Training method and device for motor imagery ability assessment model

Abstract

The present disclosure provides a training method and device for a motor imagery ability assessment model. The method relates to the technical field of motor imagery brain-computer interfaces, and includes the following steps: obtaining a training sample set, the training sample set including multiple EEG data sequences; determining, for each EEG data sequence, respective brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands; screening target features from the respective brain network reconstruction features of the EEG data sequence in the multiple characteristic frequency bands based on the correlation between the respective brain network reconstruction features of the EEG data sequence in the multiple characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence; and training an initial model using the target features of the EEG data sequence as training samples and the categories of the motor imagery ability attributes of the EEG data sequence as labels to obtain a motor imagery ability assessment model.

Description

Technical Field

[0001] The present disclosure relates to the technical field of motor imagery brain-computer interface, and more specifically, to a training method and device for a motor imagery ability assessment model. Background Art

[0002] Electroencephalogram (EEG) signals are widely used in brain-computer interface (BCI) systems due to their non-invasive nature, high temporal resolution, safety, and convenience. Related research has identified various individual information features related to motor imagery (MI) abilities, including frequency band energy and nonlinear dynamics, in EEG signals surrounding specific brain regions.

[0003] In the process of realizing the concept of the present disclosure, relevant researchers found that there are at least the following problems in the relevant technologies: the brain is a complex system, and the realization of brain functions is not only related to specific brain regions, but also depends on the joint cooperation between multiple brain regions. Focusing only on the neural response pattern of a single brain region will result in the loss of important brain connection information. Therefore, there is an urgent need for a training method for a motor imagery ability assessment model to help researchers in this field to deeply reveal individual differences in MI ability through the motor imagery ability assessment model obtained through training. Summary of the Invention

[0004] In view of this, the present disclosure provides a training method and device for a motor imagery ability assessment model.

[0005] One aspect of the present disclosure provides a training method for a motor imagery ability assessment model, comprising: obtaining a training sample set, the training sample set comprising a plurality of EEG data sequences; for each EEG data sequence, determining the brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands; based on the correlation between the brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence, screening out target features from the brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands; using the target features of the EEG data sequences as training samples and the categories of the motor imagery ability attributes of the EEG data sequences as labels, training an initial model, and obtaining a motor imagery ability assessment model.

[0006] According to an embodiment of the present disclosure, the above-mentioned determination of the brain network reconstruction features of the above-mentioned EEG data sequence in multiple characteristic frequency bands based on the above-mentioned EEG data sequence includes: preprocessing the above-mentioned EEG data sequence to obtain resting period EEG data of multiple task rounds and task period EEG data of multiple task rounds in each characteristic frequency band of the above-mentioned multiple characteristic frequency bands; using a weighted paired phase consistency method to process the resting period EEG data of the above-mentioned multiple task rounds and the task period EEG data of the above-mentioned multiple task rounds respectively to obtain a resting period brain network matrix and a task period brain network matrix; based on the above-mentioned resting period brain network matrix and the above-mentioned task period brain network matrix, obtaining a resting period brain network feature and a task period brain network feature; performing feature reconstruction based on the above-mentioned resting period brain network feature and the above-mentioned task period brain network feature to obtain the brain network reconstruction feature of the above-mentioned EEG data sequence in each characteristic frequency band.

[0007] According to an embodiment of the present disclosure, the target EEG data is the above-mentioned resting period EEG data or the above-mentioned task period EEG data, and the above-mentioned target EEG data includes sub-data corresponding to each of the multiple leads; wherein the above-mentioned weighted paired phase consistency method is used to respectively process the resting period EEG data of the above-mentioned multiple task rounds and the task period EEG data of the above-mentioned multiple task rounds to obtain a resting period brain network matrix and a task period brain network matrix, including: based on the cross-spectrum between the two sub-data corresponding to each lead pair in the target EEG data of the multiple task rounds, determining the functional connection value of each lead pair, the above-mentioned lead pair includes two leads; based on the functional connection values ​​of each of the multiple lead pairs, obtaining the brain network matrix of the above-mentioned target EEG data, the brain network matrix of the above-mentioned target EEG data is the above-mentioned resting period brain network matrix or the above-mentioned task period brain network matrix.

[0008] According to an embodiment of the present disclosure, the brain network characteristics of the target EEG data are the resting period brain network characteristics or the task period brain network characteristics, and the brain network characteristics of the target EEG data include brain network connection strength characteristics, global efficiency of the brain network, characteristic path length of the brain network, local efficiency of the brain network and clustering coefficient of the brain network; wherein, the resting period brain network characteristics and task period brain network characteristics obtained based on the resting period brain network matrix and the task period brain network matrix include: the average value of the functional connection values ​​of all lead pairs in the brain network matrix based on the target EEG data, Determine the connection strength characteristics of the above-mentioned brain network; calculate the shortest weighted path length of each lead pair based on the functional connectivity values ​​of each of the above-mentioned multiple lead pairs and the optional paths between the two leads in each lead pair; determine the global efficiency of the above-mentioned brain network and the characteristic path length of the above-mentioned brain network based on the shortest weighted path lengths of each of the above-mentioned multiple lead pairs; determine the local efficiency of the above-mentioned brain network based on the shortest weighted path lengths of each of the above-mentioned multiple lead pairs and the functional connectivity values ​​of each of the above-mentioned multiple lead pairs; determine the clustering coefficient of the above-mentioned brain network based on the functional connectivity values ​​of each of the above-mentioned multiple lead pairs.

[0009] According to an embodiment of the present disclosure, the training method of the motor imagination ability evaluation model also includes: for each of the above-mentioned EEG data sequences, using the multiple EEG data subsequences included in the above-mentioned EEG data sequence and the task category labels of the above-mentioned multiple EEG data subsequences, the motor imagination ability of the above-mentioned EEG data sequence is evaluated to obtain the motor imagination ability attributes of the above-mentioned EEG data sequence.

[0010] According to an embodiment of the present disclosure, for each of the above-mentioned EEG data sequences, the motor imagery ability of the above-mentioned EEG data sequence is evaluated by utilizing multiple EEG data subsequences and respective task category labels of the multiple EEG data subsequences to obtain the motor imagery ability attributes of the above-mentioned EEG data sequence, including: performing feature extraction on the above-mentioned multiple EEG data subsequences respectively to obtain the task feature vectors of the above-mentioned multiple EEG data subsequences; cross-validating the classification model based on the respective task feature vectors of the above-mentioned multiple EEG data subsequences and respective task category labels of the above-mentioned multiple EEG data subsequences to obtain the classification accuracy of the above-mentioned EEG data sequence; and obtaining the motor imagery ability attributes of the above-mentioned EEG data sequence based on the classification accuracy of the above-mentioned EEG data sequence.

[0011] According to an embodiment of the present disclosure, the above-mentioned multiple EEG data subsequences are respectively subjected to feature extraction to obtain task feature vectors of each of the multiple EEG data subsequences, including: for each of the above-mentioned EEG data subsequences, the above-mentioned EEG data subsequence is spatially filtered using a common spatial pattern algorithm to obtain an intermediate EEG data subsequence, and the above-mentioned intermediate EEG data subsequence is composed of multiple vectors; and the vector located at a preset position in the above-mentioned intermediate EEG data subsequence is selected to obtain the task feature vector of the above-mentioned EEG data subsequence.

[0012] According to an embodiment of the present disclosure, the cross-validation of the classification model based on the task feature vectors of each of the above-mentioned multiple EEG data subsequences and the task category labels of each of the above-mentioned multiple EEG data subsequences to obtain the classification accuracy of the above-mentioned EEG data sequence includes: for each target EEG data subsequence, determining a training set and a validation set from the task feature vectors of each of the above-mentioned multiple EEG data subsequences and the task category labels of the above-mentioned multiple EEG data subsequences, wherein the above-mentioned validation set includes the task feature vector of the above-mentioned target EEG data subsequence and the task category label of the above-mentioned target EEG data subsequence, and the above-mentioned target EEG data subsequence belongs to the above-mentioned multiple EEG data subsequences; using the above-mentioned training set to train the initial model to obtain the above-mentioned classification model; and determining the classification accuracy of the above-mentioned EEG data sequence based on the verification result of the above-mentioned classification model on the above-mentioned validation set.

[0013] According to an embodiment of the present disclosure, the target feature is screened from the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands based on the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagination ability attribute of the EEG data sequence, including: using the Pearson correlation coefficient method to calculate the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagination ability attribute of the EEG data sequence, to obtain multiple correlation coefficients; based on the size of the absolute value of each of the multiple correlation coefficients, determining the target correlation coefficient from the multiple correlation coefficients; determining the brain network reconstruction feature of the characteristic frequency band corresponding to the target correlation coefficient as the target feature.

[0014] Another aspect of the present disclosure provides a training device for a motor imagery ability evaluation model, comprising: a sample acquisition module for acquiring a training sample set, wherein the training sample set includes multiple EEG data sequences; a feature determination module for determining, for each EEG data sequence, the brain network reconstruction features of the EEG data sequence in multiple feature frequency bands; a target determination module for screening target features from the brain network reconstruction features of the EEG data sequence in multiple feature frequency bands based on the correlation between the brain network reconstruction features of the EEG data sequence in multiple feature frequency bands and the motor imagery ability attributes of the EEG data sequence; and a model determination module for using the target features of the EEG data sequence as training samples and the categories of the motor imagery ability attributes of the EEG data sequence as labels to train an initial model and obtain a motor imagery ability evaluation model.

[0015] According to the embodiments of the present disclosure, the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagination ability attributes of the EEG data sequence is used to screen out target features from the brain network reconstruction features of each characteristic frequency band. Since the target features of different characteristic frequency bands are associated with different cognitive functions, and the target features of each characteristic frequency band have a high correlation with the motor imagination ability attributes, the initial model is trained based on the target features of the EEG data sequence as training samples and the categories of the motor imagination ability attributes of the EEG data sequence as labels, which can deeply evaluate individual differences in MI ability and improve the accuracy of the motor imagination ability evaluation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0017] Figure 1 Schematically illustrates an exemplary system architecture of a training method to which a motor imagery ability assessment model can be applied according to an embodiment of the present disclosure;

[0018] Figure 2 A flowchart schematically illustrates a method for training a motor imagery ability assessment model according to an embodiment of the present disclosure;

[0019] Figure 3 Schematically shows a relationship diagram between an EEG data sequence and a target feature according to an embodiment of the present disclosure;

[0020] Figure 4 A flowchart schematically illustrates a method for training a motor imagery ability assessment model according to another embodiment of the present disclosure;

[0021] Figure 5A structural block diagram schematically illustrates a training device for a motor imagery ability assessment model according to an embodiment of the present disclosure; and

[0022] Figure 6 A block diagram of an electronic device suitable for implementing a training method for a motor imagery ability assessment model according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0023] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the detailed description below, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0024] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0025] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0026] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0027] In the embodiments of this disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of all data involved (including, but not limited to, user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to safeguard the security of user personal information, network security, and national security.

[0028] In the embodiments of the present disclosure, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0029] In the embodiments disclosed herein, brain-computer interface technology, by constructing artificial information transmission pathways outside the body, enables direct control interaction between the brain and external devices without relying on peripheral nerves and muscle tissue. The Motor Imagery-based Brain-Computer Interface (MI-BCI) is a typical active BCI system. It detects endogenous neural activity signals spontaneously generated when a user performs imagined movements and converts them into corresponding external device control commands to achieve specific functions or complete specific tasks. Imagined movements can include imagined movements of the hands, feet, or tongue.

[0030] Compared to other BCI systems, MI-BCIs require no external stimulation and allow users to autonomously modulate brain activity to achieve specific control objectives. This allows them to better reflect the brain's voluntary activity and more closely resemble the original concept of BCI as a "mind-controlled device." With the increasing development and application of MI-BCI systems, higher demands have been placed on their reliability and stability. However, due to the subjective nature of task models and the implicit nature of neural responses, users' ability to control MI-BCIs varies significantly. In particular, a certain percentage of users, even after sufficient training, struggle to generate accurate neural activity patterns and are unable to master the BCI system. These users are often referred to as "MI-BCI-blind." Accurately identifying individuals with "MI-BCI-blindness" not only helps optimize the personalized use of MI-BCI systems, saving resources and time, and improving their efficiency, but also helps uncover the underlying causes of individual variability in motor imagery performance and promote related research.

[0031] EEG signals are widely used in BCI systems due to their advantages such as non-invasiveness, high temporal resolution, safety and convenience. Related research focuses on EEG signals in specific brain areas, and can identify a variety of individual information characteristics related to MI capabilities, including frequency band energy and nonlinear dynamic characteristics. However, the brain is a complex system, and the realization of brain functions is not only related to specific brain areas, but also depends on the joint collaboration between multiple brain areas. Focusing only on the neural response pattern of a single brain area will result in the loss of important brain connection information. Relevant studies have shown that individual MI capabilities are not only related to task-state neural activity, but also that resting-state neural activity is correlated with MI capabilities.

[0032] However, current research has often focused on the relationship between neural activity in a single brain state and MI ability, neglecting changes in neural activity patterns between brain state transitions and ignoring important neural activity characteristics that reflect the brain's information processing during specific tasks. Therefore, exploring the brain's network reconfiguration patterns during the transition from resting state to MI task state and enriching the individual information characteristics related to MI ability are of great research value for further revealing the neural mechanisms underlying individual differences in MI ability.

[0033] An embodiment of the present disclosure provides a training method for a motor imagery ability assessment model, comprising: obtaining a training sample set, the training sample set comprising a plurality of EEG data sequences; for each EEG data sequence, determining the respective brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands based on the EEG data sequence and the motor imagery ability attributes of the EEG data sequence; screening target features from the respective brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands based on the correlation between the respective brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence; and using the target features of the EEG data sequences as training samples and the motor imagery ability attributes of the EEG data sequences as category labels to train an initial model and obtain a motor imagery ability assessment model.

[0034] Figure 1 The exemplary system architecture 100 of the training method of the motor imagery ability assessment model according to the embodiment of the present disclosure is schematically shown. It should be noted that, Figure 1 The examples shown are merely examples of system architectures to which the embodiments of the present disclosure may be applied, to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure may not be used in other devices, systems, environments or scenarios.

[0035] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0036] A user may use a first terminal device 101, a second terminal device 102, or a third terminal device 103 to interact with a server 105 via a network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, or the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software (for example only).

[0037] The first terminal device 101 , the second terminal device 102 , and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0038] The server 105 may be a server that provides various services, such as a background management server (for example only) that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process received data such as user requests, and feed back processing results (e.g., web pages, information, or data obtained or generated based on user requests) to the terminal devices.

[0039] It should be noted that the training method of the motor imagery ability assessment model provided in the embodiment of the present disclosure can generally be executed by the server 105. The training method of the motor imagery ability assessment model provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Alternatively, the training method of the motor imagery ability assessment model provided in the embodiment of the present disclosure can also be executed by the first terminal device 101, the second terminal device 102 or the third terminal device 103, or can also be executed by other terminal devices different from the first terminal device 101, the second terminal device 102 or the third terminal device 103.

[0040] For example, the training sample set may be originally stored in any one of the first terminal device 101, the second terminal device 102, or the third terminal device 103 (for example, the first terminal device 101, but not limited thereto), or stored on an external storage device and imported into the first terminal device 101. Then, the first terminal device 101 may locally execute the training method for the motor imagery ability assessment model provided by the embodiment of the present disclosure, or send the training sample set to other terminal devices, servers, or server clusters, and the other terminal devices, servers, or server clusters that receive the training sample set may execute the training method for the motor imagery ability assessment model provided by the embodiment of the present disclosure.

[0041] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0042] Figure 2 The flowchart of the training method of the motor imagery ability assessment model according to the embodiment of the present disclosure is schematically shown.

[0043] like Figure 2 As shown, the method includes operations S210 to S240.

[0044] In operation S210 , a training sample set is obtained, where the training sample set includes a plurality of EEG data sequences.

[0045] In operation S220 , for each EEG data sequence, respective brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands are determined.

[0046] In operation S230, based on the correlation between the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands and the motor imagery ability attribute of the EEG data sequence, a target feature is screened from the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands.

[0047] In operation S240 , the target features of the EEG data sequence are used as training samples, and the categories of the motor imagery attributes of the EEG data sequence are used as labels to train an initial model to obtain a motor imagery ability evaluation model.

[0048] According to an embodiment of the present disclosure, a training sample set can be obtained from a public database, for example, the GigaDB public database. The training sample set may include 64-channel EEG signals of multiple subjects performing left-hand and right-hand motor imagery tasks. The EEG signals are recorded by an EEG acquisition system and acquisition software, and the data sampling rate may be 512Hz. Each subject completed multiple trials of the motor imagery task, wherein one trial represents a task round, and the motor imagery task may include a left-hand motor imagery task and a right-hand motor imagery task. Each trial may include multiple sets of resting period EEG data and task period EEG data. The EEG data sequence represents a data sequence formed by changes in EEG signals in a time series. These data sequences capture the electrical activity patterns and frequencies of neurons in the cerebral cortex. An EEG data sequence is a data sequence formed by changes in EEG signals of a subject in multiple trials or over a period of time.

[0049] According to an embodiment of the present disclosure, the multiple characteristic frequency bands may include theta band, alpha band, beta band, and gamma band. The brain network reconstruction feature is used to characterize the characteristics after fusing the characteristics of multiple EEG data sequences. For each EEG data sequence, the brain network reconstruction features of the EEG data sequence in theta band, alpha band, beta band, and gamma band are determined.

[0050] According to an embodiment of the present disclosure, the motor imagery attribute of the EEG data sequence is used to characterize the motor imagery ability reflected by the EEG data sequence. The motor imagery attribute of the EEG data sequence can be the level of the motor imagery ability reflected by the EEG data sequence, etc. For example, the motor imagery attribute of the EEG data sequence can be divided into five levels, which are respectively used to characterize the size of the motor imagery ability reflected by the EEG data sequence. The target feature characterizes the EEG data sequence that meets the screening conditions, wherein the screening conditions can be that the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagery attribute of the EEG data sequence is greater than a preset threshold, wherein the correlation characterizes the degree of influence of multiple brain network reconstruction features on the motor imagery attribute, and the higher the correlation, the greater the degree of influence of the brain network reconstruction feature on the motor imagery attribute; the preset threshold can be determined based on the representation form and value range of the correlation, and the preset threshold can be in numerical form.

[0051] According to an embodiment of the present disclosure, the correlation between multiple brain network reconstruction features and motor imagination ability attributes in each characteristic frequency band is calculated respectively, and based on the correlation between multiple brain network reconstruction features and motor imagination ability attributes, the target features of each characteristic frequency band are screened out from the EEG data sequence.

[0052] Figure 3 The figure schematically shows the relationship between the EEG data sequence and the target features according to an embodiment of the present disclosure.

[0053] like Figure 3 As shown, based on n EEG data sequences, the brain network reconstruction features of the EEG data sequences in theta band, alpha band, beta band and gamma band are determined respectively, where p, q, m and z are all positive integers, and the numbers after the brain network reconstruction features are used to distinguish different brain network reconstruction features.

[0054] Based on the correlation between the brain network reconstruction features of the EEG data sequences in multiple characteristic frequency bands and the motor imagery ability attributes of the EEG data sequences, target feature 1 is screened from the brain network reconstruction features of the theta frequency band; target feature 2 is screened from the brain network reconstruction features of the alpha frequency band; target feature 3 is screened from the brain network reconstruction features of the beta frequency band; and target feature 4 is screened from the brain network reconstruction features of the gamma frequency band.

[0055] According to an embodiment of the present disclosure, the initial model can be a model for evaluating motor imagery ability constructed using a logistic regression analysis method. The target features of the EEG data sequence are used as training samples, wherein the training samples can be understood as independent variables corresponding to the initial model. The motor imagery ability attribute can be expressed in numerical form. A classification threshold can be set, for example 70%. When the motor imagery ability attribute is higher than this threshold, the motor imagery ability attribute is classified as the first level; when the motor imagery ability attribute is lower than this threshold, the motor imagery ability attribute is classified as the second level. The first level and the second level are motor imagery ability category labels. The category of the motor imagery ability attribute of the EEG data sequence is used as a label to train the initial model and obtain a motor imagery ability evaluation model.

[0056] According to the embodiments of the present disclosure, the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagination ability attributes of the EEG data sequence is used to screen out target features from the brain network reconstruction features of each characteristic frequency band. Since the target features of different characteristic frequency bands are associated with different cognitive functions, and the target features of each characteristic frequency band have a high correlation with the motor imagination ability attributes, the initial model is trained based on the target features of the EEG data sequence as training samples and the categories of the motor imagination ability attributes of the EEG data sequence as labels, which can deeply evaluate individual differences in MI ability and improve the accuracy of the motor imagination ability evaluation model.

[0057] According to an embodiment of the present disclosure, for each EEG data sequence, the motor imagery ability of the EEG data sequence is evaluated using the multiple EEG data subsequences included in the EEG data sequence and the task category labels of the multiple EEG data subsequences to obtain the motor imagery ability attributes of the EEG data sequence.

[0058] According to an embodiment of the present disclosure, each EEG data sequence includes EEG data subsequences of multiple task trials, that is, the EEG data sequence includes multiple EEG data subsequences. The task category label is used to identify the task category of the EEG data subsequence, and the task category label may include left-hand motor imagery and right-hand motor imagery. Using the multiple EEG data subsequences included in the EEG data sequence and the task category labels of each of the multiple EEG data subsequences, a model for motor imagery task category classification is trained, and other EEG data subsequences are input into the model for motor imagery task category classification, thereby performing motor imagery ability evaluation on the EEG data sequence based on the output results of the model and the task category labels, and obtaining the motor imagery ability attributes of each EEG data sequence, wherein the output results of the model may include left-hand motor imagery and right-hand motor imagery.

[0059] According to an embodiment of the present disclosure, motor imagery ability assessment is performed on an EEG data sequence. The comparison results of each EEG data subsequence can be obtained by comparing whether the model output results of multiple EEG data subsequences are the same as their respective task category labels, wherein the comparison results may include the same or different. The proportion of EEG data subsequences with the same task category label as the task category label in the multiple EEG data subsequences included in the EEG data sequence is used as the motor imagery ability attribute of the EEG data sequence.

[0060] According to the embodiments of the present disclosure, motor imagery ability is assessed for multiple EEG data subsequences and their respective task category labels, thereby identifying differences in motor imagery ability between different EEG data sequences. By characterizing the motor imagery ability of EEG data sequences using motor imagery attributes, the cognitive functions corresponding to these EEG data sequences are analyzed, improving the accuracy of MI ability assessment.

[0061] According to an embodiment of the present disclosure, for each EEG data sequence, a plurality of EEG data subsequences and respective task category labels of the plurality of EEG data subsequences are used to perform a motor imagery ability evaluation on the EEG data sequence to obtain the motor imagery ability attribute of the EEG data sequence, including: performing feature extraction on the plurality of EEG data subsequences respectively to obtain the respective task feature vectors of the plurality of EEG data subsequences; cross-validating a classification model based on the respective task feature vectors of the plurality of EEG data subsequences and the respective task category labels of the plurality of EEG data subsequences to obtain the classification accuracy of the EEG data sequence; and obtaining the motor imagery ability attribute of the EEG data sequence based on the classification accuracy of the EEG data sequence.

[0062] According to an embodiment of the present disclosure, an EEG data subsequence can represent the sequence of an EEG data sequence in a task round; an EEG data subsequence can also represent the sequence of an EEG data sequence during a motor imagery task in a task round. The task feature vector represents the feature vector of the EEG data subsequence. For each EEG data subsequence, a frequency domain filter can be applied for bandpass filtering, and the EEG data of the filtered EEG data subsequence during the motor imagery task period can be intercepted to obtain the EEG data during the motor imagery task period. The frequency of the bandpass filtering applied by the frequency domain filter can be 8~30Hz.

[0063] According to an embodiment of the present disclosure, feature extraction is performed on multiple EEG data subsequences respectively to obtain task feature vectors of each of the multiple EEG data subsequences, including: for each EEG data subsequence, spatial filtering is performed on the EEG data subsequence using a common spatial pattern algorithm to obtain an intermediate EEG data subsequence, where the intermediate EEG data subsequence is composed of multiple vectors; and a vector located at a preset position in the intermediate EEG data subsequence is selected to obtain the task feature vector of the EEG data subsequence.

[0064] According to an embodiment of the present disclosure, the common spatial pattern algorithm is a spatial domain filtering feature extraction algorithm applied to multi-channel data. The intermediate EEG data subsequence can be determined by formula (1).

[0065] (1);

[0066] in, is the intermediate EEG data subsequence after spatial filtering, is the spatial filter matrix, multiple vectors of the intermediate EEG data subsequences , is a single spatial filter vector, where n is the number of spatial filter vectors, E is the number of EEG signal leads, represents the transpose of the spatial filter matrix, is the EEG data subsequence, where The number of data points captured for each lead.

[0067] According to an embodiment of the present disclosure, a vector at a preset position in an intermediate EEG data subsequence is selected, and a task feature vector for the EEG data subsequence is constructed based on the features corresponding to the vector at the preset position. The preset position can be the first 4-dimensional vector and the last 4-dimensional vector of the intermediate EEG data subsequence, which can produce a 2×4=8-dimensional task feature vector. The first 4-dimensional and last 4-dimensional vectors may capture the EEG activity most relevant to the motor imagery task, enhancing the discrimination between signal and noise.

[0068] According to an embodiment of the present disclosure, the feature corresponding to the vector of the preset position can be determined by formula (2).

[0069] (2);

[0070] in, For the vector The corresponding features, var is the variance operator, and n is the number of spatial filter vectors.

[0071] According to an embodiment of the present disclosure, a vector at a preset position in the intermediate EEG data subsequence after spatial filtering is selected to obtain a task feature vector for the EEG data subsequence. By selecting a vector at a preset position, the task feature vector can be reduced in dimensionality, reducing computational complexity while retaining key information and improving the expressive power of the task feature vector.

[0072] According to an embodiment of the present disclosure, cross-validation of the classification model may include K-fold cross-validation, leave-one-out cross-validation, and stratified cross-validation.

[0073] According to an embodiment of the present disclosure, a classification model is cross-validated based on the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences to obtain the classification accuracy of the EEG data sequence, including: for each target EEG data subsequence, a training set and a validation set are determined from the task feature vectors of each of the multiple EEG data subsequences and the task category labels of the multiple EEG data subsequences, wherein the validation set includes the task feature vector of the target EEG data subsequence and the task category label of the target EEG data subsequence, and the target EEG data subsequence belongs to the multiple EEG data subsequences; an initial model is trained using the training set to obtain a classification model; and the classification accuracy of the EEG data sequence is determined based on the verification result of the validation set by the classification model.

[0074] According to an embodiment of the present disclosure, the target EEG data subsequence represents the EEG data subsequences that are classified into the validation set, i.e., multiple EEG data subsequences used to determine motor imagery attributes. A task feature vector belonging to the validation set can be determined by computer random selection from the task feature vectors of each of the multiple EEG data subsequences.

[0075] According to an embodiment of the present disclosure, for leave-one-out cross-validation, the task feature vector and task category label of one EEG data subsequence can be set aside each time as a validation set, and the task feature vectors and task category labels of the remaining EEG data subsequences can be used as a training set.

[0076] According to an embodiment of the present disclosure, for the K-fold cross-validation method, a computer can be used to determine K pieces of data from the task feature vectors of each of a plurality of EEG data subsequences, wherein one piece is selected as a validation set and (K-1) pieces are selected as training sets, and different data are selected in turn as validation sets, thereby obtaining K pairs of training sets and validation sets, wherein each validation set includes the task feature vectors and task category labels of each of a plurality of target EEG data subsequences.

[0077] According to an embodiment of the present disclosure, the initial model can be a support vector machine model. The initial model maps the task feature vectors of the training set to a high-dimensional feature space by introducing a kernel function. The kernel function can include a linear kernel function, a polynomial kernel function, a Gaussian kernel function, a sigmoid kernel function, and the like. The initial model is trained using the task feature vectors and task category labels of each of the multiple EEG data subsequences in the training set to obtain a classification model.

[0078] According to the embodiments of the present disclosure, classification recognition is performed using a classification model, using a linear kernel function. Preferably, a 10×10-fold cross-validation is used to determine the classification accuracy, which is used as the classification accuracy of the target EEG data subsequence. The classification accuracy of each of the multiple EEG data subsequences is used as the motor imagery ability attribute of the EEG data sequence.

[0079] For example, the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences are divided into 10 parts, each of which includes the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences. One of the 10 parts is used as a validation set, and the remaining 9 parts are used as training sets. The task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences are selected in turn as validation sets, thereby obtaining 10 pairs of training sets and validation sets. For each pair of training sets and validation sets, the training set is used to train the initial model to obtain a classification model, and the validation set is used to verify the classification accuracy of the EEG data sequence by the classification model.

[0080] According to an embodiment of the present disclosure, a classification model is cross-validated using task feature vectors and task category labels for each of a plurality of EEG data subsequences, and the resulting classification accuracy of each EEG data sequence is used as the motor imagery attribute of the EEG data sequence. Since the classification accuracy directly reflects the accuracy of the classification model in identifying the task category, and it can be assumed that the more obvious the features of each EEG data subsequence in the EEG data sequence, the higher the classification accuracy of the EEG data sequence, the motor imagery ability of the EEG data sequence can be indirectly reflected through the classification accuracy of the EEG data sequence, thereby improving the accuracy of the motor imagery ability represented by the motor imagery attribute.

[0081] According to an embodiment of the present disclosure, the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands are determined based on the EEG data sequence, including: preprocessing the EEG data sequence to obtain resting period EEG data of multiple task rounds and task period EEG data of multiple task rounds in each characteristic frequency band of the multiple characteristic frequency bands; using a weighted paired phase consistency method to process the resting period EEG data of multiple task rounds and the task period EEG data of multiple task rounds respectively to obtain a resting period brain network matrix and a task period brain network matrix; based on the resting period brain network matrix and the task period brain network matrix, obtain resting period brain network features and task period brain network features; perform feature reconstruction based on the resting period brain network features and the task period brain network features to obtain the brain network reconstruction features of the EEG data sequence in each characteristic frequency band.

[0082] According to an embodiment of the present disclosure, a task round represents a round of performing an imagery task. Resting EEG data represents the EEG data of the brain when not performing a motor imagery task. Task-period EEG data represents the EEG data when performing a specific motor imagery task. Preprocessing the EEG data sequence includes: applying a frequency domain filter to the EEG data sequence to perform bandpass filtering at 4-7 Hz (theta band), 8-13 Hz (alpha band), 14-30 Hz (beta band), and 31-45 Hz (gamma band), applying a spatial filter to remove low-frequency coupling between EEG data from different channels, and intercepting the resting EEG data in the MI task round and the task-period EEG data of multiple task rounds.

[0083] According to an embodiment of the present disclosure, in the process of determining brain network reconstruction features and determining motor imagery ability attributes, the preprocessing of the EEG data sequence is performed separately, and different preprocessing operations are used respectively.

[0084] According to an embodiment of the present disclosure, the resting-period brain network matrix characterization is a matrix composed of multiple measurement values ​​obtained by measuring the degree of non-uniform distribution of phase angle differences of resting-period EEG data in multiple task rounds using a weighted pairwise phase consistency (wPPC) method. The task-period brain network matrix characterization is a matrix composed of multiple measurement values ​​obtained by measuring the degree of non-uniform distribution of phase angle differences of task-period EEG data in multiple task rounds using a weighted pairwise phase consistency method. The measurement values ​​can characterize the degree of phase synchronization between any two task-period EEG data or resting-period EEG data.

[0085] According to embodiments of the present disclosure, resting-period brain network features are used to characterize the characteristics of resting-period EEG. Task-period brain network features are used to characterize the characteristics of task-period EEG. A complex network analysis method based on graph theory is used to measure the resting-period brain network matrix to obtain resting-period brain network features. A complex network analysis method based on graph theory is used to measure the task-period brain network matrix to obtain task-period brain network features.

[0086] According to an embodiment of the present disclosure, the brain network reconstruction feature can be determined by the following formula (3).

[0087] (3);

[0088] in, Represents the brain network characteristics of the task at each time point, represents the resting brain network characteristics at each time point during the resting period, mean represents the mean operator, and std represents the standard deviation operator. The brain network reconstruction characteristics of each frequency band can be obtained by averaging the brain network reconstruction characteristics of each frequency point within the theta, alpha, beta, and gamma bands.

[0089] According to the embodiments of the present disclosure, by fusing and analyzing the resting period EEG data of multiple task rounds and the task period EEG data of multiple task rounds, the resting period brain network characteristics and the task period brain network characteristics that can characterize the brain network connection characteristics are determined, and feature reconstruction is performed based on the resting period brain network characteristics and the task period brain network characteristics to obtain the brain network reconstruction characteristics of the EEG data sequence in each characteristic frequency band. The brain network reconstruction characteristics of each characteristic frequency band further reflect the coordination pattern of the brain in different states and the information interaction characteristics between different brain regions, effectively characterize the information processing process related to the realization of the brain's motor imagination function, and improve the accuracy of the characteristics of the input motor imagination ability evaluation model.

[0090] According to an embodiment of the present disclosure, the target EEG data is resting EEG data or task-period EEG data, and the target EEG data includes sub-data corresponding to each of the multiple leads. The resting EEG data of multiple task rounds and the task-period EEG data of multiple task rounds are processed separately using a weighted paired phase consistency method to obtain a resting brain network matrix and a task-period brain network matrix, including: based on the cross-spectrum between the two sub-data corresponding to each lead pair in the target EEG data of multiple task rounds, determining the functional connectivity value of each lead pair, where the lead pair includes two leads; based on the functional connectivity values ​​of each of the multiple lead pairs, obtaining a brain network matrix of the target EEG data, where the brain network matrix of the target EEG data is a resting brain network matrix or a task-period brain network matrix.

[0091] According to an embodiment of the present disclosure, the sub-data corresponding to each of the multiple leads represents the EEG data on the lead, that is, the EEG signal on the lead, where the lead represents the electrode measuring the EEG signal. The cross spectrum between two sub-data represents a measure of the phase difference between the two sub-data at a specific frequency and time point. The cross spectrum between the two sub-data can be obtained using formula (4).

[0092] (4);

[0093] in, It's a lead In frequency and time The spectrum at It's a lead In frequency and time The complex conjugate of the spectrum at .

[0094] According to an embodiment of the present disclosure, the functional connectivity value of a lead pair is used to characterize the degree of phase synchronization of the sub-data corresponding to each lead in the lead pair. Based on the target EEG data from multiple task rounds, the weighted pairwise phase consistency (wPPC) method is used to measure the degree of phase synchronization of the sub-data corresponding to each lead in each lead pair in the resting period EEG data or task period EEG data of each task round, thereby obtaining the functional connectivity value of each lead pair. The functional connectivity value of each lead pair can be obtained using formula (5).

[0095] (5);

[0096] in, For leads and leads The EEG signal at the frequency and time The cross spectrum at , n is the number of task rounds of the target EEG data.

[0097] According to the embodiment of the present disclosure, the functional connectivity value of each lead pair is repeatedly calculated to form a frequency and time The brain network matrix corresponding to the target EEG data. The brain network matrix is ​​preferably 64×64 in size, meaning the number of leads is preferably 64. The brain network matrix can be stored as an upper triangular matrix or a lower triangular matrix to save storage space. A larger functional connectivity value indicates a higher degree of phase synchronization between the corresponding sub-data of the lead pair.

[0098] According to an embodiment of the present disclosure, a weighted paired phase consistency method is used to measure the phase synchronization degree of each lead pair's EEG signal for target EEG data across multiple task rounds, and the functional connectivity values ​​of all lead pairs are integrated to construct a brain network matrix. Because the functional connectivity value of each lead pair measures the uneven distribution of phase angle differences between lead pairs across different task rounds, the impact of changes in task rounds on the functional connectivity value of each lead pair is reduced. The functional connectivity value of each lead pair will not deviate due to the number of task rounds, thereby improving the robustness of the brain network matrix of the target EEG data.

[0099] According to an embodiment of the present disclosure, the brain network characteristics of the target EEG data are resting period brain network characteristics or task period brain network characteristics, and the brain network characteristics of the target EEG data include brain network connection strength characteristics, global efficiency of the brain network, characteristic path length of the brain network, local efficiency of the brain network and clustering coefficient of the brain network.

[0100] According to an embodiment of the present disclosure, based on the resting period brain network matrix and the task period brain network matrix, the resting period brain network characteristics and the task period brain network characteristics are obtained, including: determining the brain network connection strength characteristics based on the average value of the functional connection values ​​of all lead pairs in the brain network matrix of the target EEG data; calculating the shortest weighted path length of each lead pair based on the functional connection values ​​of each of the multiple lead pairs and the optional paths between the two leads in each lead pair; determining the global efficiency of the brain network and the characteristic path length of the brain network based on the shortest weighted path lengths of each of the multiple lead pairs; determining the local efficiency of the brain network based on the shortest weighted path lengths of each of the multiple lead pairs and the functional connection values ​​of each of the multiple lead pairs; and determining the clustering coefficient of the brain network based on the functional connection values ​​of each of the multiple lead pairs.

[0101] According to an embodiment of the present disclosure, the average value of the functional connectivity values ​​of all lead pairs is used as the brain network connection strength. The larger the average value of the functional connectivity values ​​of all lead pairs, the stronger the brain network connection strength.

[0102] According to embodiments of the present disclosure, the global efficiency of a brain network is a global measure of brain network efficiency, reflecting the efficiency of information transmission between all leads in the brain network structure. It can be represented by the average value of the information transmission efficiency between all leads in the brain network structure. The global efficiency of a brain network can be obtained using Formula (6).

[0103] (6);

[0104] in, represents the set of all leads in the brain network, represents the number of nodes in the brain network, represents the shortest weighted path length between lead i and lead j in the brain network, where the weight between lead i and lead j in the brain network is the functional connectivity value between lead i and lead j, It can be computed by a weighted graph search algorithm, such as Dijkstra's algorithm or Floyd-Warshall's algorithm.

[0105] According to an embodiment of the present disclosure, the characteristic path length of a brain network is a global measure of the path length of the brain network, which can be represented by the average length of the shortest path between all lead pairs in the brain network structure. The characteristic path length of the brain network can be obtained by formula (7).

[0106] (7);

[0107] Among them, N represents the set of all leads in the brain network, n represents the number of leads in the brain network, represents the shortest weighted path length between lead i and lead j in the brain network. The weight between lead i and lead j in the brain network is the functional connectivity value between lead i and lead j. It can be computed by a weighted graph search algorithm, such as Dijkstra's algorithm or Floyd-Warshall's algorithm.

[0108] According to an embodiment of the present disclosure, the local efficiency of a brain network is a local measure of the efficiency of the brain network, which can be represented by the average value of the information transfer efficiency between a certain lead and its adjacent leads. The local efficiency of the brain network can be obtained by formula (8).

[0109] (8);

[0110] Among them, N represents the set of all leads in the brain network, n represents the number of leads in the brain network, represents the functional connectivity value between leads, represents the shortest weighted path length between lead j and lead h within the brain network, Represents the set of leads adjacent to lead i.

[0111] According to an embodiment of the present disclosure, the clustering coefficient of a brain network is a measure of the degree of focus of brain network nodes, reflecting the closeness of the interconnection between a lead and its adjacent leads. The clustering coefficient of a brain network can be obtained by formula (9).

[0112] (9);

[0113] Among them, N represents the set of all leads in the brain network, n represents the number of leads in the brain network, Represents the functional connectivity value between leads.

[0114] According to the embodiments of the present disclosure, the functional integration characteristics of the brain network topology are characterized by the global efficiency and characteristic path length of the brain network in the resting period and the task period. The functional separation characteristics of the brain network topology are characterized by the local efficiency and clustering coefficient of the brain network. Describing the structure and function of the brain network using multiple metrics provides data support for revealing the brain's information processing process that supports motor imagery.

[0115] Figure 4 The flowchart of the training method of the motor imagery ability assessment model according to another embodiment of the present disclosure is schematically shown.

[0116] like Figure 4 As shown, in operation S410, the EEG data sequence is preprocessed.

[0117] Specifically, different preprocessing is performed on the multiple EEG data sequences 411. For the EEG data sequences 411 entering operation S420, a frequency domain filter is applied to the multiple EEG data sequences 411 to perform bandpass filtering, and the EEG data during the task period is intercepted to determine the motor imagery ability attribute. For the EEG data sequence 411 entering operation S430, a frequency domain filter is applied to perform bandpass filtering of 4-7 Hz (theta band), 8-13 Hz (alpha band), 14-30 Hz (beta band), and 31-45 Hz (gamma band). A spatial filter is applied to remove low-frequency coupling between EEG data sequences 411 of different channels. The resting period EEG data and the task period EEG data of multiple task rounds are intercepted to determine the brain network reconstruction characteristics.

[0118] In operation S420 , motor imagery ability assessment is performed.

[0119] Specifically, feature extraction is performed on the multiple task-period EEG data obtained by filtering and intercepting the multiple EEG data sequences 411 to obtain task feature vectors 421 for each of the multiple EEG data sequences 411. A classification model 422 is cross-validated based on the task feature vectors 421 and the task category labels for each of the multiple EEG data sequences 411 to obtain classification accuracy rates for each of the multiple EEG data sequences 411. The classification accuracy rates for each of the multiple EEG data sequences 411 are used as motor imagery attributes 423 for each of the multiple EEG data sequences 411.

[0120] In operation S430 , a brain network reconstruction feature is determined.

[0121] Specifically, the weighted paired phase consistency method is used to process the resting period EEG data of multiple task rounds and the task period EEG data of multiple task rounds respectively to obtain the resting period brain network matrix and the task period brain network matrix.

[0122] Based on the resting-period brain network matrix and the task-period brain network matrix, resting-period brain network features and task-period brain network features are obtained. Both the resting-period brain network features and the task-period brain network features include brain network connection strength features 431, global network efficiency 432, characteristic path length 433, local network efficiency 434, and clustering coefficient 435. Feature reconstruction is performed based on the resting-period brain network features and the task-period brain network features to obtain brain network reconstruction features for each characteristic frequency band of the EEG data sequence.

[0123] In operation S440 , target features are screened.

[0124] Specifically, based on the correlation between the brain network reconstruction features of the EEG data sequence 411 in multiple characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence, the target feature 441 is screened from the brain network reconstruction features of the EEG data sequence 411 in multiple characteristic frequency bands.

[0125] In operation S450 , an initial model is trained.

[0126] Specifically, the target feature 441 of the EEG data sequence 411 is used as a training sample, and the motor imagery ability attribute 423 of the EEG data sequence is used as a label to train the initial model 451 and obtain the motor imagery ability evaluation model 452.

[0127] According to an embodiment of the present disclosure, based on the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagination ability attribute of the EEG data sequence, a target feature is screened from the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands, including: using the Pearson correlation coefficient method to calculate the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagination ability attribute of the EEG data sequence, to obtain multiple correlation coefficients; based on the size of the absolute value of each of the multiple correlation coefficients, determining the target correlation coefficient from the multiple correlation coefficients; and determining the brain network reconstruction feature of the characteristic frequency band corresponding to the target correlation coefficient as the target feature.

[0128] According to an embodiment of the present disclosure, for each of the multiple characteristic frequency bands, the Pearson correlation coefficient method is used to measure the strength and direction of the linear relationship between the brain network reconstruction characteristics of the EEG data sequence and the motor imagery ability attribute. The correlation coefficient ranges from -1 to 1, where a value of 0 indicates no correlation between the brain network reconstruction characteristics and the motor imagery ability attribute, and a value close to 1 or -1 indicates a positive or negative correlation between the two variables.

[0129] According to embodiments of the present disclosure, the target correlation coefficient can represent the correlation coefficient with the largest absolute value in the characteristic frequency band. For each characteristic frequency band, the brain network reconstruction feature with the largest absolute value of the correlation coefficient in the characteristic frequency band is selected as the target feature. When the number of characteristic frequency bands is four, a four-dimensional target feature can be obtained.

[0130] According to an embodiment of the present disclosure, the target correlation coefficient may represent a plurality of correlation coefficients in a characteristic frequency band whose absolute value of the correlation coefficient is greater than a preset threshold. For each characteristic frequency band, a brain network reconstruction feature corresponding to the target correlation coefficient in each characteristic frequency band is selected as the target feature.

[0131] According to an embodiment of the present disclosure, based on the correlation coefficient between the brain network reconstruction features of the EEG data sequence and the motor imagination ability attributes, the brain network reconstruction feature target feature corresponding to the higher correlation coefficient is determined as the target feature from multiple brain network reconstruction features of multiple characteristic frequency bands, thereby improving the accuracy of the brain network reconstruction features used to train the motor imagination ability evaluation model, thereby improving the accuracy of the motor imagination ability evaluation model.

[0132] Figure 5 The structural block diagram of the training device of the motor imagery ability assessment model according to an embodiment of the present disclosure is schematically shown.

[0133] like Figure 5 As shown, the training device 500 for the motor imagery ability assessment model includes: a sample acquisition module 510, a feature determination module 520, a target determination module 530, and a model determination module 540.

[0134] The sample acquisition module 510 is used to acquire a training sample set, where the training sample set includes multiple EEG data sequences.

[0135] The feature determination module 520 is configured to determine, for each EEG data sequence, the brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands.

[0136] The target determination module 530 is used to filter out target features from the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands based on the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence.

[0137] The model determination module 540 is used to use the target features of the EEG data sequence as training samples and the categories of the motor imagery ability attributes of the EEG data sequence as labels to train an initial model and obtain a motor imagery ability evaluation model.

[0138] According to an embodiment of the present disclosure, the feature determination module 520 includes: a data determination submodule, a matrix determination submodule, a feature determination submodule, and a feature reconstruction submodule.

[0139] The data determination submodule is used to preprocess the EEG data sequence to obtain resting EEG data of multiple task rounds and task period EEG data of multiple task rounds in each characteristic frequency band of multiple characteristic frequency bands.

[0140] The matrix determination submodule is used to use the weighted paired phase consistency method to process the resting period EEG data of multiple task rounds and the task period EEG data of multiple task rounds respectively, to obtain the resting period brain network matrix and the task period brain network matrix.

[0141] The feature determination submodule is used to obtain the resting period brain network features and the task period brain network features based on the resting period brain network matrix and the task period brain network matrix.

[0142] The feature reconstruction submodule is used to perform feature reconstruction based on the brain network features of the resting period and the brain network features of the task period, and obtain the brain network reconstruction features of the EEG data sequence in each characteristic frequency band.

[0143] According to an embodiment of the present disclosure, the target EEG data is resting period EEG data or task period EEG data, and the target EEG data includes sub-data corresponding to each of the multiple leads.

[0144] The matrix determination submodule includes: a connection value determination unit and a matrix determination unit.

[0145] The connection value determination unit is used to determine the functional connection value of each lead pair based on the cross spectrum between two sub-data corresponding to each lead pair in the target EEG data of multiple task rounds, where the lead pair includes two leads.

[0146] The matrix determination unit is used to obtain the brain network matrix of the target EEG data based on the functional connectivity values ​​of the multiple lead pairs. The brain network matrix of the target EEG data is a resting period brain network matrix or a task period brain network matrix.

[0147] According to an embodiment of the present disclosure, the brain network characteristics of the target EEG data are resting period brain network characteristics or task period brain network characteristics, and the brain network characteristics of the target EEG data include brain network connection strength characteristics, global efficiency of the brain network, characteristic path length of the brain network, local efficiency of the brain network and clustering coefficient of the brain network.

[0148] The feature determination submodule includes: a first determination unit, a second determination unit, a third determination unit, an efficiency determination unit, and a coefficient determination unit.

[0149] The first determining unit is used to determine the brain network connection strength feature based on the average value of the functional connection values ​​of all lead pairs in the brain network matrix of the target EEG data.

[0150] The second determining unit is configured to calculate the shortest weighted path length of each lead pair based on the functional connectivity values ​​of the plurality of lead pairs and the optional paths between the two leads in each lead pair.

[0151] The third determining unit is configured to determine the global efficiency of the brain network and the characteristic path length of the brain network based on the shortest weighted path lengths of the plurality of lead pairs.

[0152] The efficiency determination unit is configured to determine the local efficiency of the brain network based on the shortest weighted path lengths of the plurality of lead pairs and the functional connectivity values ​​of the plurality of lead pairs.

[0153] The coefficient determination unit is used to determine the clustering coefficient of the brain network based on the functional connectivity values ​​of the plurality of lead pairs.

[0154] According to an embodiment of the present disclosure, the training device 500 for the motor imagery ability assessment model further includes: an attribute determination module.

[0155] The attribute determination module is used to evaluate the motor imagery ability of each EEG data sequence using the multiple EEG data subsequences included in the EEG data sequence and the task category labels of the multiple EEG data subsequences, so as to obtain the motor imagery ability attribute of the EEG data sequence.

[0156] According to an embodiment of the present disclosure, the attribute determination module includes: a vector determination submodule, a correctness determination submodule, and an attribute determination submodule.

[0157] The vector determination submodule is used to extract features from multiple EEG data subsequences respectively to obtain task feature vectors of each of the multiple EEG data subsequences.

[0158] The accuracy determination submodule is used to perform cross-validation of the classification model based on the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences to obtain the classification accuracy of the EEG data sequence.

[0159] The attribute determination submodule is used to obtain the motor imagery ability attribute of the EEG data sequence based on the classification accuracy of the EEG data sequence.

[0160] According to an embodiment of the present disclosure, the vector determination submodule includes: a filtering unit and a vector determination unit.

[0161] The filtering unit is used to perform spatial filtering on each EEG data subsequence using a common spatial pattern algorithm to obtain an intermediate EEG data subsequence, where the intermediate EEG data subsequence consists of multiple vectors.

[0162] The vector determination unit is used to select a vector at a preset position in the intermediate EEG data subsequence to obtain a task feature vector of the EEG data subsequence.

[0163] According to an embodiment of the present disclosure, the accuracy determination submodule includes: a set determination unit, a model determination unit, and an accuracy determination unit.

[0164] A set determination unit is used to determine, for each target EEG data subsequence, a training set and a validation set from the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences, wherein the validation set includes the task feature vector of the target EEG data subsequence and the task category label of the target EEG data subsequence, and the target EEG data subsequence belongs to the multiple EEG data subsequences.

[0165] The model determination unit is used to train the initial model using the training set to obtain a classification model.

[0166] The accuracy determination unit is used to determine the classification accuracy of the EEG data sequence based on the verification result of the classification model on the verification set.

[0167] According to an embodiment of the present disclosure, the target determination module includes: a first determination submodule, a second determination submodule, and a target determination submodule.

[0168] The first determination submodule is used to calculate the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence using the Pearson correlation coefficient method to obtain multiple correlation coefficients.

[0169] The second determining submodule is configured to determine a target correlation coefficient from the multiple correlation coefficients based on the magnitude of the absolute values ​​of the multiple correlation coefficients.

[0170] The target determination submodule is used to determine the brain network reconstruction feature of the characteristic frequency band corresponding to the target correlation coefficient as the target feature.

[0171] According to the embodiments of the present invention, any number of modules, sub-modules, units, and sub-units, or at least part of the functions of any number of them, can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, sub-modules, units, and sub-units can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, sub-modules, units, and sub-units can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging the circuit, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, sub-modules, units, and sub-units can be at least partially implemented as a computer program module, which can perform the corresponding functions when the computer program module is executed.

[0172] For example, any number of the sample acquisition module 510, feature determination module 520, target determination module 530, and model determination module 540 may be combined into a single module / unit / sub-unit, or any one of these modules / units / sub-units may be split into multiple modules / units / sub-units. Alternatively, at least part of the functionality of one or more of these modules / units / sub-units may be combined with at least part of the functionality of other modules / units / sub-units and implemented in a single module / unit / sub-unit. According to an embodiment of the present disclosure, at least one of the sample acquisition module 510, feature determination module 520, target determination module 530, and model determination module 540 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or may be implemented in hardware or firmware by any other reasonable means of integrating or packaging circuits, or may be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of these. Alternatively, at least one of the sample acquisition module 510, the feature determination module 520, the target determination module 530, and the model determination module 540 can be at least partially implemented as a computer program module, which can perform corresponding functions when executed.

[0173] It should be noted that the training device part of the motor imagination ability assessment model in the embodiment of the present disclosure corresponds to the training method part of the motor imagination ability assessment model in the embodiment of the present disclosure. The description of the training device of the motor imagination ability assessment model specifically refers to the training method part of the motor imagination ability assessment model, which will not be repeated here.

[0174] Figure 6 A block diagram of an electronic device suitable for implementing a training method for a motor imagery ability assessment model according to an embodiment of the present disclosure is schematically shown. Figure 6 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0175] like Figure 6As shown, the electronic device 600 according to an embodiment of the present disclosure includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage portion 608 into a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to the embodiment of the present disclosure.

[0176] Various programs and data required for the operation of the electronic device 600 are stored in the RAM 603. The processor 601, ROM 602, and RAM 603 are connected to each other via a bus 604. The processor 601 executes the various operations of the method flow according to the embodiment of the present disclosure by executing the programs in the ROM 602 and / or RAM 603. It should be noted that the programs may also be stored in one or more memories other than the ROM 602 and RAM 603. The processor 601 may also execute the various operations of the method flow according to the embodiment of the present disclosure by executing the programs stored in the one or more memories.

[0177] According to an embodiment of the present disclosure, electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to bus 604. Electronic device 600 may also include one or more of the following components connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including devices such as a cathode ray tube (CRT), liquid crystal display (LCD), and speakers; a storage section 608 including a hard disk; and a communication section 609 including a network interface card such as a LAN card or modem. Communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. Removable media 611, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed in drive 610 as needed, so that computer programs read from the removable media can be installed into storage section 608 as needed.

[0178] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, the above-mentioned functions defined in the system of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0179] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, and when executed, implements the method according to the embodiments of the present disclosure.

[0180] According to embodiments of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium. Examples include, but are not limited to, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0181] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 602 and / or the RAM 603 described above and / or one or more memories other than the ROM 602 and the RAM 603 .

[0182] An embodiment of the present disclosure also includes a computer program product, which includes a computer program containing program code for executing the method provided by the embodiment of the present disclosure. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the training method of the motor imagination ability assessment model provided by the embodiment of the present disclosure.

[0183] When the computer program is executed by the processor 601, the above functions defined in the system / device of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0184] In one embodiment, the computer program may be stored on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may be transmitted and distributed in the form of a signal on a network medium, downloaded and installed via the communication portion 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0185] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect via the Internet).

[0186] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of the systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two boxes shown in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, as well as the combination of boxes in the block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions. It will be understood by those skilled in the art that the features described in the various embodiments of the present disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or coupled in various ways, and all of these combinations and / or couplings fall within the scope of the present disclosure.

[0187] The above describes the embodiments of the present disclosure. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although each embodiment has been described separately above, this does not mean that the measures in each embodiment cannot be advantageously used in combination. Without departing from the scope of the present disclosure, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A training method for a motor imagery ability assessment model, comprising: Acquire a training sample set, wherein the training sample set includes a plurality of EEG data sequences; For each EEG data sequence, determining the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands; Based on the correlation between the brain network reconstruction features of the EEG data sequence in each of the plurality of characteristic frequency bands and the motor imagery ability attribute of the EEG data sequence, a target feature is screened from the brain network reconstruction features of the EEG data sequence in each of the plurality of characteristic frequency bands; as well as Using the target features of the EEG data sequence as training samples and the categories of the motor imagery attributes of the EEG data sequence as labels, an initial model is trained to obtain a motor imagery ability evaluation model; The step of determining the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands includes: Preprocessing the EEG data sequence to obtain resting-period EEG data of multiple task rounds and task-period EEG data of multiple task rounds in each of the multiple characteristic frequency bands; determining a functional connectivity value for each lead pair based on a cross-spectrum between two sub-data corresponding to each lead pair in target EEG data from multiple task rounds, wherein the lead pair includes two leads, the target EEG data is the resting period EEG data or the task period EEG data, and the target EEG data includes sub-data corresponding to each of the multiple leads; Obtaining a brain network matrix of the target EEG data based on the functional connectivity values ​​of the plurality of lead pairs, wherein the brain network matrix of the target EEG data is the resting period brain network matrix or the task period brain network matrix; Based on the resting period brain network matrix and the task period brain network matrix, obtaining a resting period brain network feature and a task period brain network feature; and Feature reconstruction is performed based on the resting period brain network features and the task period brain network features to obtain the brain network reconstruction features of the EEG data sequence in each characteristic frequency band.

2. The method according to claim 1, wherein The brain network characteristics of the target EEG data are the resting period brain network characteristics or the task period brain network characteristics, and the brain network characteristics of the target EEG data include brain network connection strength characteristics, global efficiency of the brain network, characteristic path length of the brain network, local efficiency of the brain network, and clustering coefficient of the brain network; The step of obtaining the resting period brain network features and the task period brain network features based on the resting period brain network matrix and the task period brain network matrix includes: determining the brain network connection strength feature based on an average of functional connectivity values ​​of all lead pairs in a brain network matrix of the target EEG data; Calculating the shortest weighted path length of each lead pair based on the functional connectivity values ​​of the plurality of lead pairs and the optional paths between the two leads in each lead pair; determining a global efficiency of the brain network and a characteristic path length of the brain network based on the shortest weighted path lengths of the plurality of lead pairs; determining a local efficiency of the brain network based on the shortest weighted path lengths of each of the plurality of lead pairs and the functional connectivity values ​​of each of the plurality of lead pairs; and A clustering coefficient of the brain network is determined based on the functional connectivity values ​​of each of the plurality of lead pairs.

3. The method according to claim 1, further comprising: For each of the EEG data sequences, the motor imagery ability of the EEG data sequence is evaluated using the multiple EEG data subsequences included in the EEG data sequence and the task category labels of the multiple EEG data subsequences to obtain the motor imagery ability attribute of the EEG data sequence.

4. The method according to claim 3, wherein: For each of the EEG data sequences, the motor imagery ability evaluation is performed on the EEG data sequence using a plurality of EEG data subsequences and respective task category labels of the plurality of EEG data subsequences to obtain a motor imagery ability attribute of the EEG data sequence, including: Performing feature extraction on each of the multiple EEG data subsequences to obtain task feature vectors for each of the multiple EEG data subsequences; Performing cross-validation on a classification model based on the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences to obtain a classification accuracy of the EEG data sequence; and Based on the classification accuracy of the EEG data sequence, a motor imagery ability attribute of the EEG data sequence is obtained.

5. The method according to claim 4, wherein The performing feature extraction on the multiple EEG data subsequences to obtain task feature vectors of the multiple EEG data subsequences respectively includes: For each of the EEG data subsequences, spatially filtering the EEG data subsequence using a common spatial pattern algorithm to obtain an intermediate EEG data subsequence, where the intermediate EEG data subsequence consists of a plurality of vectors; and A vector at a preset position in the intermediate EEG data subsequence is selected to obtain a task feature vector of the EEG data subsequence.

6. The method according to claim 4, wherein: The cross-validation of the classification model based on the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences to obtain the classification accuracy of the EEG data sequence includes: For each target EEG data subsequence, determining a training set and a validation set from the task feature vectors of each of the multiple EEG data subsequences and the task category labels of each of the multiple EEG data subsequences, wherein the validation set includes the task feature vector of the target EEG data subsequence and the task category label of the target EEG data subsequence, and the target EEG data subsequence belongs to the multiple EEG data subsequences; Using the training set to train an initial model to obtain the classification model; and Based on the verification result of the classification model on the verification set, the classification accuracy of the EEG data sequence is determined.

7. The method according to claim 1, wherein The method of screening target features from the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands based on the correlation between the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands and the motor imagery ability attribute of the EEG data sequence includes: Calculating the correlation between the brain network reconstruction features of the EEG data sequence in multiple characteristic frequency bands and the motor imagery ability attributes of the EEG data sequence using a Pearson correlation coefficient method to obtain multiple correlation coefficients; determining a target correlation coefficient from the plurality of correlation coefficients based on the magnitude of the absolute values ​​of the respective correlation coefficients; and The brain network reconstruction feature of the characteristic frequency band corresponding to the target correlation coefficient is determined as the target feature.

8. A training device for a motor imagery ability assessment model, comprising: A sample acquisition module is used to acquire a training sample set, wherein the training sample set includes a plurality of EEG data sequences; a feature determination module, configured to determine, for each EEG data sequence, respective brain network reconstruction features of the EEG data sequence in a plurality of characteristic frequency bands; a target determination module, configured to filter and obtain a target feature from the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands based on a correlation between the brain network reconstruction features of the EEG data sequence in the plurality of characteristic frequency bands and the motor imagery ability attribute of the EEG data sequence; as well as a model determination module, configured to use the target features of the EEG data sequence as training samples and the categories of the motor imagery attributes of the EEG data sequence as labels to train an initial model and obtain a motor imagery ability assessment model; Wherein, the feature determination module is used to: Preprocessing the EEG data sequence to obtain resting-period EEG data of multiple task rounds and task-period EEG data of multiple task rounds in each of the multiple characteristic frequency bands; determining a functional connectivity value for each lead pair based on a cross-spectrum between two sub-data corresponding to each lead pair in target EEG data from multiple task rounds, wherein the lead pair includes two leads, the target EEG data is the resting period EEG data or the task period EEG data, and the target EEG data includes sub-data corresponding to each of the multiple leads; Obtaining a brain network matrix of the target EEG data based on the functional connectivity values ​​of the plurality of lead pairs, wherein the brain network matrix of the target EEG data is the resting period brain network matrix or the task period brain network matrix; Based on the resting period brain network matrix and the task period brain network matrix, obtaining a resting period brain network feature and a task period brain network feature; and Feature reconstruction is performed based on the resting period brain network features and the task period brain network features to obtain the brain network reconstruction features of the EEG data sequence in each characteristic frequency band.

Images