Music-based Alzheimer's disease cognition improvement evaluation system and method

Abstract

The invention discloses a music-based Alzheimer's disease cognition improvement evaluation system and method, and relates to the technical field of cognition improvement evaluation, and the system comprises a to-be-evaluated data acquisition module, a data processing module, and a cognition improvement evaluation module. According to the method, time domain feature extraction, frequency domain feature extraction and nonlinear dynamic feature extraction are performed on multi-channel electroencephalogram data to obtain multi-scale feature parameters before and after a music scheme is executed, and a machine learning regression model is adopted to obtain a cognitive assessment score before execution and a cognitive assessment score after execution; therefore, the cognitive improvement evaluation result corresponding to the music execution record data is obtained, and the accuracy of Alzheimer's disease cognitive improvement evaluation based on music is improved.

Description

Technical Field

[0001] This application relates to the technical field of cognitive improvement assessment, and particularly to a music-based Alzheimer's disease cognitive improvement assessment system and method. Background Art

[0002] Alzheimer's disease patients usually retain musical memories because the key brain regions associated with musical memories are relatively unaffected by the disease. Research has shown that listening to music or singing can provide emotional and cognitive function improvement effects for people with Alzheimer's disease and other types of dementia. Therefore, research on music-based Alzheimer's disease cognitive improvement assessment is needed.

[0003] In the prior art, Chinese Patent CN113082447A discloses a method for predicting the music-modulated brain plasticity effect of an fMRI brain circuit, including the following steps: S1, performing cognitive screening, collecting resting-state magnetic resonance and structural magnetic resonance data and cognitive test scores of subjects with normal cognitive function; S2, performing an 8-week music intervention training on the subjects; S3, performing cognitive function evaluation on the trained subjects; S4, dividing all subjects into an effective intervention group and an ineffective intervention group according to the third quartile of the change amount, and using the classification result as the label of the subject; S5, training a prediction model.

[0004] The above prior art does not consider the association between cognitive function and electroencephalogram data, and the accuracy of cognitive improvement assessment is poor. Summary of the Invention

[0005] This application provides a music-based Alzheimer's disease cognitive improvement assessment system and method to solve the problem that the existing cognitive improvement assessment technology does not consider the association between cognitive function and electroencephalogram data, and the accuracy of cognitive improvement assessment is poor.

[0006] On the one hand, this application provides a music-based Alzheimer's disease cognitive improvement assessment system, including: a to-be-assessed data acquisition module, a data processing module, and a cognitive improvement assessment module.

[0007] The to-be-assessed data acquisition module is used to acquire the music execution record data of the to-be-assessed user, including the music scheme and multi-channel electroencephalogram data before and after the execution of the music scheme.

[0008] The data processing module is used to extract time-domain features, frequency-domain features, and non-linear dynamics features from the multi-channel electroencephalogram data to obtain multi-scale feature parameters before and after the execution of the music scheme.

[0009] The cognitive improvement evaluation module is used to obtain the pre - execution cognitive evaluation score and the post - execution cognitive evaluation score respectively according to the multi - scale feature parameters before and after the execution of the music plan by using a machine learning regression model.

[0010] The cognitive improvement evaluation module is also used to obtain the cognitive improvement evaluation result corresponding to the music execution record data according to the pre - execution cognitive evaluation score and the post - execution cognitive evaluation score.

[0011] In a possible implementation manner, the data processing module is configured to: before performing time - domain feature extraction, frequency - domain feature extraction, and non - linear dynamics feature extraction on the multi - channel EEG data, pre - process the multi - channel EEG data first.

[0012] The pre - processing includes: electrode positioning, re - referencing, band - pass filtering, principal component analysis, and interference rejection.

[0013] In a possible implementation manner, the time - domain feature extraction includes: extracting the mean absolute value, the number of zero crossings, the slope sign change, and the waveform length.

[0014] The frequency - domain feature extraction includes: extracting the power spectral density.

[0015] The non - linear dynamics feature extraction includes: performing multi - dimensional phase - space reconstruction and extracting the correlation dimension, Lyapunov exponent, sample entropy, and approximate entropy.

[0016] In a possible implementation manner, the data processing module is configured to: after performing time - domain feature extraction, frequency - domain feature extraction, and non - linear dynamics feature extraction on the multi - channel EEG data, perform standardization processing on the multi - scale feature parameters.

[0017] In a possible implementation manner, the machine learning regression model uses a convolutional neural network and a bidirectional long short - term memory network model, introduces an attention mechanism, and builds a nine - layer neural network, including: a first input layer, a second convolutional layer, a third pooling layer, a fourth BiLSTM layer, a fifth BiLSTM layer, a sixth self - attention layer, a seventh dropout layer, an eighth fully - connected layer, and a ninth regression layer.

[0018] The machine learning regression model has been pre - trained using an EEG data training set, and each piece of data in the EEG data training set includes a multi - scale feature training parameter and a corresponding label score.

[0019] In a possible implementation manner, the Alzheimer's disease cognitive improvement evaluation system based on music further includes: an improvement evaluation database module, a to - be - executed data acquisition module, a case matching module, and a music plan generation module.

[0020] The improvement evaluation database module is used to generate an improvement evaluation dataset according to the music execution record data and the corresponding cognitive improvement evaluation results, and each piece of data in the improvement evaluation dataset includes a piece of music execution record data and the corresponding cognitive improvement evaluation result.

[0021] The to-be-executed data acquisition module is used to acquire the initial EEG data of the user to be executed.

[0022] The case matching module is used to match several pieces of music execution record data and the corresponding cognitive improvement evaluation results from the improvement evaluation dataset according to the initial EEG data as the matching result.

[0023] The music scheme generation module is used to select the music scheme in the music execution record data corresponding to the optimal cognitive improvement evaluation result from the matching result as the reference music scheme.

[0024] In a possible implementation manner, the case matching module uses the random forest algorithm to match the multi-channel EEG data before the execution of the music scheme in all the music execution record data in the improvement evaluation dataset according to the initial EEG data.

[0025] In a possible implementation manner, the music scheme is divided into three modes, including: passive music scheme mode, active music scheme mode, and re-creation music scheme mode.

[0026] The passive music scheme mode adopts the mode of playing music, the active music scheme mode adopts the mode of singing music, and the re-creation music scheme mode adopts the mode of music creation.

[0027] On the other hand, the present application provides a method for evaluating the cognitive improvement of Alzheimer's disease based on music, including the following steps:

[0028] Step 1: Acquire the music execution record data of the user to be evaluated, including the music scheme and the multi-channel EEG data before and after the execution of the music scheme.

[0029] Step 2: Extract time-domain features, frequency-domain features, and non-linear dynamic features from the multi-channel EEG data to obtain multi-scale feature parameters before and after the execution of the music scheme.

[0030] Step 3: Use a machine learning regression model to obtain the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score respectively according to the multi-scale feature parameters before and after the execution of the music scheme.

[0031] Step 4: Obtain the cognitive improvement evaluation result corresponding to the music execution record data according to the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score.

[0032] In a possible implementation, after step four, the following steps are further included:

[0033] Step five: Generate an improvement evaluation data set according to the music execution record data and the corresponding cognitive improvement evaluation results. Each piece of data in the improvement evaluation data set includes a piece of music execution record data and the corresponding cognitive improvement evaluation result.

[0034] Step six: Obtain the initial electroencephalogram data of the user to be executed.

[0035] Step seven: Match several pieces of music execution record data and the corresponding cognitive improvement evaluation results from the improvement evaluation data set according to the initial electroencephalogram data as the matching results.

[0036] Step eight: Select the music scheme in the music execution record data corresponding to the optimal cognitive improvement evaluation result from the matching results as the reference music scheme.

[0037] The Alzheimer's disease cognitive improvement evaluation system and method based on music in this application have the following advantages:

[0038] By performing time-domain feature extraction, frequency-domain feature extraction, and non-linear dynamics feature extraction on multi-channel electroencephalogram data, multi-scale feature parameters before and after the execution of the music scheme are obtained. A machine learning regression model is used to obtain the cognitive evaluation score before execution and the cognitive evaluation score after execution, and then the cognitive improvement evaluation result corresponding to the music execution record data is obtained, improving the accuracy of Alzheimer's disease cognitive improvement evaluation based on music.

[0039] The proposed time-domain feature extraction includes: extracting the mean absolute value, zero-crossing number, slope sign change, and waveform length. The frequency-domain feature extraction includes: extracting the power spectral density. The non-linear dynamics feature extraction includes: performing multi-dimensional phase space reconstruction and extracting the correlation dimension, Lyapunov exponent, sample entropy, and approximate entropy, improving the comprehensiveness of electroencephalogram data feature extraction.

[0040] The proposed machine learning regression model uses a convolutional neural network and a bidirectional long short-term memory network model, introduces an attention mechanism, and establishes a nine-layer neural network. By combining the convolutional neural network, bidirectional long short-term memory network, and attention mechanism, the accuracy of cognitive improvement evaluation is further improved.

[0041] The proposed case matching module uses the random forest algorithm to match the multi-channel electroencephalogram data before the execution of the music scheme in all music execution record data in the improvement evaluation data set according to the initial electroencephalogram data, improving the accuracy of case matching and thus improving the reliability of the reference music scheme.

[0042] The proposed passive music scheme mode, active music scheme mode, and re - creative music scheme mode improve the diversity of music schemes. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0044] Figure 1 It is a schematic diagram of the modules of the Alzheimer's disease cognitive improvement evaluation system based on music provided by the embodiments of the present application;

[0045] Figure 2 It is another schematic diagram of the modules of the Alzheimer's disease cognitive improvement evaluation system based on music provided by the embodiments of the present application;

[0046] Figure 3 It is a schematic flowchart of the method for evaluating Alzheimer's disease cognitive improvement based on music provided by the embodiments of the present application;

[0047] Figure 4 It is another schematic flowchart of the method for evaluating Alzheimer's disease cognitive improvement based on music provided by the embodiments of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0049] As Figure 1 shown, the embodiments of the present application provide an Alzheimer's disease cognitive improvement evaluation system based on music, including: a module for obtaining data to be evaluated, a data processing module, and a cognitive improvement evaluation module.

[0050] The module for obtaining data to be evaluated is used to obtain the music execution record data of the user to be evaluated, including music schemes and multi - channel electroencephalogram data before and after the execution of the music scheme.

[0051] The data processing module is used to extract time - domain features, frequency - domain features, and non - linear dynamics features from the multi - channel electroencephalogram data to obtain multi - scale feature parameters before and after the execution of the music scheme.

[0052] The cognitive improvement evaluation module is used to obtain the pre - execution cognitive evaluation score and the post - execution cognitive evaluation score respectively according to the multi - scale feature parameters before and after the implementation of the music plan by using a machine learning regression model.

[0053] The cognitive improvement evaluation module is also used to obtain the cognitive improvement evaluation result corresponding to the music execution record data according to the pre - execution cognitive evaluation score and the post - execution cognitive evaluation score.

[0054] Specifically, the to - be - evaluated data acquisition module, the data processing module, and the cognitive improvement evaluation module are electrically connected or communicatively connected in sequence.

[0055] In this embodiment, the music execution record data is historical treatment data that has not been scored and annotated and is obtained from a certain music therapy database. Among them, the acquisition devices for multi - channel electroencephalogram (EEG) data before and after the implementation of the music plan both use a portable EEG 32 - lead detection unit, which is adapted to rubber - ring ear electrodes and supports wireless ERP amplifier and trigger synchronization. Before using the acquisition device, clean the contact points of the rubber - ring ear electrodes with clean water or physiological saline to remove oxides and dirt. During wearing, through the signal quality monitoring function built in the EEG detection unit, a warning is issued when the contact is poor to remind the user to adjust the contact electrodes. Before signal acquisition, the sampling frequency of 128Hz / 256Hz can be set to collect EEG signals.

[0056] In a possible embodiment, the to - be - evaluated data acquisition module also acquires the basic information of the to - be - evaluated user, including age, gender, education level, past medical history, music hobbies, and Alzheimer's disease case reports, etc.

[0057] Exemplarily, the data processing module is configured to: before performing time - domain feature extraction, frequency - domain feature extraction, and non - linear dynamics feature extraction on the multi - channel EEG data, pre - process the multi - channel EEG data first.

[0058] The pre - processing includes: electrode positioning, re - referencing, band - pass filtering, principal component analysis, and interference rejection.

[0059] Specifically, in this embodiment, electrode positioning, re - referencing, band - pass filtering, principal component analysis, and interference rejection are set as programs and integrated in the data processing module, and can automatically perform pre - processing after receiving multi - channel EEG data. Among them, interference rejection includes removing zero - drift, electromyogram (EMG), and electro - oculogram (EOG) interference.

[0060] Exemplarily, the time - domain feature extraction includes: extracting the mean absolute value, zero - crossing number, slope sign change, and waveform length.

[0061] The frequency - domain feature extraction includes: extracting the power spectral density.

[0062] The extraction of the non-linear dynamics features includes: performing multi-dimensional phase space reconstruction, and extracting the correlation dimension, Lyapunov exponent, sample entropy, and approximate entropy.

[0063] Specifically, in the time domain feature extraction, the mean absolute value (MAV) can be calculated by the following formula:

[0064]

[0065] where n is the total number of data points of the multi-channel EEG data collected, and x i is the i-th data of the time series of the multi-channel EEG data, which can quantify the average level of the data.

[0066] The zero crossing number (ZC) refers to the number of times the signal waveform crosses the horizontal reference line within a certain time, and the calculation formula is as follows:

[0067]

[0068] where sgn is the sign function.

[0069] The slope sign change (SSC) refers to the number of times the sign of the amplitude change rate changes in the waveform of the signal, and its calculation formula is:

[0070]

[0071] The waveform length (WL) refers to the cumulative sum of the waveform lengths of the signal within a certain time window, indicating the complexity of the signal, and its calculation formula is:

[0072]

[0073] Specifically, in the frequency domain feature extraction, the power spectral density is calculated by the formula:

[0074]

[0075] where k = 1, 2, 3... N, and N is the number of data points of the selected finite sequence.

[0076] Specifically, the extraction of non-linear dynamics features is effective for EEG cognitive function analysis, and the calculation process is divided into two steps: the first step is multi-dimensional phase space reconstruction. For the time series x(n), the reconstructed vector X i is:

[0077] X i = [x(i), x(i + τ),..., x(i + (m - 1)τ)], i = 1, 2,..., M.

[0078] Among them, m is the embedding dimension, τ is the time delay, and M is the number of vectors in the reconstructed phase space, where M = N - (m - 1)τ.

[0079] The second step is to extract the correlation dimension, Lyapunov exponent, sample entropy, and approximate entropy.

[0080] The calculation formula for the correlation dimension D2 is as follows:

[0081] Among them

[0082] where r represents any given positive value, and θ is the Heaviside step function.

[0083] Assume that in a nonlinear system, x n+1 = F(x n ). The discrete exponent is the Lyapunov exponent, which satisfies the condition under small fluctuations ε:

[0084]

[0085] When ε → 0 and n → ∞, the Lyapunov exponent is:

[0086]

[0087] In this embodiment, the largest Lyapunov exponent is extracted.

[0088] Sample entropy can quantitatively evaluate the complexity of electroencephalogram signal changes, effectively reflecting the characteristics of brain changes under physiological and pathological conditions and the characteristics of the effects of different drugs. In the reconstructed m-dimensional phase space, the distance between two vectors is defined as the maximum value of the differences between their corresponding components, that is, d[X(i), X(j)]. For a given positive value r, calculate the number of vector distances less than r, calculate the ratio of their count to the total number, and then calculate C m (r) as follows:

[0089]

[0090] Continue to calculate C m+1 (r) in m + 1 dimensions. The sample entropy SampEn is as follows:

[0091]

[0092] The calculation formula for the approximate entropy ApEn is as follows:

[0093]

[0094] ApEn(m,r,N) = Φ m (r) - Φ m+1 (r).

[0095] Exemplarily, the data processing module is configured to: after performing time-domain feature extraction, frequency-domain feature extraction, and non-linear dynamics feature extraction on the multi-channel EEG data, perform normalization processing on the multi-scale feature parameters.

[0096] Exemplarily, the machine learning regression model adopts a convolutional neural network and a bidirectional long short-term memory network model, introduces an attention mechanism, and establishes a nine-layer neural network, including: a first input layer, a second convolutional layer, a third pooling layer, a fourth BiLSTM layer, a fifth BiLSTM layer, a sixth self-attention layer, a seventh dropout layer, an eighth fully connected layer, and a ninth regression layer.

[0097] The machine learning regression model has been pre-trained using an EEG data training set, and each piece of data in the EEG data training set includes a multi-scale feature training parameter and a corresponding label score.

[0098] Specifically, in this embodiment, the first input layer is used to input multi-scale feature parameters; the second convolutional layer is used to perform convolution on the time-series data of the multi-channel EEG data using a CNN sequence folding layer, extract features using a 2D convolutional layer with 16 filters, and select ReLU as the activation function; the third pooling layer is a max pooling layer, which divides the input data into rectangular pooling regions and outputs the maximum value in each region, and then unfolds the folded data and inputs it into the fourth BiLSTM layer and the fifth BiLSTM layer. The number of neurons in the fourth BiLSTM layer and the fifth BiLSTM layer is 30 each; the parameters of the sixth self-attention layer are set to a single head and two query-key bonding channels; the dropout probability of the seventh dropout layer is set to 0.2 - 0.4; the eighth fully connected layer and the ninth regression layer are used to output the pre-execution cognitive assessment score or the post-execution cognitive assessment score.

[0099] Specifically, the process of obtaining the multi-scale feature training parameters in the EEG data training set is as follows: Obtain historical treatment data (music execution record data) from a certain music therapy database, and perform score annotation on each multi-channel EEG data in these historical treatment data by means of expert evaluation or extracting the scores of the Mini-Mental State Examination (MMSE), emotion measurement scale, life ability measurement scale, and Pittsburgh Sleep Quality Index (PSQI) in the medical record report to obtain label scores. At the same time, extract multi-scale feature parameters for each multi-channel EEG data in these historical treatment data to form a set of multi-scale feature parameters with label scores, that is, the EEG data training set.

[0100] Such as Figure 2As shown, exemplarily, the Alzheimer's disease cognitive improvement evaluation system based on music further includes: an improvement evaluation database module, a to-be-executed data acquisition module, a case matching module, and a music plan generation module.

[0101] The improvement evaluation database module is used to generate an improvement evaluation data set according to the music execution record data and the corresponding cognitive improvement evaluation results. Each piece of data in the improvement evaluation data set includes a piece of music execution record data and the corresponding cognitive improvement evaluation result.

[0102] The to-be-executed data acquisition module is used to acquire the initial EEG data of the user to be executed.

[0103] The case matching module is used to match several pieces of music execution record data and the corresponding cognitive improvement evaluation results from the improvement evaluation data set according to the initial EEG data as the matching result.

[0104] The music plan generation module is used to select the music plan in the music execution record data corresponding to the optimal cognitive improvement evaluation result from the matching result as the reference music plan.

[0105] Specifically, the cognitive improvement evaluation module is electrically connected or communicatively connected to the improvement evaluation database module. The improvement evaluation database module and the to-be-executed data acquisition module are both electrically connected or communicatively connected to the case matching module. The case matching module is electrically connected or communicatively connected to the music plan generation module.

[0106] In this embodiment, the initial EEG data is the EEG data of a case that has not undergone music therapy obtained from a certain case database. The initial EEG data is also multi-channel EEG data, and the acquisition device of the initial EEG data is the same as the acquisition device of the multi-channel EEG data before and after the execution of the music plan.

[0107] In a possible embodiment, the to-be-executed data acquisition module further acquires the basic information of the user to be executed, including content such as age, gender, education level, past medical history, music hobbies, and Alzheimer's disease case reports.

[0108] Exemplarily, the case matching module uses the random forest algorithm to match the multi-channel EEG data before the execution of the music plan in all the music execution record data in the improvement evaluation data set according to the initial EEG data.

[0109] In this embodiment, each piece of music execution record data in the improvement evaluation dataset is classified according to its pre-execution cognitive evaluation score. Taking the pre-execution cognitive evaluation score ranging from 0 to 100 as an example, the music execution record data corresponding to the pre-execution cognitive evaluation scores in the ranges of [0, 40), [40, 70), and [70, 100] are classified as severe cognitive impairment, moderate cognitive impairment, and mild cognitive impairment in sequence. The random forest algorithm is used to match the multi-channel EEG data corresponding to the music execution record data of several music programs close to the initial EEG data in the improvement evaluation dataset, and the music program in the music execution record data corresponding to the optimal cognitive improvement evaluation result is selected from the several pieces of music execution record data as the reference music program.

[0110] Exemplarily, the music program is divided into three modes, including: passive music program mode, active music program mode, and re-creation music program mode.

[0111] The passive music program mode adopts the mode of playing music, the active music program mode adopts the mode of singing music, and the re-creation music program mode adopts the mode of music creation.

[0112] Specifically, in a possible embodiment, the music program is executed by a music execution module. The music execution module includes: a player, a microphone, and musical instruments. The function of playing music can be realized through the player (corresponding to the passive music program mode), the functions of singing songs, dramas and other music can be realized through the microphone (corresponding to the active music program mode), and the function of music creation can be realized through musical instruments (such as African drums, cymbals, triangles, flutes, etc.) (corresponding to the re-creation music program mode), which helps to increase the relevant activities of the cerebral cortex of Alzheimer's patients and regulate emotions.

[0113] In a possible embodiment, feature extraction and cognitive improvement evaluation can also be performed on the multi-channel EEG data before and after the execution of the reference music program to obtain the corresponding cognitive improvement evaluation result, so as to provide more references. For example, when the cognitive improvement evaluation result is improvement (that is, the cognitive evaluation score after the execution of the reference music program is better than the cognitive evaluation score before the execution of the reference music program), some more challenging music activities can be added (the challenges of the passive music program mode, the active music program mode, and the re-creation music program mode increase in sequence). When the cognitive improvement evaluation result is no improvement (that is, the cognitive evaluation score after the execution of the reference music program is equal to or worse than the cognitive evaluation score before the execution of the reference music program), the reference music program or a music program with lower challenge is still adopted, and then the music type, melody, rhythm, time, and frequency are adjusted according to the user's condition and personalized needs.

[0114] Such as Figure 3As shown in the figure, the embodiment of the present application also provides a method for evaluating the cognitive improvement of Alzheimer's disease based on music, including the following steps:

[0115] Step 1: Obtain the music execution record data of the user to be evaluated, including the music plan and multi-channel EEG data before and after the execution of the music plan.

[0116] Step 2: Extract time-domain features, frequency-domain features, and non-linear dynamics features from the multi-channel EEG data to obtain multi-scale feature parameters before and after the execution of the music plan.

[0117] Step 3: Use a machine learning regression model to obtain the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score respectively based on the multi-scale feature parameters before and after the execution of the music plan.

[0118] Step 4: Obtain the cognitive improvement evaluation result corresponding to the music execution record data according to the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score.

[0119] As Figure 4 shown, exemplarily, after Step 4, it further includes:

[0120] Step 5: Generate an improvement evaluation data set according to the music execution record data and the corresponding cognitive improvement evaluation result. Each data in the improvement evaluation data set includes a piece of music execution record data and the corresponding cognitive improvement evaluation result.

[0121] Step 6: Obtain the initial EEG data of the user to be executed.

[0122] Step 7: Match several pieces of music execution record data and the corresponding cognitive improvement evaluation results from the improvement evaluation data set according to the initial EEG data as the matching result.

[0123] Step 8: Select the music plan in the music execution record data corresponding to the optimal cognitive improvement evaluation result from the matching result as the reference music plan.

[0124] By extracting time-domain features, frequency-domain features, and non-linear dynamics features from the multi-channel EEG data, the embodiment of the present application obtains multi-scale feature parameters before and after the execution of the music plan, uses a machine learning regression model to obtain the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score, and then obtains the cognitive improvement evaluation result corresponding to the music execution record data, improving the accuracy of the evaluation of the cognitive improvement of Alzheimer's disease based on music.

[0125] The proposed time-domain feature extraction includes: extracting the average absolute value, the number of zero crossings, the slope sign change, and the waveform length. The frequency-domain feature extraction includes: extracting the power spectral density. The non-linear dynamics feature extraction includes: performing multi-dimensional phase space reconstruction and extracting the correlation dimension, Lyapunov exponent, sample entropy, and approximate entropy, which improves the comprehensiveness of electroencephalogram data feature extraction.

[0126] The proposed machine learning regression model uses a convolutional neural network and a bidirectional long short-term memory network model, introduces an attention mechanism, and establishes a nine-layer neural network. By combining the convolutional neural network, bidirectional long short-term memory network, and attention mechanism, the accuracy of cognitive improvement assessment is further improved.

[0127] The proposed case matching module uses the random forest algorithm to match the multi-channel electroencephalogram data before the execution of the music plan in all music execution record data in the improvement assessment dataset according to the initial electroencephalogram data, which improves the accuracy of case matching and further improves the reliability of the reference music plan.

[0128] The proposed passive music plan mode, active music plan mode, and re-creation music plan mode improve the diversity of music plans.

[0129] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be interpreted to include the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0130] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

1. Alzheimer's disease cognitive improvement assessment system based on music, characterized in that, Including: A module for obtaining data to be evaluated, a data processing module, and a cognitive improvement evaluation module; The module for obtaining data to be evaluated is used to obtain the music execution record data of the user to be evaluated, including music programs and multi-channel EEG data before and after the execution of the music program; The data processing module is used to extract time-domain features, frequency-domain features, and non-linear dynamics features from the multi-channel EEG data to obtain multi-scale feature parameters before and after the execution of the music program; The cognitive improvement evaluation module is used to obtain the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score respectively according to the multi-scale feature parameters before and after the execution of the music program by using a machine learning regression model; The cognitive improvement evaluation module is also used to obtain the cognitive improvement evaluation result corresponding to the music execution record data according to the pre-execution cognitive evaluation score and the post-execution cognitive evaluation score.

2. The music-based Alzheimer's disease cognitive improvement assessment system according to claim 1, wherein The data processing module is set to: before extracting time-domain features, frequency-domain features, and non-linear dynamics features from the multi-channel EEG data, preprocess the multi-channel EEG data first; The preprocessing includes: electrode positioning, re-referencing, band-pass filtering, principal component analysis, and interference removal.

3. The music-based Alzheimer's disease cognitive improvement assessment system according to claim 1, wherein The time-domain feature extraction includes: extracting the mean absolute value, zero crossing number, slope sign change, and waveform length; The frequency-domain feature extraction includes: extracting the power spectral density; The non-linear dynamics feature extraction includes: performing multi-dimensional phase space reconstruction and extracting the correlation dimension, Lyapunov exponent, sample entropy, and approximate entropy.

4. The music-based Alzheimer's disease cognitive improvement assessment system according to claim 1, wherein The data processing module is set to: after extracting time-domain features, frequency-domain features, and non-linear dynamics features from the multi-channel EEG data, standardize the multi-scale feature parameters.

5. The music-based Alzheimer's disease cognitive improvement evaluation system according to claim 1, characterized in that The machine learning regression model uses a convolutional neural network and a bidirectional long short-term memory network model, introduces an attention mechanism, and establishes a nine-layer neural network, including: a first input layer, a second convolutional layer, a third pooling layer, a fourth BiLSTM layer, a fifth BiLSTM layer, a sixth self-attention layer, a seventh dropout layer, an eighth fully connected layer, and a ninth regression layer; The machine learning regression model has been pre-trained using an EEG data training set, and each piece of data in the EEG data training set includes a multi-scale feature training parameter and a corresponding label score.

6. The music-based Alzheimer's disease cognitive improvement evaluation system according to claim 1, characterized in that Also including: An improvement evaluation database module, a module for obtaining data to be executed, a case matching module, and a music program generation module; The improvement evaluation database module is used to generate an improvement evaluation data set according to the music execution record data and the corresponding cognitive improvement evaluation result, and each piece of data in the improvement evaluation data set includes a music execution record data and the corresponding cognitive improvement evaluation result; The module for obtaining data to be executed is used to obtain the initial EEG data of the user to be executed; The case matching module is used to match several music execution record data and the corresponding cognitive improvement evaluation results from the improvement evaluation data set according to the initial EEG data as the matching result; The music scheme generation module is used to select the music scheme in the music execution record data corresponding to the optimal cognitive improvement evaluation result from the matching result as the reference music scheme.

7. The music-based Alzheimer's disease cognitive improvement assessment system according to claim 6, characterized in that, The case matching module uses the random forest algorithm to match the multi-channel EEG data before the execution of the music scheme in all the music execution record data in the improvement evaluation dataset according to the initial EEG data.

8. The music-based Alzheimer's disease cognitive improvement assessment system according to claim 1, wherein The music scheme is divided into three modes, including: passive music scheme mode, active music scheme mode, and re-creation music scheme mode. The passive music scheme mode adopts the mode of playing music, the active music scheme mode adopts the mode of singing music, and the re-creation music scheme mode adopts the mode of music creation.

9. The method for evaluating the improvement of Alzheimer's disease cognition based on music, which uses the system for evaluating the improvement of Alzheimer's disease cognition based on music according to any one of claims 1 to 8, is characterized in that, It includes the following steps: Step 1, obtain the music execution record data of the user to be evaluated, including the music scheme and the multi-channel EEG data before and after the execution of the music scheme. Step 2, perform time-domain feature extraction, frequency-domain feature extraction, and non-linear dynamics feature extraction on the multi-channel EEG data to obtain the multi-scale feature parameters before and after the execution of the music scheme. Step 3, use the machine learning regression model to obtain the cognitive evaluation score before execution and the cognitive evaluation score after execution respectively according to the multi-scale feature parameters before and after the execution of the music scheme. Step 4, obtain the cognitive improvement evaluation result corresponding to the music execution record data according to the cognitive evaluation score before execution and the cognitive evaluation score after execution.

10. The method for evaluating the improvement of Alzheimer's disease cognition based on music according to claim 9, wherein, After step 4, it further includes: Step 5, generate an improvement evaluation dataset according to the music execution record data and the corresponding cognitive improvement evaluation result. Each piece of data in the improvement evaluation dataset includes a piece of music execution record data and the corresponding cognitive improvement evaluation result. Step 6, obtain the initial EEG data of the user to be executed. Step 7, match several pieces of music execution record data and the corresponding cognitive improvement evaluation results from the improvement evaluation dataset according to the initial EEG data as the matching result. Step 8, select the music scheme in the music execution record data corresponding to the optimal cognitive improvement evaluation result from the matching result as the reference music scheme.

Images