model. In this way, the processed electroencephalogram input data can be analyzed by the trained electroencephalogram analysis model, so that whether the target to be analyzed has the possibility of Parkinson's disease can be found in time and accurately.

[0041] For example, the embodiments of the present application can be applied toFigure 1 In the example scenario shown. In this scenario, the target to be analyzed A 110 can be a patient with pre-existing non-motor symptoms of Parkinson's disease, a patient with pre-existing motor symptoms of Parkinson's disease, or a normal person in a physical examination or consultation scenario, or a subject participating in a medical auxiliary instrument test. The display device C 130 can be a related medical image display device, or a computer, a projector, a mobile phone, or other device with display function.

[0042] In the above application scenarios, the target to be analyzed A 110 can directly wear the electroencephalogram signal analysis device B 120 on the head, or the electroencephalogram signal analysis device B 120 is connected to the brain of the target to be analyzed A 110 in other forms. The electroencephalogram signal analysis device B 120 is connected to the display device. The electroencephalogram signal of the target to be analyzed A 110 can be collected and analyzed by the electroencephalogram signal analysis device B 120, and finally the analysis result is displayed by the display device C 30. The displayed analysis result can include a text type conclusion, a digital type conclusion, and a graphical type conclusion.

[0043] The electroencephalogram signal analysis device in the embodiment of the present application can also be applied to a personal health detection scene. The electroencephalogram signal of a person is analyzed by the electroencephalogram signal analysis device to make a simple and qualitative judgment, help the user to understand the health status of the person in time, and facilitate timely medical consultation.

[0044] The electroencephalogram signal analysis device in the embodiment of the present application can also be applied to a clinic and the like. Through analysis of the electroencephalogram signal of a patient, guidance is provided for the patient.

[0045] The electroencephalogram signal analysis device in the embodiment of the present application can also be applied to a hospital. In combination with a large computer, the electroencephalogram signal is analyzed and judged multiple times, and other diagnostic methods and medical equipment in the medical system are comprehensively used to give the final diagnosis result and opinion of the patient. On the other hand, the electroencephalogram signal of a confirmed patient can be collected regularly and analyzed as a means of follow-up consultation to help doctors better analyze the cause and condition of the disease.

[0046] The electroencephalogram signal analysis device in the embodiment of the present application is described in detail as follows in combination with other drawings of the specification:

[0047] Figure 2 is a structural schematic diagram of the electroencephalogram signal analysis device 200 provided by an embodiment of the present application, which is described with reference to Figure 2The electroencephalogram signal analysis device 120 can include a collection module 210, a processing module 220, and an analysis module 230. The collection module is configured to collect an electroencephalogram signal of a target to be analyzed and transmit the electroencephalogram signal to the processing module. The processing module is configured to receive the electroencephalogram signal, pre-process the electroencephalogram signal to obtain a plurality of electroencephalogram input data, and transmit the plurality of electroencephalogram input data to the analysis module. The analysis module is provided with a trained electroencephalogram analysis model, and is configured to receive the plurality of electroencephalogram input data, input the plurality of electroencephalogram input data into the electroencephalogram analysis model, and obtain an analysis result output by the electroencephalogram analysis model, the analysis result representing a possibility that the target to be analyzed has Parkinson's disease.

[0048] In some embodiments, the electroencephalogram analysis model described above can be a trained convolutional neural network model.

[0049] Optionally, referring to Figure 3 The architecture of the electroencephalogram analysis model 300 described above can include a first convolutional module 310, a second convolutional module 320, a third convolutional module 330, a fourth convolutional module 340, a fifth convolutional module 350, a first fully connected layer 360, a second fully connected layer 370, and a SoftMax layer 380 connected in sequence.

[0050] Optionally, the first convolutional module includes a first convolutional layer, a first batch normalization layer, and a first pooling layer. The second convolutional module includes a second convolutional layer, a second batch normalization layer, and a second pooling layer. The third convolutional module includes a third convolutional layer, a third batch normalization layer, and a third pooling layer. The fourth convolutional module includes a fourth convolutional layer, a fourth batch normalization layer, and a fourth pooling layer. The fifth convolutional module includes a fifth convolutional layer, a fifth batch normalization layer, and a fifth pooling layer.

[0051] Optionally, the first convolutional layer uses a first convolutional kernel of 128; the second convolutional layer uses a second convolutional kernel and the third convolutional layer uses a third convolutional kernel, both of which are 64; the fourth convolutional layer uses a fourth convolutional kernel and the fifth convolutional layer uses a fifth convolutional kernel, both of which are 32. The size of the first convolutional kernel, the second convolutional kernel, and the third convolutional kernel is 5x1, the size of the fourth convolutional kernel is 3x1, and the size of the fifth convolutional kernel is 2x1. The most number of convolutional kernels is used in the first convolutional layer, which can extract a sufficient number of features of the electroencephalogram signal, and these features contain the basic characteristics of the electroencephalogram signal, which can be referred to as low-level features of the electroencephalogram signal. With the deepening of the network, the features of the electroencephalogram signal are gradually extracted, and the number and size of the convolutional kernels affect the level of these features. In the fifth convolutional layer of the fifth convolutional module, deeper features of the electroencephalogram signal can be extracted, and these features contain information about the impact of Parkinson's disease on the brain, which can be referred to as high-level features extracted by the fifth convolutional layer.

[0052] Optionally, the activation function of the first batch normalization layer, the second batch normalization layer, the third batch normalization layer, the fourth batch normalization layer and the fifth batch normalization layer is LeakyReLU. The batch normalization layer accelerates the training of the above-mentioned convolutional neural network, and the setting of the batch normalization layer can improve the accuracy and generalization ability of the above-mentioned convolutional neural network.

[0053] Optionally, the first pooling layer, the second pooling layer, the third pooling layer, the fourth pooling layer and the fifth pooling layer are all max pooling layers, and the size of the convolution kernel is 2x1. Since the size of the electroencephalogram input data is time dimension dataxspace dimension data, the above-mentioned max pooling layer can downsample the input electroencephalogram data in the time dimension, prevent overfitting of the above-mentioned convolutional neural network, and further improve the extraction ability of the convolutional neural network for the key features of the electroencephalogram signal.

[0054] Optionally, the number of neurons of the first full connection layer is 16, and the probability of dropout is 0.3. The number of neurons of the second full connection layer is 8, and the probability of dropout is 0.2. After the processing of the above-mentioned five convolution modules, the features reflecting the influence of Parkinson's disease on the brain in the electroencephalogram input data have been extracted, so considering the miniaturization of the network, setting a small number of neurons in the full connection layer can achieve the expected effect.

[0055] Optionally, the number of neurons of the SoftMax layer is 2. As the last layer of the above-mentioned convolutional neural network, the SoftMax layer normalizes the results analyzed in the front to a probability, realizes the classification of the analysis results through two neurons, and finally obtains clear and explicit analysis results. The analysis result can include a first classification value and a second classification value, the first classification value indicating that the current analyzed electroencephalogram input data corresponds to a target to be analyzed that may have Parkinson's disease, and the second classification value indicating that the current analyzed electroencephalogram input data corresponds to a target to be analyzed that does not have Parkinson's disease.

[0056] Optionally, in the training process of the above-mentioned neural network, the loss function used when the network performs deep learning is a cross-entropy loss function:

[0057]

[0058] where i is the i-th sample, y i is the true label of the i-th sample, is the predicted label of the i-th sample, and N is the total number of samples.

[0059] The EEG analysis model in the embodiment of the application contains 18 processing layers, and uses EEG input data in the time dimension and the space dimension as input data of the model, so that not only the time information of the EEG signal is considered, but also the space information is considered. Through the attribute setting of each convolution layer, batch normalization layer, pooling layer and full connection layer, network miniaturization is realized while preventing network overfitting and underfitting, the robustness of the network is improved, and the accuracy of the analysis result is improved.

[0060] Optionally, the data set used in the training of the convolutional neural network can include resting EEG data of a plurality of Parkinson's disease patients and a plurality of normal people in an open-eye condition, wherein the number of Parkinson's disease patients and the number of normal people are the same. In order to make the analysis result of the EEG analysis model more accurate, the following conditions are set for the data in the data set:

[0061] 1. The distribution of the data of the Parkinson's disease patient group and the data of the normal person group is uniform and symmetrical enough.

[0062] 2. The collection time is greater than a first preset time, and the number of electrodes and electrode channels is greater than a first preset number.

[0063] 3. The collected resting EEG data is original data without artificial screening or computer processing.

[0064] Table 1 is related information of the training data set of the EEG analysis model provided by an embodiment of the application. Referring to Table 1, the number of samples, age, gender and disease age in the data set are as follows:

[0065] Table 1

[0066]

[0067] Among them, moCA is the result of the Montreal Cognitive Assessment, and UODRS III is the result of the Unified Parkinson's Disease Rating Scale.

[0068] In some embodiments, the collection module 210 can include a plurality of electrodes and a plurality of electrode channels, and the plurality of electrodes include a reference electrode and a non-reference electrode. The EEG signal is collected by the plurality of electrodes and transmitted to the processing module 220 through the plurality of electrode channels.

[0069] In an application scenario of the embodiment of the application, a plurality of electrodes can be attached to a plurality of points on the head of the target to be analyzed to collect EEG signals of each brain region of the target to be analyzed.

[0070] Optionally, the number of the electrodes is 64, and the number of the electrode channels is 64. The sampling frequency of the electrodes is 512 Hz, and the continuous sampling time length of a set of electroencephalogram signals is 155 seconds. In the embodiments of the present application, the positions of the electrodes are set according to the International 10-20 System for the Recording of the Electroencephalogram, and the electrode names are also based on the International 10-20 System for the Recording of the Electroencephalogram.

[0071] Optionally, the Pz electrode is a reference electrode, and the other 63 electrodes are non-reference electrodes.

[0072] In some embodiments, the processing module 220 can be specifically configured to receive electroencephalogram signals, and screen out available electroencephalogram signals, which are electroencephalogram data of non-reference electrodes without damage. The available electroencephalogram signals are subjected to denoising processing to obtain available electroencephalogram data. The available electroencephalogram data are sliced by a sliding window of a preset size to obtain a plurality of electroencephalogram input data of an analysis target.

[0073] In a possible implementation, the implementation process of receiving electroencephalogram signals and screening out available electroencephalogram signals can include: receiving a set of electroencephalogram signals transmitted by 64 electrode channels, the set of electroencephalogram signals including all electroencephalogram data of 64 electrode positions. The electroencephalogram data of the reference electrode are removed, and the electroencephalogram data of the corresponding electrodes with data damage are also removed, and the remaining available electroencephalogram signals, i.e., all electroencephalogram data of non-reference electrodes without damage, are screened out. In this implementation, the available electroencephalogram signals include electroencephalogram data of 60 electrodes.

[0074] Optionally, the denoising processing of the available electroencephalogram signals includes: removing ocular artifact noise in the electroencephalogram signals due to blinking by independent component analysis.

[0075] In a possible implementation, after the available electroencephalogram signals are subjected to denoising processing to obtain available electroencephalogram data, the available electroencephalogram data are no longer subjected to high-pass filtering or low-pass filtering processing, so that more original electroencephalogram data can be protected as much as possible to protect the time-space information.

[0076] In some embodiments, the implementation process of slicing the available electroencephalogram data by a sliding window of a preset size to obtain a plurality of electroencephalogram input data of an analysis target can include: adopting a sliding window with a length of 2 seconds and setting an overlap rate of 50%, slicing the available electroencephalogram data to obtain electroencephalogram input data with a sampling point of 1024 in the time dimension and 60 electrode channels in the space dimension, i.e., a data block with a size of 1024 x 60 after slicing.

[0077] In the above embodiment, since the sampling frequency of the electrode is 512 Hz, a 2-second sliding window can extract 512*2=1024 sample data. For a set of EEG signals with a length of 155 seconds, data slicing is performed at an overlap rate of 50%, and 155-1=154 data blocks, i.e., 154 EEG input data, can be obtained. In order to preserve the complete spatial information of the EEG signal, no data processing is performed in the spatial dimension in the embodiment of the present application.

[0078] Through the above slicing processing, the excessive available EEG data can be converted into data convenient for the EEG analysis model processing. The sliding window is 2 seconds, which is consistent with the blinking period of a person and is conducive to feature extraction. Setting the overlap rate can offset the phase shift effect in the data, and also increase the number of EEG input data without data enhancement. Before training the convolutional neural network, the sample EEG signals also need to be processed in the above manner. If a total of 28 sample EEG signals are collected, these EEG signals can be converted into 28*154=4312 EEG input data after processing, which is sufficient to train the small convolutional neural network (EEG analysis model) in the embodiment of the present application, and can avoid the network overfitting problem caused by too few training samples.

[0079] Figure 4 is a schematic diagram of the matrix of the EEG input data provided by an embodiment of the present application, referring to Figure 4 The rows of the matrix represent the sample data in the time dimension, the columns of the matrix represent the sample data of different electrode channels, and the depth of the color represents the size of the voltage value corresponding to the sample data.

[0080] In some embodiments, the analysis module 230 can be specifically configured to: receive a plurality of EEG input data. The plurality of EEG input data is input into the EEG analysis model, the features of the plurality of EEG input data are gradually extracted through a plurality of convolution modules, the features extracted by the last set of convolution modules are weighted, fused and classified and predicted through a plurality of fully connected layers, and finally the analysis result corresponding to the EEG signal of the target to be analyzed is obtained.

[0081] Optionally, the analysis module 230 can be configured to receive a plurality of electroencephalogram input data. The plurality of electroencephalogram input data is input into the electroencephalogram analysis model. Low-level features in the electroencephalogram input data are extracted by a first convolution layer in the first convolution module. Then, the features are extracted step by step by a second convolution layer in the second convolution module, a third convolution layer in the third convolution module, a fourth convolution layer in the fourth convolution module, and a fifth convolution layer in the fifth convolution module. The output data of the fifth convolution layer is a high-level feature of the electroencephalogram input data. Then, the output data of the fifth pooling layer in the fifth convolution module is weighted and fused by the first full connection layer, and classified to obtain first fused features. The first fused features are classified by the second full connection layer to obtain second fused features. Finally, the SoftMax layer is used for classification prediction to obtain an analysis result corresponding to the electroencephalogram signal of the target to be analyzed.

[0082] Optionally, if the analysis result is 0, the target to be analyzed has no possibility of Parkinson's disease, i.e., it can be preliminarily determined that the target to be analyzed is not a Parkinson's disease patient. If the analysis result is 1, the target to be analyzed has a possibility of Parkinson's disease, i.e., it can be preliminarily determined that the target to be analyzed is a Parkinson's disease patient.

[0083] Referring to Figure 5 In some embodiments, the analysis result is determined based on Figure 2 In the embodiment shown, the electroencephalogram signal analysis device 200 can further include a display module 240 configured to receive the intermediate layer output data of at least one specified pooling layer output sent by the analysis module, and display a visualized graph of the intermediate layer output data of the at least one specified pooling layer output.

[0084] In some embodiments, the display module 240 can be configured to, for each electroencephalogram input data, obtain a first heat vector of each specified pooling layer output in the electroencephalogram analysis model, the first heat vector being a vector corresponding to the intermediate layer output data of the specified pooling layer output, the first heat vector including a convolution channel dimension, a time dimension, and a space dimension. The first heat vector corresponding to each specified pooling layer is averaged in the convolution channel dimension and the time dimension to obtain a second heat vector corresponding to each specified pooling layer. For each specified pooling layer, the second heat vectors corresponding to the specified pooling layer in the plurality of electroencephalogram input data are averaged to obtain a third heat vector corresponding to the specified pooling layer. According to the third heat vector corresponding to each specified pooling layer, a visualized graph of the output data corresponding to the corresponding pooling layer is drawn.

[0085] In one possible implementation, the specified pooling layer is the fifth pooling layer, and the preset number is 50. Referring to Figure 6For the electroencephalogram input data with a size of 1024x60, the data outputted by the fifth pooling layer in the electroencephalogram analysis model, i.e., the intermediate layer output data, is obtained. The intermediate layer output data is taken as the first heat vector. According to the architecture of the electroencephalogram analysis model and the related information of each processing layer provided in the above embodiment, the first heat vector with a size of 32x30x60 can be obtained through the processing of the respective convolutional layers and pooling layers. The first dimension is the convolutional channel dimension, the second dimension is the time dimension, and the third dimension is the spatial dimension.

[0086] For the first heat vector, refer to Figure 6 , first, the average value is taken on the 32 convolutional channels to obtain a heat vector with a size of 30x60, and then the average value is taken on the 30 time sampling points to obtain a second heat vector with a size of 1x60.

[0087] For a group of electroencephalogram signals of a target to be analyzed, if the sampling time of the electrodes is 155 seconds, the sampling frequency is 512 Hz, and the electroencephalogram data of 60 electrode channels is retained, then according to the group of electroencephalogram signals, 154 electroencephalogram input data can be finally obtained. Through the processing of the electroencephalogram analysis model, 154 first heat vectors can be obtained, and then 154 second heat vectors can be obtained. Next, 50 second heat vectors are selected from the 154 second heat vectors, and a third heat vector is obtained by averaging the 50 second heat vectors, as shown in Figure 6 , the size of the third heat vector is still 1x60. According to the third heat vector, in combination with the positions of the electrodes, a visualized graph of the output data corresponding to the fifth pooling layer is drawn.

[0088] In another possible implementation, the above-mentioned 154 second heat vectors can also be normalized and mapped to the positions of the corresponding electrodes, and then another visualized graph of the output data corresponding to the fifth pooling layer is drawn. Through the heat map, the brain regions or electrodes related to Parkinson's disease can be displayed.

[0089] In the above embodiment, a visualized graph can also be drawn directly according to the second heat vector in combination with the positions of the electrodes. However, the visualized graph obtained from the second heat vector of only one electroencephalogram input data cannot reflect the real situation of the electroencephalogram signals of the target to be analyzed, and the displayed second heat vector can be a special case. Therefore, by averaging a plurality of second heat vectors, a more actual and comprehensive electroencephalogram signal feature graph can be obtained. Moreover, the result obtained by averaging is more macroscopic than that obtained by normalization, which is convenient for drawing and observing the graph.

[0090] Figure 6The graph in the upper right corner is a heat map corresponding to the second heat vector according to an embodiment of the present application, Figure 6 The graph in the lower right corner is an average heat map corresponding to the third heat vector according to the embodiment. Referring to Figure 6 The heat map and the average heat map are obtained by processing the electroencephalogram of the same target to be analyzed, but the brain regions corresponding to the characteristic reactions in the two heat maps are quite different. Therefore, data with a too small time length cannot truly reflect the real situation of the electroencephalogram of the target to be analyzed, and a certain amount of data needs to be integrated to obtain a result closer to the actual situation.

[0091] Figure 7 is an average heat map corresponding to a normal test sample according to an embodiment of the present application, Figure 8 is an average heat map corresponding to a Parkinson's disease patient test sample according to an embodiment of the present application. Referring to Figure 7 For the analysis of the normal test sample, the characteristics extracted by the electroencephalogram analysis model are evenly distributed on the left and right sides of the brain, that is, the electrode values of the left and right brains of the normal person are small. Referring to Figure 8 For the analysis of the Parkinson's disease patient test sample, the characteristics extracted by the electroencephalogram analysis model are mainly concentrated on the left upper side, and the corresponding electrodes are 9 electrodes with FC5 as the center, and the corresponding brain structure is the intersection of the frontal lobe and the parietal lobe. The functions of the frontal lobe and the parietal lobe are to control the movement of the body, language and memory, etc., and have a high correlation with the symptoms of Parkinson's disease.

[0092] As shown in Figure 5 The electroencephalogram analysis device 200 can further include a prediction module 250, which can be configured to output an electroencephalogram analysis conclusion of the target to be analyzed according to the analysis result and the visualization graph, the electroencephalogram analysis conclusion including characteristics of each brain region of the target to be analyzed and a possibility of the target to be analyzed suffering from Parkinson's disease.

[0093] In one specific embodiment of the present application, the performance of the electroencephalogram analysis model in the above embodiment is evaluated. The accuracy Acc, precision Pre, recall Re and F1 score of the electroencephalogram analysis model are calculated according to the following formula:

[0094]

[0095]

[0096]

[0097]

[0098] Wherein, TP represents true positive, true positive represents the case that the sample of Parkinson's disease patient is correctly classified, that is, for the sample of Parkinson's disease patient, the predicted label and the real label are both 1. TN represents true negative, true negative represents the case that the sample of normal person is correctly classified, that is, for the sample of normal person, the predicted label and the real label are both 0. FP represents false positive, false positive represents the case that the sample of Parkinson's disease patient is incorrectly classified, that is, for the sample of Parkinson's disease patient, the predicted label is 0 and the real label is 1. FN represents false negative, false negative represents the case that the sample of normal person is incorrectly classified, that is, for the sample of normal person, the predicted label is 1 and the real label is 0.

[0099] Through calculation, the above-mentioned electroencephalogram analysis model has an accuracy of 96.91%, a precision of 99.79%, a recall of 93.97% and an F1 score of 96.79 for the binary classification of the electroencephalogram signals of the Parkinson's disease patients and the normal persons. In addition, the AUC is calculated as 0.996.

[0100] The electroencephalogram signal analysis device provided by the embodiment of the present application comprises a collection module, a processing module and an analysis module display module. The electroencephalogram signal of the target to be analyzed is collected by the collection module, and the electroencephalogram signal is preprocessed by the processing module to obtain a plurality of electroencephalogram input data. Then, the plurality of electroencephalogram input data is input into the trained electroencephalogram analysis model in the analysis module to obtain the analysis result output by the electroencephalogram analysis model. In this way, the processed electroencephalogram input data can be analyzed by the trained electroencephalogram analysis model, so that whether the target to be analyzed has the possibility of Parkinson's disease can be found in time and accurately.

[0101] For the electroencephalogram signal analysis device described in the above embodiment, Figure 9 The flowchart of the electroencephalogram signal analysis method provided by the embodiment of the present application is shown, and the electroencephalogram signal analysis method is described as follows:

[0102] S901, acquiring the electroencephalogram signal of the target to be analyzed.

[0103] S902, preprocessing the electroencephalogram signal to obtain a plurality of electroencephalogram input data.

[0104] S903, inputting the plurality of electroencephalogram input data into the trained electroencephalogram analysis model to obtain the analysis result output by the electroencephalogram analysis model.

[0105] In some embodiments, the training process of the above-mentioned electroencephalogram analysis model can comprise:

[0106] S1, acquire sample electroencephalogram signals of a plurality of sample analysis targets, the sample analysis targets including normal sample analysis targets and diseased sample analysis targets; the real labels of the diseased sample analysis targets are first classification values, and the real labels of the normal sample analysis targets are second classification values.

[0107] S2, respectively pre-process the sample electroencephalogram signals of each sample analysis target to obtain a plurality of sample input data of each sample analysis target.

[0108] S3, input the plurality of sample input data to an electroencephalogram analysis model for training.

[0109] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0110] The embodiments of the present application also provide an electronic device, referring to Figure 10 The electronic device 1000 can include at least one processor 1010, a memory 1020, and a computer program stored in the memory 1020 and executable on the at least one processor 1010, wherein the processor 1010 implements the steps in any of the above method embodiments when executing the computer program, for example Figure 9 Steps S901 to S903 in the illustrated embodiments.

[0111] For example, the computer program can be divided into one or more modules / units, one or more modules / units are stored in the memory 1020 and executed by the processor 1010 to complete the present application. The one or more modules / units can be a series of computer program segments capable of completing a specific function, which are used to describe the execution process of the computer program in the electronic device 1000.

[0112] Those skilled in the art can understand, Figure 10 It is only an example of an electronic device and does not constitute a limitation on the electronic device, which can include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, buses, etc.

[0113] The processor 1010 can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0114] The memory 1020 can be an internal storage unit of the electronic device, and can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. The memory 1020 is used to store the computer program and other programs and data required by the electronic device. The memory 1020 can also be used to temporarily store data that has been output or will be output.

[0115] The bus can be an industry standard architecture (ISA) bus, a peripheral component (PCI) bus, an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, the bus in the drawings of the present application does not limit to only one bus or one type of bus.

[0116] The electroencephalogram signal analysis method provided by the embodiments of the present application can be applied to a computer, a wearable device, a vehicle-mounted device, a tablet computer, a notebook computer, a netbook, a personal digital assistant (PDA), an augmented reality (AR) / virtual reality (VR) device, a mobile phone and other electronic devices. The embodiments of the present application do not make any limitation on the specific type of the electronic device.

[0117] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to realize the steps in each of the above-mentioned embodiments of the electroencephalogram signal analysis method.

[0118] The embodiment of the present application provides a computer program product. When the computer program product is run on a mobile terminal, the mobile terminal is caused to execute the steps in each of the above-mentioned embodiments of the electroencephalogram signal analysis method.

[0119] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the embodiment of the present application realizes all or part of the processes in the above-mentioned embodiments, which can be completed by a computer program instructing related hardware. The computer program can be stored in a computer readable storage medium. The computer program is executed by a processor to realize the steps in each of the above-mentioned embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file or some intermediate form. The computer readable medium at least includes any entity or device capable of carrying the computer program code to the photographing device / electronic device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk and the like.

[0120] In the above-mentioned embodiments, the description of each embodiment has its own focus. The parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0121] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0122] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / network device and method can be implemented in other manners. For example, the embodiments of the apparatus / network device described above are merely illustrative. For example, the division of the modules or units is merely logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0123] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0124] The above-described embodiments are merely used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalent replacements; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. An electroencephalogram signal analysis device, characterized by, The device comprises a collection module, a processing module and an analysis module. The collection module is configured to collect electroencephalogram signals of a target to be analyzed and transmit the electroencephalogram signals to the processing module. The processing module is configured to receive the electroencephalogram signals, pre-process the electroencephalogram signals to obtain a plurality of electroencephalogram input data, and transmit the plurality of electroencephalogram input data to the analysis module. The analysis module is internally provided with a trained electroencephalogram analysis model, and is configured to receive the plurality of electroencephalogram input data, input the plurality of electroencephalogram input data into the electroencephalogram analysis model, and obtain an analysis result output by the electroencephalogram analysis model, wherein the analysis result represents a possibility that the target to be analyzed has Parkinson's disease. The electroencephalogram analysis model comprises a plurality of convolution modules, each of which comprises a convolution layer, a batch normalization layer and a pooling layer, and the output data of each pooling layer is an intermediate layer output data. The electroencephalogram signal analysis device further comprises a display module configured to receive at least one intermediate layer output data output by a specified pooling layer of the analysis module, and display a visual graph of the at least one intermediate layer output data output by the specified pooling layer. The display module is specifically configured to: For each electroencephalogram input data, obtain a first heat vector output by each specified pooling layer in the electroencephalogram analysis model, wherein the first heat vector is a vector corresponding to the intermediate layer output data output by the specified pooling layer, and the first heat vector comprises a convolution channel dimension, a time dimension and a space dimension; and average the first heat vector corresponding to each specified pooling layer in the convolution channel dimension and the time dimension to obtain a second heat vector corresponding to each specified pooling layer. For each specified pooling layer, average a preset number of second heat vectors corresponding to the specified pooling layer in the plurality of electroencephalogram input data to obtain a third heat vector corresponding to the specified pooling layer. According to the third heat vector corresponding to each specified pooling layer, a visual graph of the output data corresponding to the respective pooling layer is drawn.

2. The electroencephalographic signal analysis device of claim 1, wherein, The electroencephalogram analysis model comprises a plurality of fully connected layers and a SoftMax layer arranged in sequence after the plurality of convolution modules; and the analysis module is specifically configured to: receive the plurality of electroencephalogram input data; input the plurality of electroencephalogram input data into the electroencephalogram analysis model, gradually extract features of the plurality of electroencephalogram input data through the plurality of convolution modules, and perform weighted fusion and classification prediction on the features extracted by the last convolution module through the plurality of fully connected layers and the SoftMax layer, to finally obtain an analysis result corresponding to the electroencephalogram signals of the target to be analyzed.

3. The electroencephalographic signal analysis device of claim 1 or 2, wherein The collection module comprises a plurality of electrodes and a plurality of electrode channels, and the plurality of electrodes comprises a reference electrode and a non-reference electrode. The electroencephalogram signals are collected by the plurality of electrodes and transmitted to the processing module through the plurality of electrode channels, and the processing module is specifically configured to: receive the electroencephalogram signals and screen out usable electroencephalogram signals, wherein the usable electroencephalogram signals are electroencephalogram data of non-damaged non-reference electrodes. The available electroencephalogram signal is denoised to obtain available electroencephalogram data; The available electroencephalogram data is sliced by a preset size sliding window to obtain a plurality of electroencephalogram input data of the target to be analyzed.

4. The electroencephalographic signal analysis device of claim 1, wherein, The electroencephalogram signal analysis device further comprises: The prediction module is configured to output an electroencephalogram signal analysis conclusion of the target to be analyzed according to the analysis result and the visualization figure, wherein the electroencephalogram signal analysis conclusion comprises characteristics of each brain region of the target to be analyzed and a possibility of the target to be analyzed suffering from Parkinson's disease.

5. An electroencephalogram signal analysis method characterized by, The method is applied to the electroencephalogram signal analysis device according to any one of claims 1-4, and the method comprises: obtaining an electroencephalogram signal of a target to be analyzed; preprocessing the electroencephalogram signal to obtain a plurality of electroencephalogram input data; inputting the plurality of electroencephalogram input data into the trained electroencephalogram analysis model to obtain an analysis result output by the electroencephalogram analysis model.

6. The electroencephalogram signal analysis method of claim 5, wherein, The training process of the electroencephalogram analysis model comprises: obtaining sample electroencephalogram signals of a plurality of sample analysis targets, wherein the sample analysis targets comprise normal sample analysis targets and diseased sample analysis targets; a true label corresponding to the diseased sample analysis target is a first classification value, and a true label corresponding to the normal sample analysis target is a second classification value; respectively preprocessing the sample electroencephalogram signals of the sample analysis targets to obtain a plurality of sample input data of the sample analysis targets; inputting the plurality of sample input data into the electroencephalogram analysis model for training.

7. An electronic device comprising a memory and a processor, said memory having stored therein a computer program operable on said processor, characterized in that, The processor executes the computer program to implement the method according to any one of claims 5-6.

8. A computer-readable storage medium storing a computer program, the computer-readable storage medium comprising: The computer program is executed by the processor to implement the method according to any one of claims 5-6.

Images