Artificial intelligence-based cognitive impairment detection device, computer device and medium

Abstract

The application belongs to the field of artificial intelligence and the field of digital medicine, and relates to a cognitive impairment detection device based on artificial intelligence, a computer device and a storage medium.The device obtains original brain electrical signals of a user, pre-processes the original brain electrical signals to obtain a brain electrical signal sequence, adds a classification mark and position coding to the brain electrical signal sequence to obtain a target brain electrical signal sequence, processes the target brain electrical signal sequence based on an anomaly detection layer in an anomaly detection model to obtain target features, performs prediction processing on the target features based on a full connection layer in the anomaly detection model to obtain a prediction result, generates a numerical comparison result of the prediction result and a preset threshold value, and generates a shallow cognitive impairment detection result corresponding to the user based on the numerical comparison result.In addition, the shallow cognitive impairment detection result can be stored in a blockchain.The application adopts a mode of shallow cognitive impairment detection based on an anomaly detection model, and improves the accuracy of shallow cognitive impairment detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence development and the field of digital medicine, and in particular relates to a cognitive impairment detection device based on artificial intelligence, a computer device and a storage medium. BACKGROUND

[0002] In the medical field, electroencephalogram (EEG) as a non-invasive neurophysiological monitoring method has been widely used in the detection of mild cognitive impairment (MCI). Traditional EEG analysis methods mainly rely on time domain, frequency domain and time-frequency domain analysis. These methods identify specific EEG features of MCI patients to provide important basis for diagnosis. However, with the in-depth study of technology and the growth of application demand, the existing technology gradually reveals its limitations in the detection of mild cognitive impairment.

[0003] The traditional processing method of detecting mild cognitive impairment (MCI) based on feature extraction of electroencephalogram analysis usually requires professionals to have deep knowledge of neuroelectrophysiology and signal processing skills, and to extract feature parameters such as amplitude and frequency through complex calculation process. This highly dependent artificial way not only is inefficient, but also is prone to subjectivity and instability of the detection results, which is difficult to meet the demand of obtaining detection results quickly and accurately in clinical practice. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a cognitive impairment detection device based on artificial intelligence, a computer device and a storage medium, to solve the technical problem that the existing processing method of detecting mild cognitive impairment (MCI) based on feature extraction of electroencephalogram analysis is highly dependent on artificial, which is not only inefficient, but also is prone to subjectivity and instability of the detection results.

[0005] In order to solve the above technical problems, the embodiments of the present application provide a cognitive impairment detection device based on artificial intelligence, which adopts the following technical solutions:

[0006] The first acquisition module is configured to acquire the original electroencephalogram signal of a user.

[0007] The preprocessing module is configured to preprocess the original electroencephalogram signal to obtain a corresponding electroencephalogram signal sequence.

[0008] The first processing module is configured to add classification labels and position encodings to the electroencephalogram signal sequence to obtain a corresponding target electroencephalogram signal sequence.

[0009] The second processing module is configured to process the target electroencephalogram signal sequence based on an anomaly detection layer in a trained anomaly detection model to obtain target features corresponding to the target electroencephalogram signal sequence.

[0010] The prediction module is configured to perform prediction processing on the target feature based on a full connection layer in the anomaly detection model to obtain a corresponding prediction result.

[0011] The first generation module is configured to obtain a preset threshold value and generate a numerical comparison result between the prediction result and the preset threshold value.

[0012] The second generation module is configured to generate a mild cognitive impairment detection result corresponding to the user based on the numerical comparison result.

[0013] Further, the preprocessing module comprises:

[0014] The filtering submodule is configured to perform filtering processing on the original electroencephalogram signal to obtain a corresponding first electroencephalogram signal.

[0015] The segmentation submodule is configured to perform channel segmentation processing on the first electroencephalogram signal to obtain a corresponding second electroencephalogram signal.

[0016] The flattening submodule is configured to perform data flattening operation on the second electroencephalogram signal to obtain a corresponding third electroencephalogram signal.

[0017] The projection submodule is configured to perform linear projection processing on the third electroencephalogram signal to obtain a corresponding fourth electroencephalogram signal.

[0018] The determination submodule is configured to take the fourth electroencephalogram signal as the target electroencephalogram signal.

[0019] Further, the second generation module comprises:

[0020] The first judgment submodule is configured to judge whether the numerical comparison result is that the prediction result is greater than the preset threshold value.

[0021] The first generation submodule is configured to generate a first mild cognitive impairment detection result that the user has mild cognitive impairment if the prediction result is greater than the preset threshold value.

[0022] The second generation submodule is configured to generate a second mild cognitive impairment detection result that the user does not have mild cognitive impairment if the prediction result is less than the preset threshold value.

[0023] Further, the cognitive impairment detection device based on artificial intelligence further comprises:

[0024] The second acquisition module is configured to acquire a pre-constructed electroencephalogram signal sample.

[0025] The division module is configured to divide the electroencephalogram signal sample into a training set and a test set.

[0026] The third obtaining module is configured to obtain a preset initial anomaly detection model and perform model initialization on the initial anomaly detection model.

[0027] The third processing module is configured to input the training set into the initial anomaly detection model for forward propagation and calculate an error between a model output of the initial anomaly detection model and a true label of the training set.

[0028] The fourth processing module is configured to calculate a gradient based on the error, update a model weight of the initial anomaly detection model through a back propagation algorithm, and optimize the error by using a minimization strategy and a maximization strategy.

[0029] The fifth processing module is configured to repeat the forward propagation and the back propagation process until a preset iteration condition is met, and obtain a corresponding specified anomaly detection model.

[0030] The test module is configured to test the specified anomaly detection model based on the test set.

[0031] The determination module is configured to determine that the specified anomaly detection model is the anomaly detection model if the specified anomaly detection model passes the test.

[0032] Further, the second obtaining module comprises:

[0033] The first obtaining submodule is configured to obtain a pre-acquired initial electroencephalogram signal sample.

[0034] The calling submodule is configured to call a preset sample construction model.

[0035] The construction submodule is configured to perform sample construction processing on the initial electroencephalogram signal sample based on the sample construction model, and obtain a corresponding expanded sample.

[0036] The third generation submodule is configured to generate the electroencephalogram signal sample based on the initial electroencephalogram signal sample and the expanded sample.

[0037] Further, the test module comprises:

[0038] The second obtaining submodule is configured to input the test set into the specified anomaly detection model and obtain a performance index value of the specified anomaly detection model. The performance index value is a value corresponding to a preset performance index, and the number of performance indexes includes multiple.

[0039] The second judgment submodule is configured to determine whether all the performance index values are within a corresponding preset value range.

[0040] a first determining sub-module, configured to determine that the specified abnormality detection model passes the test if all of the evaluation indexes are within the corresponding preset numerical range;

[0041] a second determining sub-module, configured to determine that the specified abnormality detection model fails the test if all of the evaluation indexes are not within the corresponding preset numerical range.

[0042] Further, the artificial intelligence-based cognitive impairment detection apparatus further comprises:

[0043] a third generating module, configured to generate a corresponding target detection report based on the mild cognitive impairment detection result;

[0044] a fourth acquiring module, configured to acquire communication information of the user;

[0045] a pushing module, configured to push the target detection report to the user based on the communication information.

[0046] Further, the third generating module comprises:

[0047] a converting sub-module, configured to convert the mild cognitive impairment detection result into a target detection result corresponding to a preset understanding style;

[0048] a third acquiring sub-module, configured to acquire medical advice information corresponding to the mild cognitive impairment detection result;

[0049] a fourth acquiring sub-module, configured to acquire lifestyle advice information corresponding to the mild cognitive impairment detection result;

[0050] a fifth acquiring sub-module, configured to acquire a preset detection report template;

[0051] a filling sub-module, configured to fill the target detection result, the medical advice information and the lifestyle advice information into the detection report template to obtain the target detection report.

[0052] To solve the above technical problems, the embodiment of the present application further provides a computer device comprising a memory and a processor, the memory stores computer readable instructions, and the processor executes the computer readable instructions to realize the functions of each module in the above artificial intelligence-based cognitive impairment detection apparatus.

[0053] To solve the above technical problems, the embodiment of the present application further provides a computer readable storage medium, which stores computer readable instructions, and the computer readable instructions are executed by a processor to realize the functions of each module in the above artificial intelligence-based cognitive impairment detection apparatus.

[0054] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0055] In the present application, the cognitive impairment detection device based on artificial intelligence first acquires the original electroencephalogram signal of a user; and pre-processes the original electroencephalogram signal to obtain a corresponding electroencephalogram signal sequence; then adds classification labels and position encodings to the electroencephalogram signal sequence to obtain a corresponding target electroencephalogram signal sequence; then processes the target electroencephalogram signal sequence based on an anomaly detection layer in a trained anomaly detection model to obtain target features corresponding to the target electroencephalogram signal sequence; subsequently, the target features are predicted and processed based on a fully connected layer in the anomaly detection model to obtain a corresponding prediction result; further, a preset threshold value is acquired, and a numerical comparison result between the prediction result and the preset threshold value is generated; finally, a mild cognitive impairment detection result corresponding to the user is generated based on the numerical comparison result. Through pre-processing of the acquired original electroencephalogram signal of the user and adding classification labels and position encodings to obtain a target electroencephalogram signal sequence, and then using the trained anomaly detection model to predict and process the target electroencephalogram signal sequence to obtain a corresponding prediction result, and finally analyzing the prediction result based on a preset threshold value, a mild cognitive impairment detection result corresponding to the user can be quickly and accurately generated. The present application uses a mild cognitive impairment detection method based on an anomaly detection model, which no longer needs to be highly dependent on manual processing, and effectively improves the sensitivity and accuracy of mild cognitive impairment detection. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the schemes in the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0057] Figure 1 is an exemplary system architecture diagram to which the present application can be applied;

[0058] Figure 2 is a structural schematic diagram of an embodiment of the cognitive impairment detection device based on artificial intelligence according to the present application;

[0059] Figure 3 is a structural schematic diagram of an embodiment of the computer device according to the present application. DETAILED DESCRIPTION

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description herein is for describing particular embodiments only and is not intended to be limiting of the application; the use herein of terms such as "comprise" and "comprising", "have" and "having", "include" and "including" and "contain" and "containing" are to be construed in an open, non-exhaustive way, i.e. in the sense of "including, but not limited to"; the use herein of terms such as "first", "second", and "other" are used to distinguish one object from another, and are not used to describe a particular sequential order.

[0061] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another.

[0062] In order to make the persons skilled in the art better understand the scheme of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings.

[0063] As shown in Figure 1 The system architecture 100 can include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is a medium for providing a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0064] A user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0065] The terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smartphones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop portable computers, and desktop computers, etc.

[0066] The server 105 can be a server providing various services, for example, a background server providing support for a page displayed on the terminal device 101, 102, 103.

[0067] It should be noted that the cognitive impairment detection apparatus based on artificial intelligence provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the cognitive impairment detection apparatus based on artificial intelligence is generally arranged in a server / terminal device.

[0068] It should be understood that Figure 1 The number of terminal devices, networks and servers in

[0069] With reference to Figure 2 , a structural diagram of one embodiment of the cognitive impairment detection apparatus based on artificial intelligence according to the present application is shown. The order of the modules in the structural diagram can be changed, and some modules can be omitted, according to different needs. The cognitive impairment detection apparatus based on artificial intelligence provided in the embodiments of the present application can be applied to any processing scenario requiring shallow cognitive impairment detection, and then the cognitive impairment detection apparatus based on artificial intelligence can be applied to products in these scenarios. The cognitive impairment detection apparatus based on artificial intelligence 200 includes a first acquisition module 201, a preprocessing module 202, a first processing module 203, a second processing module 204, a prediction module 205, a first generation module 206, and a second generation module 207. Among them:

[0070] The first acquisition module 201 is configured to acquire the original brain electrical signal of the user.

[0071] In the present embodiment, the electronic device (for example Figure 1The server / terminal device shown) can obtain the original brain electrical signal through a wired connection or a wireless connection. It should be noted that the wireless connection can include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other now known or future developed wireless connection. The present application can be applied to the field of digital medicine, and can be specifically used in the medical business scenario of evaluating whether the user has the risk of mild cognitive impairment. The user can be a potential user who may have the risk of mild cognitive impairment, and the original brain electrical signal can be the multi-channel EEG signal data obtained by recording the brain electrical activity of the user during daily activities or specific tasks using an EEG (electroencephalogram) device during the evaluation test of mild cognitive impairment. Each channel represents the brain electrical activity recorded by the electrode at different positions on the scalp.

[0072] The preprocessing module 202 is configured to preprocess the original brain electrical signal to obtain a corresponding brain electrical signal sequence.

[0073] In the embodiment, the specific implementation process of preprocessing the original brain electrical signal to obtain a corresponding brain electrical signal sequence will be further described in detail in subsequent specific embodiments, and will not be described here.

[0074] The first processing module 203 is configured to add a classification label and a position code to the brain electrical signal sequence to obtain a corresponding target brain electrical signal sequence.

[0075] In the embodiment, a CLS label (classification label) can be added at the beginning of each time window of the brain electrical signal sequence to represent the global information of the entire time window. This step helps to retain the global context information in the subsequent Transformer model. Further, a position code is added to each time window to retain the time and channel information. This position code enables the Transformer model to utilize the time and spatial order information of the input data, enhancing the learning ability of the model.

[0076] The second processing module 204 is configured to process the target brain electrical signal sequence based on the anomaly detection layer in the trained anomaly detection model to obtain a target feature corresponding to the target brain electrical signal sequence.

[0077] In this embodiment, the target electroencephalogram signal sequence is input into the trained anomaly detection model. The anomaly detection model is specifically composed of an Anomaly Transformer and a fully connected layer, and the anomaly detection layer refers to the Anomaly Transformer. The Anomaly Transformer is alternately stacked by multiple Anomaly-Attention blocks and feedforward neural network layers. Each layer calculates the attention weight of the input signal and further processes it through the feedforward neural network. The stacking of these layers enables the model to extract and integrate deep features of the signal, i.e., the target features, layer by layer.

[0078] The prediction module 205 is configured to perform prediction processing on the target features based on the fully connected layer in the anomaly detection model to obtain a corresponding prediction result.

[0079] In this embodiment, the output of the anomaly detection layer, i.e., the target features, is input into the fully connected layer, and the corresponding prediction result is output by the fully connected layer. The prediction result output by the fully connected layer is a probability distribution, representing the prediction probability for each class ("existence of mild cognitive impairment" and "non-existence of mild cognitive impairment").

[0080] The first generation module 206 is configured to obtain a preset threshold and generate a numerical comparison result between the prediction result and the preset threshold.

[0081] In this embodiment, the value of the preset threshold is not specifically limited and can be specifically set to 0.5. This threshold can be adjusted according to actual conditions to optimize the performance of the model, for example, the best threshold can be determined by cross-validation and other methods. The numerical comparison result includes that the prediction result is greater than the preset threshold or the prediction result is less than the preset threshold.

[0082] The second generation module 207 is configured to generate a mild cognitive impairment detection result corresponding to the user based on the numerical comparison result.

[0083] In this embodiment, the specific implementation process of generating the mild cognitive impairment detection result corresponding to the user based on the numerical comparison result will be further described in detail in subsequent specific embodiments, and will not be described in detail here.

[0084] In the present application, the cognitive impairment detection device based on artificial intelligence first acquires the original electroencephalogram signal of a user; and pre-processes the original electroencephalogram signal to obtain a corresponding electroencephalogram signal sequence; then adds classification labels and position encodings to the electroencephalogram signal sequence to obtain a corresponding target electroencephalogram signal sequence; then processes the target electroencephalogram signal sequence based on an anomaly detection layer in a trained anomaly detection model to obtain target features corresponding to the target electroencephalogram signal sequence; subsequently, the target features are predicted and processed based on a fully connected layer in the anomaly detection model to obtain a corresponding prediction result; further, a preset threshold value is acquired, and a numerical comparison result between the prediction result and the preset threshold value is generated; finally, a mild cognitive impairment detection result corresponding to the user is generated based on the numerical comparison result. The present application pre-processes the acquired original electroencephalogram signal of the user and adds classification labels and position encodings to obtain a target electroencephalogram signal sequence, and then uses a trained anomaly detection model to predict and process the target electroencephalogram signal sequence to obtain a corresponding prediction result, which is finally analyzed based on a preset threshold value, so that a mild cognitive impairment detection result corresponding to the user can be quickly and accurately generated. The present application uses an anomaly detection model to detect mild cognitive impairment, which no longer needs to be highly dependent on manual processing, and effectively improves the sensitivity and accuracy of mild cognitive impairment detection.

[0085] In some optional implementations, the pre-processing module 202 described above includes:

[0086] A filtering sub-module configured to filter the original electroencephalogram signal to obtain a corresponding first electroencephalogram signal.

[0087] In the present embodiment, the original electroencephalogram signal can be filtered using a filter to extract signal components of a specific frequency band, thereby obtaining the first electroencephalogram signal. The specific frequency band can be determined according to actual business needs, for example, alpha waves (8-13 Hz) and beta waves (13-30 Hz). The first electroencephalogram signal after filtering removes most of the noise and retains the required signal components of interest.

[0088] A segmentation sub-module configured to segment the first electroencephalogram signal by channels to obtain a corresponding second electroencephalogram signal.

[0089] In the present embodiment, the first electroencephalogram signal can be segmented by channels according to a preset time window. Each time window contains data frames within a certain time, and each frame contains data of multiple channels. The appropriate size of the time window can be determined according to experimental needs and analysis purposes. The size of the time window contains enough data frames for subsequent analysis while avoiding data redundancy.

[0090] A flattening sub-module is configured to perform data flattening on the second EEG signal to obtain a third EEG signal.

[0091] In this embodiment, the data flattening operation refers to performing flattening on the signal data of each channel, that is, converting the multi-dimensional signal into a one-dimensional sequence.

[0092] A projection sub-module is configured to perform linear projection on the third EEG signal to obtain a fourth EEG signal.

[0093] In this embodiment, a dimension reduction method such as PCA, LDA, etc. can be used to remove redundant information in the third EEG signal while retaining the most important features. The linear projection processing can improve the quality of the data and reduce the impact of noise and redundancy on subsequent analysis.

[0094] A determination sub-module is configured to determine the fourth EEG signal as the target EEG signal.

[0095] The original EEG signal is filtered to obtain a first EEG signal. The first EEG signal is then segmented into channels to obtain a second EEG signal. The second EEG signal is then flattened to obtain a third EEG signal. The third EEG signal is then projected linearly to obtain a fourth EEG signal, which is determined as the target EEG signal. The specific preprocessing procedure is designed for the multi-channel characteristics of the EEG signal. The spatial and temporal characteristics of the EEG signal are retained by filtering, channel segmentation, data flattening, and linear projection, which improves the representation quality of the EEG signal data, thereby better capturing global and local features related to the detection of mild cognitive impairment and improving the accuracy of the detection of mild cognitive impairment.

[0096] In some optional implementations, the second generation module 307 includes:

[0097] A first determination sub-module is configured to determine whether the numerical comparison result is that the prediction result is greater than the preset threshold.

[0098] In this embodiment, the content of the numerical comparison result includes that the prediction result is greater than the preset threshold or the prediction result is less than the preset threshold.

[0099] A first generation sub-module is configured to generate a first mild cognitive impairment detection result that the user has mild cognitive impairment if the prediction result is greater than the preset threshold.

[0100] In the embodiment, if the numerical comparison result is that the prediction result is greater than the preset threshold, the prediction result is correspondingly converted into the first mild cognitive impairment detection result that the user has mild cognitive impairment.

[0101] The second generation sub-module is configured to generate a second mild cognitive impairment detection result that the user does not have mild cognitive impairment if the prediction result is less than the preset threshold.

[0102] In the embodiment, if the numerical comparison result is that the prediction result is less than the preset threshold, the prediction result is correspondingly converted into the first mild cognitive impairment detection result that the user does not have mild cognitive impairment.

[0103] The application determines whether the numerical comparison result is that the prediction result is greater than the preset threshold, generates the first mild cognitive impairment detection result that the user has mild cognitive impairment if the prediction result is greater than the preset threshold, and generates the second mild cognitive impairment detection result that the user does not have mild cognitive impairment if the prediction result is less than the preset threshold. The application can quickly and accurately generate the corresponding mild cognitive impairment detection result of the user by analyzing the content of the numerical comparison result, thereby ensuring the accuracy of the generated mild cognitive impairment detection result.

[0104] In some optional implementation manners of the embodiment, the cognitive impairment detection device based on artificial intelligence further includes:

[0105] The second acquisition module is configured to acquire a pre-constructed electroencephalogram signal sample.

[0106] In the embodiment, the specific implementation process of acquiring the pre-constructed electroencephalogram signal sample will be further described in detail in subsequent specific embodiments, and will not be described in detail here.

[0107] The division module is configured to divide the electroencephalogram signal sample into a training set and a test set.

[0108] In the embodiment, the electroencephalogram signal sample can be divided into the training set and the test set according to a preset division ratio. The numerical value of the division ratio is not specifically limited and can be set according to actual use requirements, for example, can be set to 7:3.

[0109] The third acquisition module is configured to acquire a preset initial anomaly detection model and perform model initialization on the initial anomaly detection model.

[0110] In the present embodiment, the initial anomaly detection model described above is specifically composed of an Anomaly Transformer model and a fully connected layer. Specifically, the overall structure of the Anomaly Transformer model is formed by alternating stacking of Anomaly-Attention blocks and feedforward layers, which facilitates the model to learn potential temporal correlations from multi-level features. The Anomaly-Attention mechanism is a key component of the model, which considers both prior correlations and sequence correlations. Specifically, the Anomaly-Attention blocks include: multi-layer stacking: the Anomaly Transformer model uses multi-layer stacked Anomaly-Attention blocks to process the input signal layer by layer. Each layer contains a self-attention mechanism and a feedforward neural network, which are used to capture the context information and features in the input sequence. Attention weight calculation: in each layer, the Anomaly-Attention block calculates the Attention weights between each element (i.e., the EEG signal at each time point) in the input sequence and other elements, to evaluate their correlation. These weights reflect the degree of influence between signals at different time points. Feedforward neural network: the calculated Attention weights are used for weighted summation to generate the context representation of each element. Then, these representations are further processed by the feedforward neural network to extract higher-level features. Anomaly attention mechanism: the Anomaly Transformer model introduces an anomaly attention mechanism to distinguish between normal and abnormal points through the distribution characteristics of the Attention weights. Abnormal points usually exhibit different patterns or distributions in the Attention weights than normal points.

[0111] The third processing module is configured to input the training set into the initial anomaly detection model for forward propagation, and calculate an error between a model output of the initial anomaly detection model and a true label of the training set; wherein the error includes a reconstruction error and a correlation difference.

[0112] In this embodiment, specifically, the process of inputting the training set into the initial anomaly detection model for forward propagation and calculating the error between the model output of the initial anomaly detection model and the true label of the training set includes: 1) calculating a priori correlation matrix. Gaussian kernel function: calculate the priori relationship between time points by using Gaussian kernel function. Gaussian kernel function can evaluate the similarity or correlation between time points according to the relative distance of time points. By adjusting the parameters (such as standard deviation σ) of Gaussian kernel, the characteristics and patterns of different time series can be adapted. Calculation method: for each time point i, calculate the priori correlation weight p_ij between it and other time points j in the sequence. These weights constitute a priori correlation matrix P, where each element p_ij represents the priori correlation strength between time points i and j. 2) Calculate sequence correlation matrix. Attention weight capture: the Attention weight calculated by the Anomaly-Attention block captures the patterns and anomalies in the signal. These weights reflect the actual correlation between different time points of the signal, which is a representation of the inherent structure of the model in the input sequence. 3) Generate sequence correlation matrix: normalize the Attention weight (such as using Softmax function) to generate sequence correlation matrix S. Each element s_ij in matrix S represents the actual correlation strength between time points i and j. 4) Correlation difference evaluation. Calculate difference (error): calculate the difference between priori correlation matrix P and sequence correlation matrix S by using symmetric KL divergence. This difference reflects the deviation between the actual correlation pattern in the time series and the priori correlation pattern, which is an important indicator for evaluating the degree of anomaly in the signal. Anomaly degree evaluation: according to the calculated correlation difference value, the degree of anomaly in the signal can be evaluated. Anomaly points usually result in larger correlation difference values because they have significant differences in correlation patterns with normal points.

[0113] The fourth processing module is configured to calculate the gradient based on the error, update the model weight of the initial anomaly detection model through the back propagation algorithm, and optimize the error by using the minimization strategy and the maximization strategy.

[0114] In this embodiment, the process of back propagation includes: 1. Calculate the gradient: according to the error (including reconstruction error and correlation difference) calculated by forward propagation, use chain rule to calculate the gradient of each parameter in the model. 2. Update model weight: use gradient descent algorithm or Adam algorithm to update the weight and bias of the model to minimize the reconstruction error with minimization strategy and maximize the correlation difference with maximization strategy, so that the model can better detect abnormal signals and improve the detection ability of the model.

[0115] The fifth processing module is configured to repeat the forward propagation and back propagation process until the preset iteration condition is met, and obtain a corresponding specified anomaly detection model.

[0116] In the embodiment, the preset iteration condition can refer to a preset iteration number or meeting a certain stop condition (e.g., performance on a validation set no longer significantly improves). The iteration number is not limited in particular, and can be determined according to actual business requirements.

[0117] The test module is configured to test the specified anomaly detection model based on the test set.

[0118] In the embodiment, the specific implementation process of testing the specified anomaly detection model based on the test set will be further described in detail in subsequent embodiments, and will not be elaborated here.

[0119] The determination module is configured to determine the specified anomaly detection model as the anomaly detection model if the specified anomaly detection model passes the test.

[0120] In the embodiment, if the specified anomaly detection model passes the test, it indicates that the model effect of the generated specified anomaly detection model has reached the expected effect, and then the specified anomaly detection model is determined as the anomaly detection model. In the present application, the Attention mechanism combining prior correlation and sequence correlation is introduced into the Anomaly Transformer model. The prior correlation represents the prior relationship between time points through a Gaussian kernel function, and the sequence correlation captures patterns and anomalies in the signal through Attention weights. This mechanism enables the model to more accurately identify small changes and abnormal patterns in the EEG signal, which may be related to early symptoms of mild cognitive impairment (MCI). By capturing these subtle changes, the model can more effectively detect MCI in the early stages.

[0121] The application obtains a pre-constructed electroencephalogram signal sample, divides the electroencephalogram signal sample into a training set and a test set, obtains a preset initial anomaly detection model, and performs model initialization on the initial anomaly detection model. Then, the training set is input into the initial anomaly detection model for forward propagation, and the error between the model output of the initial anomaly detection model and the true label of the training set is calculated. Subsequently, the gradient is calculated based on the error, the model weight of the initial anomaly detection model is updated through a back propagation algorithm, and the error is optimized by using a minimization strategy and a maximization strategy. The forward propagation and back propagation processes are repeated until the preset iteration condition is met, and a corresponding specified anomaly detection model is obtained. Finally, the specified anomaly detection model is tested based on the test set. If the specified anomaly detection model passes the test, the specified anomaly detection model is used as the anomaly detection model. The application obtains a pre-constructed electroencephalogram signal sample, and then uses the Attention mechanism combined with priori and sequence association and an optimized training strategy to train the initial anomaly detection model using the electroencephalogram signal sample to obtain a trained anomaly detection model, thereby effectively improving the detection performance and reliability of the anomaly detection model for shallow cognitive impairment detection.

[0122] In some optional implementations, the second obtaining module includes:

[0123] The first obtaining sub-module is configured to obtain an initial electroencephalogram signal sample collected in advance.

[0124] In this embodiment, the initial electroencephalogram signal sample can be a sample data obtained by selecting 1.5 s of electroencephalogram signal data from the sampling data of the electroencephalogram signal at 500 Hz and performing manual labeling. The label of manual labeling includes the presence of shallow cognitive impairment or the absence of shallow cognitive impairment.

[0125] The calling sub-module is configured to call a preset sample construction model.

[0126] In the embodiment, the sample construction model can be specifically a model based on a generative adversarial network. The sample construction model is composed of a generator and a discriminator. The generator is responsible for generating fake samples similar to the initial electroencephalogram signal samples. The generator is a neural network, and the input is random noise, and the output is the generated fake sample. The discriminator is responsible for distinguishing whether the input sample is real or fake generated by the generator. The discriminator is also a neural network, and the input is a sample, and the output is the probability that the sample is a real sample. The training process of the sample generation model includes: alternating training: in each round of training, first fix the parameters of the generator, train the discriminator to distinguish between real samples and fake samples; then fix the parameters of the discriminator, train the generator to generate more realistic fake samples, so that the discriminator cannot distinguish. Optimization goal: the goal of the generator is to generate as realistic fake samples as possible, so that the discriminator cannot distinguish; the goal of the discriminator is to distinguish real samples and fake samples as accurately as possible. After the training of the sample construction model is completed, a large number of fake samples similar to the real samples but with diversity can be generated by using the sample construction model, so as to enhance the diversity of the sample data.

[0127] The construction sub-module is configured to perform sample construction processing on the initial electroencephalogram signal samples based on the sample construction model to obtain corresponding expanded samples.

[0128] In the embodiment, a large number of fake samples (i.e., the expanded samples) similar to the real samples (i.e., the initial electroencephalogram signal samples) but with diversity can be generated by using the sample construction model to perform sample construction processing on the initial electroencephalogram signal samples, so as to effectively enhance the diversity of the sample data.

[0129] The third generation sub-module is configured to generate the electroencephalogram signal samples based on the initial electroencephalogram signal samples and the expanded samples.

[0130] In the embodiment, the corresponding integrated samples can be obtained by integrating the initial electroencephalogram signal samples and the expanded samples, and the integrated samples can be used as the final electroencephalogram signal samples.

[0131] The application obtains a pre-acquired initial electroencephalogram signal sample, then calls a preset sample construction model, then performs sample construction processing on the initial electroencephalogram signal sample based on the sample construction model to obtain a corresponding expanded sample, and subsequently generates the electroencephalogram signal sample based on the initial electroencephalogram signal sample and the expanded sample. The application can intelligently and quickly generate a corresponding expanded sample based on the initial electroencephalogram signal sample based on the use of the sample construction model, and then generate the required electroencephalogram signal sample based on the initial electroencephalogram signal sample and the expanded sample to complete the expansion of the electroencephalogram signal sample, effectively enhancing the diversity of the electroencephalogram signal sample, thereby facilitating the improvement of the generalization ability and accuracy of the subsequently trained abnormality detection model.

[0132] In some optional implementations of the embodiment, the test module includes:

[0133] The second obtaining sub-module is configured to input the test set into the specified abnormality detection model and obtain a performance index value of the specified abnormality detection model, wherein the performance index value is a value corresponding to a preset performance index, and the number of performance indexes includes a plurality of performance indexes.

[0134] In the embodiment, the test data set is input into the specified abnormality detection model for testing, and the performance index value of the specified abnormality detection model corresponding to the preset performance index is recorded after the model test is completed. The performance index includes a recall rate index and an accuracy rate index. Correspondingly, the performance index value includes a recall rate value and an accuracy rate value.

[0135] The second judgment sub-module is configured to judge whether all the performance index values are within the corresponding preset value range.

[0136] In the embodiment, according to the actual model performance test requirement, a preset value range conforming to the normal expectation is constructed for each performance index. Specifically, a recall rate value range matching the recall rate index and an accuracy rate value range matching the accuracy rate index are constructed in advance.

[0137] The first determination sub-module is configured to determine that the specified abnormality detection model passes the test if all the performance index values are within the corresponding preset value range.

[0138] In the embodiment, if all the performance index values are within the corresponding preset value range, it indicates that the model effect (recall rate and accuracy rate) of the generated specified abnormality detection model reaches the expected effect, and then it is determined that the specified abnormality detection model passes the test.

[0139] The second determining sub-module is configured to determine that the specified anomaly detection model fails the test if none of the performance indicator values is within the corresponding preset value range.

[0140] In this embodiment, if at least one of the performance indicator values is not within the corresponding preset value range, it indicates that the model effect of the generated specified anomaly detection model does not achieve the expected effect, and then it is determined that the specified anomaly detection model fails the test.

[0141] In this embodiment, the test set is input into the specified anomaly detection model, and the performance indicator values of the specified anomaly detection model are obtained. Then, it is determined whether all the performance indicator values are within the corresponding preset value range. If all the performance indicator values are within the corresponding preset value range, it is determined that the specified anomaly detection model passes the test. If none of the performance indicator values is within the corresponding preset value range, it is determined that the specified anomaly detection model fails the test. In this embodiment, the test set is input into the specified anomaly detection model, and the performance indicator values of the specified anomaly detection model are obtained. Then, the performance indicator values of the specified anomaly detection model are evaluated based on the use of the preset value range, so that the performance test result corresponding to the specified anomaly detection model can be accurately generated, and the accuracy of the obtained performance test result is ensured.

[0142] In some optional implementation manners of this embodiment, the above-mentioned artificial intelligence-based cognitive impairment detection device further includes:

[0143] The third generating module is configured to generate a corresponding target detection report based on the mild cognitive impairment detection result.

[0144] In this embodiment, the specific implementation process of generating the target detection report based on the mild cognitive impairment detection result will be further described in detail in subsequent specific embodiments, and will not be described here in more detail.

[0145] The fourth obtaining module is configured to obtain communication information of the user.

[0146] In this embodiment, the communication information can refer to a mobile phone number or an email address of the user.

[0147] The pushing module is configured to push the target detection report to the user based on the communication information.

[0148] In this embodiment, the target detection report can be sent to the communication terminal corresponding to the user according to the obtained communication information, so as to complete the pushing process of the target detection report.

[0149] The application generates a corresponding target detection report based on the mild cognitive impairment detection result; then acquires the communication information of the user; and subsequently pushes the target detection report to the user based on the communication information. After generating the mild cognitive impairment detection result corresponding to the user, the application intelligently generates a corresponding target detection report based on the mild cognitive impairment detection result, realizes the automatic generation of the target detection report, so that subsequent manual report generation operations are not required, and the generation efficiency of the target detection report is improved. In addition, the target detection report is pushed to the user based on the communication information of the user, so that the user can timely understand the mild cognitive impairment detection situation of the user and perform corresponding subsequent processing, and the use experience of the user is improved.

[0150] In some optional implementations of the embodiment, the third generation module includes:

[0151] A conversion submodule is configured to convert the mild cognitive impairment detection result into a target detection result corresponding to a preset understanding style.

[0152] In the embodiment, the understanding style refers to a non-professional style that is easy to understand. If the mild cognitive impairment detection result indicates that the user has mild cognitive impairment, the content of the corresponding target detection result can be: "According to our assessment, you are currently facing a certain risk of mild cognitive impairment." If the mild cognitive impairment detection result indicates that the user does not have mild cognitive impairment, the content of the corresponding target detection result can be: "Our assessment result shows that you are currently not at significant risk of mild cognitive impairment."

[0153] A third acquisition submodule is configured to acquire medical advice information corresponding to the mild cognitive impairment detection result.

[0154] In the embodiment, if the mild cognitive impairment detection result indicates that the user has mild cognitive impairment, the corresponding medical advice information includes: "Given that you are at risk of mild cognitive impairment, we recommend that you undergo further medical evaluation to better understand your cognitive status." If the mild cognitive impairment detection result indicates that the user does not have mild cognitive impairment, the corresponding medical advice information includes: "We recommend that you regularly monitor your cognitive health to timely discover and address any potential issues."

[0155] A fourth acquisition submodule is configured to acquire lifestyle advice information corresponding to the mild cognitive impairment detection result.

[0156] In the embodiment, if the mild cognitive impairment detection result is that the user has mild cognitive impairment, the corresponding lifestyle suggestion information includes: "At the same time, we suggest you maintain a healthy lifestyle, which helps prevent and delay the development of cognitive impairment." If the mild cognitive impairment detection result is that the user does not have mild cognitive impairment, the corresponding lifestyle suggestion information includes: "Although you currently do not have significant risk of mild cognitive impairment, we still suggest you maintain a healthy lifestyle to prevent future cognitive problems."

[0157] The fifth obtaining sub-module is configured to obtain a preset detection report template.

[0158] In the embodiment, the detection report template can be a template file constructed according to actual business needs, which includes a detection result area, a medical advice area, and a lifestyle suggestion area.

[0159] The filling sub-module is configured to fill the target detection result, the medical advice information, and the lifestyle suggestion information into the detection report template to obtain the target detection report.

[0160] In the embodiment, the target detection report can be obtained by filling the target detection result, the medical advice information, and the lifestyle suggestion information into the detection result area, the medical advice area, and the lifestyle suggestion area of the detection report template, respectively.

[0161] The application converts the mild cognitive impairment detection result into a target detection result corresponding to a preset understanding style, then obtains medical advice information corresponding to the mild cognitive impairment detection result, and obtains lifestyle suggestion information corresponding to the mild cognitive impairment detection result, and then obtains a preset detection report template. Subsequently, the target detection result, the medical advice information, and the detection report template are filled into the detection report template to obtain the target detection report. The application converts the mild cognitive impairment detection result into a target detection result corresponding to a preset understanding style, and obtains medical advice information corresponding to the mild cognitive impairment detection result and lifestyle suggestion information corresponding to the mild cognitive impairment detection result. Then, the target detection result, the medical advice information, and the lifestyle suggestion information are filled into the preset detection report template, thereby automatically and intelligently completing the construction of the target detection report and improving the content richness of the target detection report.

[0162] In some optional implementations of the present embodiment, there are also some machine learning-based mild cognitive impairment detection methods in the prior art, however, these detection methods lack automation and real-time processing capability, mostly rely on fixed feature sets, and are difficult to adapt to individual differences and dynamically changing electroencephalogram signals. In addition, these methods often face the problems of dimension disaster and high computational complexity when processing large-scale, multi-dimensional EEG data, resulting in unstable detection effect and insufficient generalization ability. Finally, the existing detection system usually lacks sensitivity to abnormal brain electrical activity, and cannot timely capture and identify weak MCI early signals.

[0163] The present application can effectively solve the problems of existing methods, such as complex and time-consuming manual feature extraction, reliance on fixed feature sets, difficulty in adapting to individual differences and dynamically changing electroencephalogram signals, and facing dimension disaster and high computational complexity when processing large-scale, multi-dimensional EEG data. Anomaly Transformer can automatically learn complex patterns and features in EEG signals by modeling point representation and global association of time series, reducing the dependence on manual feature extraction. At the same time, the anomaly attention mechanism improves the sensitivity and accuracy of MCI detection by capturing small abnormal changes in EEG signals.

[0164] Specifically, Anomaly Transformer generates association weight distribution through self-attention mechanism, which reflects the rich association of each time point with the entire sequence. Due to its rarity, abnormal points are difficult to establish significant association with the entire sequence, which makes the association of abnormal points mainly concentrated in its adjacent time points. This neighborhood concentration bias can effectively distinguish normal and abnormal points. By introducing an anomaly detection criterion based on association difference, Anomaly Transformer uses the minimum-maximum strategy to amplify the discrimination of normal and abnormal points, thereby realizing real-time and automated processing of EEG signals and improving the generalization ability and robustness of the model. This method not only achieves advanced detection results on multiple unsupervised time series anomaly detection benchmarks, but also significantly improves the efficiency and accuracy of MCI early detection.

[0165] The present application can be applied in the field of smart medical treatment, thereby promoting the construction of smart city.

[0166] It should be emphasized that, in order to further ensure the privacy and security of the above mild cognitive impairment detection results, the above mild cognitive impairment detection results can also be stored in a node of a block chain.

[0167] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0168] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0169] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0170] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 3 , Figure 3 This is a basic structural block diagram of the computer device in this embodiment.

[0171] The computer device 3 includes a memory 31, a processor 32, and a network interface 33 that are interconnected via a system bus. It should be noted that only the computer device 3 with components 31-33 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0172] The computer device can be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The computer device can interact with a user through a keyboard, a mouse, a remote controller, a touchpad, a voice control device, or the like.

[0173] The memory 31 includes at least one type of readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, or the like), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, or the like. In some embodiments, the memory 31 can be an internal storage unit of the computer device 3, such as a hard disk or a memory of the computer device 3. In other embodiments, the memory 31 can also be an external storage device of the computer device 3, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, or the like. Of course, the memory 31 can include both an internal storage unit and an external storage device of the computer device 3. In this embodiment, the memory 31 is generally used to store an operating system and various application software installed in the computer device 3, such as computer readable instructions of the cognitive impairment detection apparatus based on artificial intelligence, or the like. In addition, the memory 31 can also be used to temporarily store various data that have been output or will be output.

[0174] The processor 32 can be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 32 is generally used to control the overall operation of the computer device 3. In this embodiment, the processor 32 is used to run computer readable instructions or process data stored in the memory 31, such as computer readable instructions of the cognitive impairment detection apparatus based on artificial intelligence.

[0175] The network interface 33 can include a wireless network interface or a wired network interface, and is generally used to establish a communication connection between the computer device 3 and other electronic devices.

[0176] The functions of the modules in the cognitive impairment detection apparatus based on artificial intelligence of the above embodiments are implemented by the processor executing computer readable instructions stored in the memory.

[0177] In the embodiment, the cognitive impairment detection device based on artificial intelligence first acquires the original electroencephalogram signal of a user; and pre-processes the original electroencephalogram signal to obtain a corresponding electroencephalogram signal sequence; then adds classification labels and position encodings to the electroencephalogram signal sequence to obtain a corresponding target electroencephalogram signal sequence; then processes the target electroencephalogram signal sequence based on an anomaly detection layer in a trained anomaly detection model to obtain target features corresponding to the target electroencephalogram signal sequence; subsequently performs prediction processing on the target features based on a fully connected layer in the anomaly detection model to obtain a corresponding prediction result; further acquires a preset threshold and generates a numerical comparison result between the prediction result and the preset threshold; and finally generates a mild cognitive impairment detection result corresponding to the user based on the numerical comparison result. The present application obtains a target electroencephalogram signal sequence by pre-processing the acquired original electroencephalogram signal of a user and adding classification labels and position encodings, and then uses a trained anomaly detection model to perform prediction processing on the target electroencephalogram signal sequence to obtain a corresponding prediction result, and finally analyzes the prediction result based on a preset threshold, so as to quickly and accurately generate a mild cognitive impairment detection result corresponding to the user. The present application uses the method of detecting mild cognitive impairment based on an anomaly detection model, no longer needs to highly rely on manual processing, and effectively improves the sensitivity and accuracy of mild cognitive impairment detection.

[0178] The present application also provides another implementation, namely providing a computer readable storage medium storing computer readable instructions executable by at least one processor to cause the at least one processor to perform the functions of each module in the cognitive impairment detection device based on artificial intelligence as described above.

[0179] In the embodiment, the cognitive impairment detection device based on artificial intelligence first acquires the original electroencephalogram signal of the user; and pre-processes the original electroencephalogram signal to obtain a corresponding electroencephalogram signal sequence; then adds classification labels and position encodings to the electroencephalogram signal sequence to obtain a corresponding target electroencephalogram signal sequence; then processes the target electroencephalogram signal sequence based on an anomaly detection layer in a trained anomaly detection model to obtain target features corresponding to the target electroencephalogram signal sequence; subsequently, the target features are predicted and processed based on a fully connected layer in the anomaly detection model to obtain a corresponding prediction result; further, a preset threshold value is acquired, and a numerical comparison result between the prediction result and the preset threshold value is generated; finally, a shallow cognitive impairment detection result corresponding to the user is generated based on the numerical comparison result. The present application obtains a target electroencephalogram signal sequence by pre-processing the acquired original electroencephalogram signal of the user and adding classification labels and position encodings, and then uses a trained anomaly detection model to predict and process the target electroencephalogram signal sequence to obtain a corresponding prediction result, and finally analyzes the prediction result based on a preset threshold value, which can quickly and accurately generate a shallow cognitive impairment detection result corresponding to the user. The present application uses an anomaly detection model to detect shallow cognitive impairment, which no longer needs to rely heavily on manual processing, and effectively improves the sensitivity and accuracy of shallow cognitive impairment detection.

[0180] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be realized by means of software and a necessary general hardware platform, of course, they can also be realized by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, an optical disk), and includes a plurality of instructions for making a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in various embodiments of the present application.

[0181] Obviously, the above-described embodiments are only some embodiments but not all the embodiments of the present application, the preferred embodiments of the present application are shown in the drawings, but do not limit the patent scope of the present application. The present application can be implemented in many different forms, and conversely, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent replacements to some technical features therein. Any equivalent structure made by using the content of the specification and drawings, directly or indirectly applied to other related technical fields, is also within the patent protection scope of the present application.

Claims

1. A cognitive impairment detection device based on artificial intelligence, characterized in that, include: The first acquisition module is used to acquire the user's raw EEG signals; The preprocessing module is used to preprocess the raw EEG signals to obtain the corresponding EEG signal sequence; The first processing module is used to add classification markers and position codes to the EEG signal sequence to obtain the corresponding target EEG signal sequence; wherein, a classification marker is added at the beginning of each time window of the EEG signal sequence to represent the global information of the entire time window; and a position code is added to each time window to retain time and channel information; The second processing module is used to process the target EEG signal sequence based on the anomaly detection layer in the trained anomaly detection model to obtain target features corresponding to the target EEG signal sequence. The prediction module is used to perform prediction processing on the target features based on the fully connected layer in the anomaly detection model to obtain the corresponding prediction results; The first generation module is used to obtain a preset threshold and generate a numerical comparison result between the prediction result and the preset threshold; The second generation module is used to generate a shallow cognitive impairment detection result corresponding to the user based on the numerical comparison result; The artificial intelligence-based cognitive impairment detection device further includes: The second acquisition module is used to acquire pre-constructed EEG signal samples; A partitioning module is used to divide the EEG signal samples into a training set and a test set; The third acquisition module is used to acquire a preset initial anomaly detection model and initialize the initial anomaly detection model. The third processing module is used to input the training set into the initial anomaly detection model for forward propagation, and to calculate the error between the model output of the initial anomaly detection model and the true label of the training set. The fourth processing module is used to calculate the gradient based on the error, update the model weights of the initial anomaly detection model through the backpropagation algorithm, and optimize the error using a minimization strategy and a maximization strategy. The fifth processing module is used to repeat the forward and backward propagation process until the preset iteration conditions are met, and obtain the corresponding specified anomaly detection model. The testing module is used to test the specified anomaly detection model based on the test set; The determination module is used to determine the specified anomaly detection model as the anomaly detection model if the specified anomaly detection model passes the test.

2. The artificial intelligence-based cognitive impairment detection device according to claim 1, characterized in that, The preprocessing module includes: The filtering submodule is used to filter the original EEG signal to obtain the corresponding first EEG signal; The segmentation submodule is used to perform channel segmentation processing on the first EEG signal to obtain the corresponding second EEG signal; The flattening submodule is used to perform data flattening operations on the second EEG signal to obtain the corresponding third EEG signal; The projection submodule is used to perform linear projection processing on the third EEG signal to obtain the corresponding fourth EEG signal. A determination submodule is used to use the fourth EEG signal as the target EEG signal.

3. The artificial intelligence-based cognitive impairment detection device according to claim 1, characterized in that, The second generation module includes: The first judgment submodule is used to determine whether the numerical comparison result is that the predicted result is greater than the preset threshold. The first generation submodule is used to generate a first shallow cognitive impairment detection result if the prediction result is greater than the preset threshold. The second generation submodule is used to generate a second shallow cognitive impairment detection result if the prediction result is less than the preset threshold, indicating that the user does not have shallow cognitive impairment.

4. The artificial intelligence-based cognitive impairment detection device according to claim 1, characterized in that, The second acquisition module includes: The first acquisition submodule is used to acquire pre-collected initial EEG signal samples; Call the submodule to invoke the preset sample to build the model; A construction submodule is used to perform sample construction processing on the initial EEG signal sample based on the sample construction model to obtain the corresponding expanded sample; The third generation submodule is used to generate the EEG signal sample based on the initial EEG signal sample and the expanded sample.

5. The artificial intelligence-based cognitive impairment detection device according to claim 1, characterized in that, The testing module includes: The second acquisition submodule is used to input the test set into the specified anomaly detection model and acquire the performance index value of the specified anomaly detection model; wherein, the performance index value is a value corresponding to a preset performance index, and the number of performance indices includes multiple; The second judgment submodule is used to determine whether all the performance index values ​​are within the corresponding preset value range; The first determination submodule is used to determine that the specified anomaly detection model passes the test if all values ​​are within the corresponding preset value range. The second determination submodule is used to determine that the specified anomaly detection model has failed the test if all values ​​are not within the corresponding preset value range.

6. The artificial intelligence-based cognitive impairment detection device according to claim 1, characterized in that, The AI-based cognitive impairment detection device also includes: The third generation module is used to generate a corresponding target detection report based on the shallow cognitive impairment detection results; The fourth acquisition module is used to acquire the user's communication information; The push module is used to push the target detection report to the user based on the communication information.

7. The artificial intelligence-based cognitive impairment detection device according to claim 6, characterized in that, The third generation module includes: The conversion submodule is used to convert the shallow cognitive impairment detection results into target detection results corresponding to a preset comprehension style; The third acquisition submodule is used to acquire medical advice information corresponding to the shallow cognitive impairment detection results; The fourth acquisition submodule is used to acquire lifestyle recommendation information corresponding to the shallow cognitive impairment detection results; The fifth submodule is used to obtain a preset test report template; The fill submodule is used to fill the target detection results, the medical advice information, and the lifestyle advice information into the detection report template to obtain the target detection report.

8. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the functions of each module in the artificial intelligence-based cognitive impairment detection device as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the functions of each module in the artificial intelligence-based cognitive impairment detection device as described in any one of claims 1 to 7.

Images