Psychological disease diagnosis and identification method, device and equipment

Abstract

The invention relates to the technical field of psychological disciplines, discloses a psychological disease automatic diagnosis method, and particularly relates to a psychological type disease diagnosis and recognition method, device and equipment. The method comprises the following steps: extracting differential entropy characteristics of a plurality of electroencephalogram signals, performing superposition enhancement on the differential entropy characteristics, constructing initial graph data based on the differential entropy characteristics and a graph construction rule, and constructing the initial graph data into an optimal coding tree to obtain a clustering distribution matrix corresponding to nodes; and inputting the initial graph data and the clustering distribution matrix into a recognition model to obtain a recognition result. According to the method, the electroencephalogram signals are subjected to computerization processing to realize automatic diagnosis, and the objectivity and the accuracy of psychological fatigue disease diagnosis are improved.

Description

Technical Field

[0001] The present application relates to the technical field of psychology, and is an automated method for diagnosing mental illness, specifically a method, device and equipment for diagnosing and identifying mental illnesses. Background Art

[0002] Psychological diseases, especially depression, are the most common type of psychological diseases in modern people. Psychological changes are based on the physiological activities of the brain. The occurrence of mental illness indicates that the structure or function of the brain has changed. Depression, in turn, may lead to greater stress and dysfunction, affecting the patient's daily life and aggravating the symptoms of the disease. The prevalence of depression is on the rise. Faced with the growing prevalence, the diagnosis and treatment rates of depression are far from meeting the coping level. Only 9.5% have received services related to mental illness, and only 0.5% have received adequate treatment. Objective diagnostic methods, lack of resources, and social discrimination against mental illness affect the implementation of effective treatment. Traditional depression diagnosis methods are mainly based on depression detection scales and doctor interviews, which are highly subjective and the results are often not accurate enough. Therefore, the development of reliable, objective and advanced methods is the key to improving the accuracy of clinical diagnosis, and can also further assist doctors in making decisions. Summary of the invention

[0003] In order to solve the above technical problems, the embodiments of the present application provide a method, device and equipment for diagnosing and identifying psychological diseases, which realizes automated diagnosis by computerizing EEG signals, thereby improving the objectivity and accuracy of psychological disease diagnosis.

[0004] In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0005] In a first aspect, a method for diagnosing and identifying psychological diseases is provided, the method being applied to a psychological diagnosis system, the psychological diagnosis system comprising an electroencephalogram (EEG) signal acquisition device and a server, the EEG signal acquisition device comprising a plurality of EEG acquisition electrode channels, the plurality of EEG acquisition electrode channels being used to acquire a plurality of EEG signals from a plurality of regions of a patient's head to be diagnosed, and a data terminal electrically connected to the EEG acquisition electrodes, the data terminal communicating with the server, the method being configured in the server; the method comprising: extracting differential entropy features of the plurality of EEG signals, superimposing and enhancing the differential entropy features, and based on the differential entropy features and the graph structure Establish rules to construct initial graph data, and construct the initial graph data into an optimal coding tree to obtain a cluster allocation matrix corresponding to the node; the initial graph data includes node features and an adjacency matrix; the initial graph data and the cluster allocation matrix are input into a recognition model to obtain a recognition result; the recognition model includes a feature extraction module and a classifier module, the feature extraction module is used to concatenate the node features, the adjacency matrix, and the cluster allocation matrix to obtain a feature vector, the classifier module is used to receive the feature vector and store the changes in the EEG signal time series, and convert the feature vector and output it to a linear layer for classification results.

[0006] Furthermore, the initial graph data is constructed based on the differential entropy features and graph construction rules, including: taking multiple EEG acquisition electrode channels as nodes, using the differential entropy features as node features, calculating Pearson coefficients between multiple EEG acquisition channels, and based on a preset Pearson coefficient threshold, connecting two channels whose values ​​are greater than the threshold, and not connecting channels whose values ​​are less than the threshold, determining the adjacency matrix based on the edge structure, and combining the node features and the adjacency matrix into initial graph data.

[0007] Furthermore, the coding tree includes a root node, an internal node and a leaf node; the initial graph data is constructed into an optimal coding tree, including: forming a plurality of initial coding trees based on the root node, the internal node and the leaf node, and determining the structural entropy of each of the initial coding trees based on a structural entropy minimization algorithm, and determining the coding tree corresponding to the minimum structural entropy as the optimal coding tree; the structural entropy is determined based on the following formula: Where V is the node set in the initial graph data G, and is also the initial leaf node set of the initial coding tree T. i is a non-root node in T, Yes i The parent node of v i The number of edges with endpoints in the leaf node partition of , vol(V) and vol(v i ) represent V and v respectively iThe degree of the leaf node.

[0008] Furthermore, the structural entropy minimization algorithm is used to determine the structural entropy of each of the initial coding trees, and the coding tree corresponding to the minimum structural entropy is determined as the optimal coding tree, including: specifying the height of the initial coding tree, the initial coding tree has only one root node and multiple leaf nodes, the number of leaf nodes is the same as the number of EEG acquisition electrode channels, and the leaf nodes of the initial coding tree are all child nodes of the root node; based on the merge sort algorithm, a full-height coding tree is generated in a bottom-up manner, and each iteration forms a new partition by merging two child nodes of the root node; and internal nodes are deleted to compress the height of the coding tree to obtain an intermediate coding tree; nodes in the intermediate coding tree that do not directly inherit nodes in the next layer are supplemented to obtain the optimal coding tree.

[0009] Furthermore, the initial graph data and the cluster allocation matrix are input into the recognition model to obtain the recognition result, including: inputting the initial graph data and the cluster allocation matrix into the feature extraction module for feature splicing, and reducing the dimension of the spliced ​​features to obtain a feature vector; inputting the feature vector into the classifier module, calculating and storing the brain wave features of the time series, and finally outputting the result to the linear layer to obtain the classification result.

[0010] Furthermore, the feature extraction module includes multiple graph neural convolution layers and corresponding multiple graph pooling layers, the first graph neural convolution layer is used to receive the node features and the adjacency matrix, and extract the features of the node features and the adjacency matrix, the first graph pooling layer is used to receive the output of the graph neural convolution layer and the clustering assignment matrix and form a new subgraph, and input the subgraph into the second graph neural convolution layer.

[0011] Furthermore, the rules of the graph neural convolution layer are as follows: in represents the adjacency matrix of the graph, I represents the identity matrix, express The degree matrix, W (l) represents a trainable weight matrix, σ() represents a nonlinear activation function, and H (l) ∈R N×D represents the feature matrix of layer l, H (0) is the node feature.

[0012] Furthermore, the classifier module is constructed based on an LSTM network.

[0013] In a second aspect, a psychological type disease diagnosis and identification device is provided, the device is applied to a psychological diagnosis system, the psychological diagnosis system includes an electroencephalogram signal acquisition device and a server, the electroencephalogram signal acquisition device includes a plurality of electroencephalogram acquisition electrode channels, the plurality of electroencephalogram acquisition electrode channels are used to acquire a plurality of electroencephalogram signals from a plurality of regions of a patient's head to be diagnosed, and a data terminal electrically connected to the electroencephalogram acquisition electrodes, the data terminal communicates with the server, and the device is configured in the server; the device includes: a brain map data construction module, used to extract differential entropy features of the plurality of electroencephalogram signals, and to superimpose and enhance the differential entropy features, based on Initial graph data is constructed based on the differential entropy features and graph construction rules, and the initial graph data is constructed as an optimal coding tree to obtain a cluster allocation matrix corresponding to the node; a recognition module is used to input the initial graph data and the cluster allocation matrix into a recognition model to obtain a recognition result; the recognition model includes a feature extraction module and a classifier module, the feature extraction module is used to splice the node features, the adjacency matrix, and the cluster allocation matrix to obtain a feature vector, the classifier module is used to receive the feature vector and store the changes in the EEG signal time series, and convert the feature vector and output it to a linear layer for classification results.

[0014] In a third aspect, a terminal device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any of the above-mentioned methods for diagnosing and identifying psychological diseases when executing the computer program.

[0015] In the technical solution provided by the embodiment of the present application, by extracting differential entropy features of multiple EEG signals, and superimposing and enhancing the differential entropy features, constructing initial graph data based on the differential entropy features and graph construction rules, and constructing the initial graph data into an optimal coding tree, a clustering allocation matrix corresponding to the node is obtained; the initial graph data and the clustering allocation matrix are input into the recognition model to obtain the recognition result. The present invention realizes automated diagnosis by computerizing EEG signals, thereby improving the objectivity and accuracy of diagnosis of mental illness. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] The methods, systems and / or programs in the accompanying drawings will be further described according to exemplary embodiments. These exemplary embodiments will be described in detail with reference to the drawings. These exemplary embodiments are non-limiting exemplary embodiments, wherein example numbers represent similar mechanisms in the various views of the accompanying drawings.

[0018] Figure 1 It is a schematic diagram of the structure of the psychological type disease diagnosis system provided in the embodiment of the present application.

[0019] Figure 2 It is a flowchart of the method for diagnosing and identifying psychological diseases provided in an embodiment of the present application.

[0020] Figure 3 It is a schematic diagram of the structure of the infusion management device provided in an embodiment of the present application.

[0021] Figure 4 It is a schematic diagram of the structure of the infusion management device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0023] In the following detailed description, numerous specific details are set forth by way of example in order to provide a comprehensive understanding of the relevant guidance. However, it will be apparent to those skilled in the art that the present application may be practiced without these details. In other cases, well-known methods, procedures, systems, compositions and / or circuits have been described at a relatively high level, without detail, in order to avoid unnecessarily obscuring aspects of the present application.

[0024] Flowcharts are used in the present application to illustrate the execution process performed by the system according to the embodiment of the present application. It should be clearly understood that the execution process of the flowchart may not be performed in order. On the contrary, these execution processes may be performed in reverse order or simultaneously. In addition, at least one other execution process may be added to the flowchart. One or more execution processes may be deleted from the flowchart.

[0025] Before further describing the embodiments of the present invention in detail, the nouns and terms involved in the embodiments of the present invention are described. The nouns and terms involved in the embodiments of the present invention are subject to the following interpretations.

[0026] (1) In response, it is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or with a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.

[0027] (2) Based on is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or have a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.

[0028] Depression is characterized by high morbidity, high disability and high mortality, but most people still lack scientific knowledge about depression, and even fewer people seek timely medical treatment, receive accurate diagnosis and treatment intervention. In addition, traditional diagnostic methods often rely on scales, such as the Hamilton Depression Rating Scale and the Beck Depression Inventory, combined with doctor inquiries for diagnosis. The effectiveness of this diagnostic method is based on the patient's honesty and full understanding of the scale. It is easily affected by the doctor's experience level and the patient's mood when seeking medical treatment. It contains many subjective factors, resulting in a relatively high misdiagnosis rate. Many patients are delayed in their condition, and some non-depressed people are mistakenly prescribed antidepressants. Therefore, finding a way to identify depression is an indispensable prerequisite for the formulation and effective implementation of treatment plans.

[0029] In recent years, the continuous development of some objective indicators has provided new ideas for the identification of depression, such as head movement, facial movement, eye movement, gait, speech and brain signals. Among them, brain signals are fast, economical and convenient, and are favored by many researchers and used for the identification of depression. However, in the existing technology, EEG-based depression brain function network analysis and depression identification model construction have achieved fruitful results, but the relevant research methods still have certain deficiencies and need to be improved and perfected.

[0030] The EEG data of patients with depression and normal people have different changing patterns in parameters such as frequency band, power, and amplitude, but some studies will obtain different results. This may be due to the influence of factors such as data preprocessing, coupling methods, and volume conduction on different connection measurements. If these factors can be well controlled, then mental illnesses including depression can obtain more accurate and consistent results from brain functional connectivity and network analysis.

[0031] The construction of depression recognition model is generally to identify the most discriminative features through specific feature evaluation methods as the feature set of classification task, so that EEG signals can be fully mined and used for classification decisions. Existing studies have extracted different types of features to characterize EEG signals, but there is no consensus on which feature contributes more to depression recognition model.

[0032] Feature selection methods can reduce feature dimensions and remove noise and redundancy, and classification methods can distinguish between patients with depression and normal people. However, in the face of individual differences in data sets, studying a robust depression identification model is more valuable for clinical applications.

[0033] Based on this, see Figure 1 The embodiment of the present invention provides a psychological disease diagnosis system 100, which is mainly used to diagnose depression in a patient to be diagnosed. The system includes an electroencephalogram signal acquisition device 120 and a server 110, wherein the electroencephalogram signal acquisition device includes a plurality of electroencephalogram acquisition electrode channels, the plurality of electroencephalogram acquisition electrode channels are used to acquire a plurality of electroencephalogram signals from a plurality of regions of the head of the patient to be diagnosed, and a data terminal electrically connected to the electroencephalogram acquisition electrode, the data terminal communicating with the server.

[0034] Specifically, in the embodiment of the present invention, the EEG signal acquisition device adopts an EEG acquisition instrument produced by Electrical Geodesics Inc (EGI), the EEG cap is a 128-lead HydroGelGeodesic Sensor Net, the impedance of each lead is below 50 kilo-ohms, and the amplifier is NetAmps300. The EEG acquisition software is Net Station 4.5 version, and the EEG sampling rate is 250Hz. The EEG electrodes are placed according to the international 10-20 system, and the Cz point is the reference electrode.

[0035] The EEG signal acquisition process is as follows: 1) Turn on the acquisition equipment and computer switch; 2) The patient cleans the scalp as required, measures the size of the patient's head, and chooses a hat of the corresponding size; 3) Choose a hat of the right size, put it in saline, massage and soak for 8 minutes; 4) Adjust the position of each electrode, and add a small amount of saline with a straw; 5) Formal acquisition, debug all leads to less than 50 kilo-ohms. Run the program to start recording EEG, and stop EEG recording when it ends. The data during the acquisition process is uploaded to the server according to the preset timing.

[0036] See also Figure 2 In an embodiment of the present invention, a method for diagnosing and identifying psychological diseases is configured in a server, and the method includes the following steps:

[0037] Step S210. Extract differential entropy features of multiple EEG signals, and superimpose and enhance the differential entropy features, construct initial graph data based on the differential entropy features and graph construction rules, and construct the initial graph data into an optimal coding tree to obtain a clustering allocation matrix corresponding to the node.

[0038] In the embodiment of the present invention, the initial graph data includes node features and an adjacency matrix. The embodiment of the present invention mainly constructs graph data, namely a mind map, wherein the initial graph data is a preliminarily constructed mind map.

[0039] Among them, the construction method for the initial graph data uses multiple EEG acquisition electrode channels as nodes. In the embodiment of the present application, the number of EEG acquisition electrode channels is 128, and the number of nodes for the initial graph data is 128. The differential entropy feature (DE feature) is extracted for each channel of the EEG signal to obtain a differential entropy feature matrix with a dimension of 128×9. And each time 4 time-continuous time series data are taken, the differential entropy features are superimposed to achieve the purpose of feature enhancement. After superposition, there are 40 time series signals and 40 differential entropy feature matrices.

[0040] Specifically, for the composition method in the embodiment of the present application, 128 channels are used as nodes; differential entropy is used as the node feature; for the 128 channels of the time series signal, the Pearson coefficient is calculated between each two, where the Pearson coefficient threshold is set to 0.6, and the two channels greater than the threshold are connected, and the channels less than the threshold are not connected, thereby determining the adjacency matrix.

[0041] After obtaining the graph data, it is necessary to construct the graph data into an optimal coding tree with a height of 3 through the structural entropy minimization algorithm, and then obtain the clustering allocation matrix of the nodes. Among them, the coding tree consists of a root node, internal nodes, and leaf nodes. Among them, the highest layer is the root node, which is generated when the initial coding tree is generated, and there is only one; the middle layer is composed of newly added internal nodes, which are generated by the structural entropy algorithm and can have multiple; the lowest layer is the leaf node, which has 128 nodes, derived from the 128 nodes in the graph data.

[0042] The coding tree can better reveal the connection between the pooling layers, establish connections between different pooling layers, and help eliminate the noise structure in the graph. According to the graph structure, multiple different coding trees can be constructed. The quality of the coding tree will directly determine the quality of the clustering allocation scheme used for hierarchical pooling. In order to obtain the best clustering allocation scheme, it is necessary to minimize the structural entropy as a basic principle throughout the algorithm for constructing the coding tree.

[0043] In the embodiment of the present invention, the structural entropy is determined based on the following formula:

[0044] Where V is the node set in the initial graph data G, and is also the initial leaf node set of the initial coding tree T. i is a non-root node in T, Yes i The parent node of v i The number of edges with endpoints in the leaf node partition of , vol(V) and vol(v i ) represent V and v respectively i The degree of the leaf node.

[0045] The optimal coding tree is realized through the structural entropy minimization algorithm. The minimum entropy is calculated by the algorithm and used as the structural entropy of the graph G. The coding tree with this minimum value is the optimal coding tree and also the global optimal solution of the clustering assignment algorithm.

[0046] The objective equation of the algorithm is:

[0047]

[0048] In addition to minimizing the structural entropy, in order to adapt to the number of pooling layers, it is usually necessary to specify the height of the coding tree. The optimal coding tree with a height of k is defined as:

[0049]

[0050] The method of determining the structural entropy of each of the initial coding trees based on the structural entropy minimization algorithm and determining the coding tree corresponding to the minimum structural entropy as the optimal coding tree includes:

[0051] Based on the merge sort algorithm, a full-height coding tree is generated in a bottom-up manner. Each iteration forms a new partition by merging the two child nodes of the root node. The purpose of this process is to increase the height of the coding tree by adding r and its child node v i 、v j Insert a new layer of nodes v ε , v ε Will be used as v i and v j The parent node of v r The child nodes of .

[0052] And delete the internal nodes to compress the height of the coding tree to obtain the intermediate coding tree. This process is used to compress the height of the coding tree by i The child nodes of v are merged into j , and then remove v from the initial encoding tree i To achieve this, where v i Yes j The child node ofi ≠v j and

[0053] The optimal coding tree is obtained by supplementing the nodes in the intermediate coding tree that do not have direct successor nodes in the next layer. This process is performed by inserting v ε , fill v i The layer and v j The gap between the layers, where v i Yes j The child node of i and v j The layer height interval of the coding tree exceeds one layer.

[0054] Through the above steps, the optimal coding tree T for the given graph G can be obtained.

[0055] The encoding tree can be expressed as:

[0056] T=(V T ,E T );in, Represents the node set of each layer of the coding tree T, Corresponding to 128 initial nodes. Thus, a set of clustering assignment matrices can be obtained:

[0057] S=(S 1 ,S 2 ,...,S k );in, represents the clustering assignment matrix of the i-th pooling layer, where n i+1 and n i They are the number of nodes before and after graph coarsening. There are only two values ​​in the matrix, 0 and 1. 0 means that the node corresponding to the horizontal axis does not belong to the class corresponding to the vertical axis after graph coarsening, and 1 means that the node belongs to the class corresponding to the vertical axis after coarsening.

[0058] Step S220: Input the initial graph data and the cluster allocation matrix into a recognition model to obtain a recognition result.

[0059] In an embodiment of the present invention, the recognition model includes a feature extraction module and a classifier module; wherein the feature extraction module is used to concatenate the node features, the adjacency matrix, and the cluster allocation matrix to obtain a feature vector, and the classifier module is used to receive the feature vector and store the changes in the EEG signal time series, convert the feature vector and output it to the linear layer for classification results.

[0060] Specifically, the feature extraction module includes multiple graph neural convolution layers and corresponding multiple graph pooling layers. The first graph neural convolution layer is used to receive the node features and the adjacency matrix, and extract the features of the node features and the adjacency matrix. The first graph pooling layer is used to receive the output of the graph neural convolution layer and the clustering assignment matrix and form a new subgraph, and input the subgraph into the second graph neural convolution layer.

[0061] The input of the feature extraction module includes node features, adjacency matrix and cluster assignment matrix. The processing process of this module includes the following steps:

[0062] In the first step, the node features and adjacency matrix are input into the graph neural convolution layer, and messages are passed between nodes. The adjacency matrix describes the relationship between EEG signal electrodes, and the graph neural convolution layer can effectively calculate and extract features based on the surrounding neighbor relationships. Each node sums the features of the surrounding nodes according to the weight W of its adjacency matrix, adds its own eigenvalue, and then generates new features through a Relu activation function and passes them to the next layer. The graph convolution rules are as follows:

[0063] in represents the adjacency matrix of the graph, I represents the identity matrix, express The degree matrix, W (l) represents a trainable weight matrix, σ() represents a nonlinear activation function, and H (l) ∈R N×D represents the feature matrix of layer l, H (0) is the node feature.

[0064] The output of the graph neural convolution layer and the clustering assignment matrix are input into the graph pooling layer together. The graph data containing 128 nodes is coarsened through the hierarchical pooling structure. The main purpose of this step is to mine a representative subset of nodes to form a new subgraph. After pooling, the new adjacency matrix and node feature matrix are expressed as follows:

[0065] P i+1 =S i H i ; where T represents transpose, is the adjacency matrix of the i-th layer, S i is the clustering assignment matrix of the i-th layer, is the node feature matrix generated by the i-th graph neural convolutional layer. The pooling process can also be expressed as, P i+1 Receive node hidden feature H i , and according to the clustering distribution matrix S i Merge hidden features to get n in the new graphi+1 The initial table of clusters.

[0066] Global pooling is performed through the readout layer to obtain the final representation of the graph.

[0067] The above three steps are repeated three times, and then the results of each time are concatenated to obtain the concatenated features. Finally, the concatenated features are reduced in dimension and finally compressed into a feature vector containing 6 features.

[0068] The feature vector obtained above is input into the classifier module, the brain wave features of the time series are calculated and stored, and finally the result is output to the linear layer to obtain the classification result.

[0069] Specifically, in the embodiment of the present invention, the classifier module is constructed based on the LSTM network. The classifier module is divided into three parts: the first part receives the feature vector output from the feature extraction module; the second part is used to memorize the changes in the EEG signal time series; the third part performs feature matrix conversion and outputs it to the linear layer for classification.

[0070] LSTM controls information through "gates", including forget gate, update gate and output gate. The sigmoid unit of the forget gate processes which information should be discarded; the update gate consists of two parts. First, the operation of the sigmoid input gate is used to determine which information to update, and then the tanh layer creates a new unit current storage state; then the new unit state is updated; the output gate first passes through a sigmoid layer to obtain the judgment condition, and then the tanh layer determines the final unit output. Among them, the number of LSTM hidden layer units in the embodiment of the present invention is 128.

[0071] See also Figure 3 , a psychological type disease diagnosis and identification device 300 is provided, the device is configured in the server, and the device includes:

[0072] A brain map data construction module 310 is used to extract differential entropy features of the plurality of EEG signals, and to perform superposition enhancement on the differential entropy features, to construct initial graph data based on the differential entropy features and graph construction rules, and to construct the initial graph data into an optimal coding tree to obtain a clustering allocation matrix corresponding to a node;

[0073] The recognition module 320 is used to input the initial graph data and the cluster allocation matrix into a recognition model to obtain a recognition result.

[0074] In an embodiment of the present application, the recognition model includes a feature extraction module and a classifier module. The feature extraction module is used to concatenate the node features, the adjacency matrix, and the cluster allocation matrix to obtain a feature vector. The classifier module is used to receive the feature vector and store the changes in the EEG signal time series, convert the feature vector and output it to the linear layer for classification results.

[0075] See also Figure 3 , a minimally invasive surgery preoperative image estimation device 300 is provided, the device is applied to a server, and the device 300 includes:

[0076] A cost volume acquisition module 310 is used to extract features from the left view and the right view to obtain corresponding left feature maps and right feature maps, and concatenate the left feature maps and the right feature maps to form a first cost volume and a second cost volume corresponding to the left view and the right view;

[0077] A disparity estimation module 320, configured to perform a three-dimensional convolution operation on the first cost volume and the second cost volume respectively to obtain a first disparity estimation map and a second disparity estimation map corresponding to the left view and the right view;

[0078] A correction module 330, configured to continuously apply two-dimensional convolution and deconvolution operations to the left feature map and the right feature map to generate a first correction map and a second correction map in a parallax-vertical direction corresponding to the left view and the right view;

[0079] The reconstruction module 340 is used to obtain a left view reconstructed image and a right view reconstructed image according to preset disparity parameters and bilinear interpolation of the left view, the right view, the first disparity estimation map, the second disparity estimation map, the first correction map and the second correction map.

[0080] See also Figure 4, the above method can also be integrated into the provided terminal device 400, and the device may have relatively large differences due to different configurations or performances, and may include one or more processors 401 and memory 402, and the memory 402 may store one or more storage applications or data. Among them, the memory 402 can be a temporary storage or a permanent storage. The application stored in the memory 402 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the terminal device. Furthermore, the processor 401 can be configured to communicate with the memory 402, and the terminal device executes a series of computer executable instructions in the memory 402. The terminal device may also include one or more power supplies 403, one or more wired / wireless network interfaces 404, one or more input / output interfaces 405, one or more keyboards 406, etc.

[0081] In a specific embodiment, the terminal device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer executable instructions in the terminal device, and the one or more programs are configured to be executed by one or more processors, including computer executable instructions for performing the following:

[0082] Extracting differential entropy features of a plurality of the EEG signals, and superimposing and enhancing the differential entropy features, constructing initial graph data based on the differential entropy features and graph construction rules, and constructing the initial graph data into an optimal coding tree to obtain a clustering allocation matrix corresponding to the nodes;

[0083] The initial graph data and the cluster allocation matrix are input into a recognition model to obtain a recognition result.

[0084] The following is a detailed introduction to the various components of the processor:

[0085] In this embodiment, the processor is an application specific integrated circuit (ASIC), or is configured to implement one or more integrated circuits of the embodiments of the present application, such as one or more microprocessors (digital signal processor, DSP), or one or more field programmable gate arrays (field programmable gate array, FPGA).

[0086] Optionally, the processor may execute various functions by running or executing a software program stored in the memory and calling data stored in the memory, such as executing the above-mentioned Figure 2 The method shown.

[0087] In a specific implementation, as an embodiment, the processor may include one or more microprocessors.

[0088] Among them, the memory is used to store the software program that executes the solution of the present application, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0089] Optionally, the memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may be integrated with the processor, or may exist independently and be coupled to the processing unit through the interface circuit of the processor, and the embodiments of the present application do not specifically limit this.

[0090] It should be noted that the structure of the processor shown in this embodiment does not constitute a limitation on the device, and the actual device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0091] In addition, the technical effects of the processor can refer to the technical effects of the method described in the above method embodiment, which will not be repeated here.

[0092] It should be understood that the processor in the embodiments of the present application may be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0093] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0094] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0095] In this application, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0096] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0097] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0098] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0099] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0100] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0101] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0102] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0103] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for diagnosing and identifying psychological diseases, characterized in that: The method is applied to a psychological disease diagnosis system, the psychological diagnosis system includes an electroencephalogram signal acquisition device and a server, the electroencephalogram signal acquisition device includes a plurality of electroencephalogram acquisition electrode channels, the plurality of electroencephalogram acquisition electrode channels are used to acquire a plurality of electroencephalogram signals from a plurality of regions of a patient's head to be diagnosed, and a data terminal electrically connected to the electroencephalogram acquisition electrode, the data terminal communicates with the server, and the method is configured in the server; the method includes: Extracting differential entropy features of a plurality of the EEG signals, and superimposing and enhancing the differential entropy features, constructing initial graph data based on the differential entropy features and graph construction rules, and constructing the initial graph data into an optimal coding tree to obtain a clustering allocation matrix corresponding to the nodes; the initial graph data includes node features and an adjacency matrix; The initial graph data and the cluster allocation matrix are input into the recognition model to obtain the recognition result; the recognition model includes a feature extraction module and a classifier module, the feature extraction module is used to splice the node features, the adjacency matrix, and the cluster allocation matrix to obtain a feature vector, the classifier module is used to receive the feature vector and store the changes in the EEG signal time series, and convert the feature vector and output it to the linear layer for classification result.

2. The method for diagnosing and identifying psychological diseases according to claim 1, characterized in that: The initial graph data is constructed based on the differential entropy feature and the graph construction rule, including: taking the multiple EEG acquisition electrode channels as nodes, taking the differential entropy feature as the node feature, calculating the Pearson coefficient between the multiple EEG acquisition channels, and based on the preset Pearson coefficient threshold, when two channels with a value greater than the threshold are connected, and channels with a value less than the threshold are not connected, the adjacency matrix is ​​determined based on the edge structure, and the node features and the adjacency matrix are combined into the initial graph data.

3. The method for diagnosing and identifying psychological diseases according to claim 2, characterized in that: The coding tree includes a root node, an internal node and a leaf node; the initial graph data is constructed into an optimal coding tree, including: forming a plurality of initial coding trees based on the root node, the internal node and the leaf node, and determining the structural entropy of each of the initial coding trees based on a structural entropy minimization algorithm, and determining the coding tree corresponding to the minimum structural entropy as the optimal coding tree; the structural entropy is determined based on the following formula: Where V is the node set in the initial graph data G, and is also the initial leaf node set of the initial coding tree T. i is a non-root node in T, Yes i The parent node of v i The number of edges with endpoints in the leaf node partition of , vol(V) and vol(v i ) represent V and v respectively i The degree of the leaf node.

4. The method for diagnosing and identifying psychological diseases according to claim 3, characterized in that: The method of determining the structural entropy of each of the initial coding trees based on the structural entropy minimization algorithm and determining the coding tree corresponding to the minimum structural entropy as the optimal coding tree includes: specifying the height of the initial coding tree, the initial coding tree has only one root node and multiple leaf nodes, the number of the leaf nodes is the same as the number of the EEG acquisition electrode channels, and the leaf nodes of the initial coding tree are all child nodes of the root node; based on the merge sort algorithm, generating a full-height coding tree in a bottom-up manner, and each iteration forms a new partition by merging the two child nodes of the root node; and deleting internal nodes to compress the height of the coding tree to obtain an intermediate coding tree; and supplementing the nodes in the intermediate coding tree that do not directly inherit nodes in the next layer to obtain the optimal coding tree.

5. The method for diagnosing and identifying psychological diseases according to claim 4, characterized in that: The step of inputting the initial graph data and the cluster allocation matrix into the recognition model to obtain the recognition result includes: inputting the initial graph data and the cluster allocation matrix into the feature extraction module for feature splicing, and reducing the dimension of the spliced ​​features to obtain a feature vector; inputting the feature vector into the classifier module, calculating and storing the brain wave features of the time series, and finally outputting the results to the linear layer to obtain the classification result.

6. The method for diagnosing and identifying psychological diseases according to claim 5, characterized in that: The feature extraction module includes multiple graph neural convolution layers and corresponding multiple graph pooling layers. The first graph neural convolution layer is used to receive the node features and the adjacency matrix, and extract the features of the node features and the adjacency matrix. The first graph pooling layer is used to receive the output of the graph neural convolution layer and the clustering assignment matrix and form a new subgraph, and input the subgraph into the second graph neural convolution layer.

7. The method for diagnosing and identifying psychological diseases according to claim 6, characterized in that: The rules of the graph neural convolution layer are as follows: in represents the adjacency matrix of the graph, I represents the identity matrix, express The degree matrix, W (l) represents a trainable weight matrix, σ() represents a nonlinear activation function, and H (l) ∈R N×D represents the feature matrix of layer l, H (0) is the node feature.

8. The method for diagnosing and identifying psychological diseases according to claim 7, characterized in that: The classifier module is built based on the LSTM network.

9. A device for diagnosing and identifying psychological diseases, characterized in that: The device is applied to a psychological diagnosis system, which includes an electroencephalogram (EEG) signal acquisition device and a server. The electroencephalogram (EEG) signal acquisition device includes a plurality of electroencephalogram (EEG) acquisition electrode channels, which are used to acquire a plurality of electroencephalogram (EEG) signals from a plurality of regions of a patient's head to be diagnosed, and a data terminal electrically connected to the electroencephalogram (EEG) acquisition electrode, which communicates with the server, and the device is configured in the server. The device includes: A brain map data construction module is used to extract differential entropy features of the plurality of EEG signals, and to superimpose and enhance the differential entropy features, to construct initial graph data based on the differential entropy features and graph construction rules, and to construct the initial graph data into an optimal coding tree to obtain a clustering allocation matrix corresponding to the nodes; The recognition module is used to input the initial graph data and the cluster allocation matrix into the recognition model to obtain the recognition result; the recognition model includes a feature extraction module and a classifier module, the feature extraction module is used to concatenate the node features, the adjacency matrix, and the cluster allocation matrix to obtain a feature vector, the classifier module is used to receive the feature vector and store the changes in the EEG signal time series, and convert the feature vector and output it to the linear layer for classification results.

10. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the diagnosis and identification of psychological type diseases as described in any one of claims 1 to 8 when executing the computer program.

Images