Alzheimer's disease early screening method and system based on electroencephalogram signals

Abstract

This application provides a method and system for early screening of Alzheimer's disease based on electroencephalogram (EEG) signals, comprising: acquiring the resting-state EEG signals of the subject as raw EEG signals, and performing data preprocessing on the raw EEG signals; extracting the average power value of the preprocessed signals based on the multi-band Morlet wavelet transform algorithm, constructing a functional connectivity matrix based on the extracted average power value and performing multi-dimensional feature fusion, and using the resulting multi-dimensional functional connectivity tensor as the core feature; inputting the multi-dimensional functional connectivity tensor into a pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features, and using the network output as the early screening result for Alzheimer's disease. This application covers three core links: automatic preprocessing, data analysis, and classification detection, which work together to achieve efficient and convenient early screening for Alzheimer's disease.

Description

Technical Field

[0001] This application belongs to the field of biomedical engineering technology, and in particular relates to a method and system for early screening of Alzheimer's disease based on electroencephalogram (EEG) signals. Background Technology

[0002] Alzheimer's disease (AD) is a progressive neurodegenerative disease with an insidious onset. Its core clinical feature is memory impairment, often accompanied by aphasia, visuospatial skill impairment, executive dysfunction, and ultimately, generalized dementia. With the increasing aging of the global population, the incidence of AD is rising annually, placing a heavy burden on patients' families and society. Early diagnosis and intervention are crucial to slowing the progression of Alzheimer's disease; however, existing diagnostic methods still have many shortcomings in terms of sensitivity, specificity, and accessibility, necessitating the development of more efficient and non-invasive early diagnostic technologies.

[0003] Currently, commonly used clinical diagnostic methods for Alzheimer's disease (AD) mainly include neuropsychological scale assessment, imaging techniques, and biomarker detection.

[0004] Among them, neuropsychological scale assessments (such as the Mini-Mental State Examination) serve as a basic tool for preliminary screening and cognitive status assessment. Although they are relatively simple to operate, low in cost, and easy to promote, they also have significant limitations: their results are affected by a variety of factors such as the tester's experience, the subject's education level, cooperation, emotional state, and testing environment, making them highly subjective; they lack sensitivity to early or mild cognitive impairment, making it difficult to accurately distinguish AD from other types of dementia; and they mainly reflect the current state of cognitive function, failing to directly reveal potential pathophysiological changes.

[0005] While imaging technologies such as magnetic resonance imaging and positron emission tomography (PET) are highly accurate in showing changes in brain structure (such as hippocampal atrophy) or pathological protein deposits (such as Aβ and Tau), they suffer from problems such as expensive equipment, complex testing procedures, and difficulty in widespread adoption.

[0006] Cerebrospinal fluid biomarker testing (such as Aβ42, t-Tau, p-Tau) also has limitations such as being invasive, costly, and technically demanding. Procedures such as lumbar puncture may cause complications such as infection and headache, resulting in low patient acceptance; the testing itself is expensive and requires strict techniques for sample processing, storage, and laboratory analysis; and the time from sample collection to obtaining results often takes several days or even weeks, making it difficult to meet the needs of rapid diagnosis. Summary of the Invention

[0007] In view of this, this application aims to propose a method and system for early screening of Alzheimer's disease based on electroencephalogram (EEG) signals, in order to solve at least one of the above-mentioned problems.

[0008] To achieve the above objectives, the technical solution of this application is implemented as follows: In a first aspect, this application provides a method for early screening of Alzheimer's disease based on electroencephalogram (EEG) signals, including: The subject's resting-state EEG signal was acquired as the raw EEG signal, and the raw EEG signal was preprocessed. Based on the preprocessed EEG signal, the average power value of the signal is extracted using the multi-band Morlet wavelet transform algorithm. A functional connectivity matrix is ​​constructed based on the extracted average power value, and multi-dimensional feature fusion is performed. The resulting multi-dimensional functional connectivity tensor is used as the core feature. The generated multidimensional functional connection tensor is input into a pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features. The output of the deep separable residual convolutional network is used as the early screening result for Alzheimer's disease.

[0009] Secondly, based on the same inventive concept, this application also provides an early screening system for Alzheimer's disease based on electroencephalogram (EEG) signals, comprising: The data preprocessing module is configured to acquire the subject's resting-state EEG signal as the raw EEG signal and perform data preprocessing on the raw EEG signal; The data analysis module is configured to extract the average power value of the preprocessed EEG signal based on the multi-band Morlet wavelet transform algorithm, construct a functional connectivity matrix based on the extracted average power value, perform multi-dimensional feature fusion, and use the resulting multi-dimensional functional connectivity tensor as the core feature. The classification and detection module is configured to input the generated multidimensional functional connection tensor into a pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features, and use the output of the deep separable residual convolutional network as the early screening result for Alzheimer's disease.

[0010] Thirdly, based on the same inventive concept, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect.

[0011] Fourthly, based on the same inventive concept, this application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method as described in the first aspect.

[0012] Compared with existing technologies, the early screening method and system for Alzheimer's disease based on electroencephalogram (EEG) signals described in this application have the following beneficial effects: The method for early screening of Alzheimer's disease based on EEG signals described in this application constructs a cross-band, cross-brain region functional connectivity matrix through multi-band Morlet wavelet transform and mutual information calculation. This matrix can effectively capture nonlinear, high-resolution coupling patterns in the global dynamics of the brain, providing a compact and comprehensive representation of the whole-brain functional network. Furthermore, a deep separable residual convolutional network is used to learn and classify the features of the constructed multidimensional functional connectivity tensor, which significantly reduces the number of parameters while alleviating the gradient vanishing problem, improving the model's generalization ability, and supporting deployment and application in mobile computing environments. This method can collaboratively achieve early auxiliary diagnosis of Alzheimer's disease. Attached Figure Description

[0013] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of an early screening method for Alzheimer's disease based on electroencephalogram (EEG) signals, as described in an embodiment of this application. Figure 2 This application presents a 3D MMMIFC matrix for AD patients across frequencies (1-19 Hz) and a 2D functional connectivity matrix diagram in four representative frequency bands (1-2 Hz), (5-6 Hz), (9-10 Hz), and (13-14 Hz). Figure 3 The diagram shows the 3D MMMIFC matrix of healthy subjects across frequencies (1-19 Hz) as described in the embodiments of this application, and the 2D functional connectivity matrix diagram in four representative frequency bands (1-2 Hz), (5-6 Hz), (9-10 Hz) and (13-14 Hz). Figure 4 This is a schematic diagram of the depthwise separable residual convolutional network structure described in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of an early Alzheimer's disease screening system based on electroencephalogram (EEG) signals, as described in an embodiment of this application. Figure 6 This is a schematic diagram of the hardware structure of the electronic device described in an embodiment of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0015] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0016] The embodiments of this application are described in detail below with reference to the accompanying drawings.

[0017] Please participate Figure 1 As shown, this embodiment provides a method for early screening of Alzheimer's disease based on electroencephalogram (EEG) signals, which specifically includes the following steps: Step S101: Obtain the subject's resting-state EEG signal as the raw EEG signal, and perform data preprocessing on the raw EEG signal.

[0018] Specifically, in this embodiment, the resting-state EEG signal of the subject is collected as the raw EEG signal, and the raw EEG signal is downsampled, bandpass filtered, notch filtered and independent component analysis is performed to remove artifacts, and bad segments are removed by target artifact space reconstruction algorithm, and normalization and sliding window segmentation are performed.

[0019] Step S102: Based on the preprocessed EEG signal, extract the average power value of the signal using the multi-band Morlet wavelet transform algorithm, construct the functional connectivity matrix based on the extracted average power value, and perform multi-dimensional feature fusion, using the resulting multi-dimensional functional connectivity tensor as the core feature.

[0020] Specifically, in this embodiment, for each 5-second resting-state EEG signal segment, a complex Morlet wavelet transform is used for time-frequency decomposition, with the frequency analysis range covering 1 to 20 Hz. Nineteen consecutive sub-bands are precisely divided in 1 Hz increments (in order: 1-2 Hz, 2-3 Hz, ..., 19-20 Hz). Then, for each lead electrode and each sub-band, the square of the wavelet coefficient modulus within that band is calculated and averaged to obtain the average power of that electrode in that frequency band.

[0021] Furthermore, for each center frequency A Morlet wavelet was constructed: ; In the formula, the time scale parameter A is the normalization coefficient, ensuring that the wavelet energy is 1. Used to control wavelet width, in practical calculations, to balance computational efficiency and signal fidelity, the effective range of wavelets in the time domain is limited to [specific range]. The signal is truncated within the range to retain most of the energy signal while maintaining computational efficiency.

[0022] The EEG signal from each lead is convolved with the constructed Morlet wavelet to obtain continuous wavelet coefficients: ; In the formula, This represents the EEG signal of one lead, where b is the time shift parameter. Describing wavelets The complex conjugate form, Reflects the channel Signal at time point and frequency The energy distribution in the vicinity.

[0023] To further extract time-frequency power information, the square of the modulus of the wavelet coefficients is taken, and then averaged over the time dimension to obtain the channel. In frequency Nearby average power: .

[0024] For each 1 Hz sub-band, based on the average power value extracted from each channel, the mutual information algorithm is used to calculate the nonlinear functional connectivity strength between each pair of all 19 electrode channels. To eliminate the influence of marginal entropy differences and enhance the comparability between different channel pairs, the mutual information values ​​are normalized using the self-information of each channel, and the resulting asymmetric matrix is ​​symmetricized to form a symmetric functional connectivity matrix. Finally, the functional connectivity matrices corresponding to the 19 sub-bands are stacked along the frequency dimension to construct a three-dimensional functional connectivity tensor. It includes two spatial dimensions and one frequency dimension. The two spatial dimensions define the functional connection strength between the electrode pairs, and the frequency dimension characterizes the distribution of the functional connection strength with frequency.

[0025] Furthermore, for each frequency band, the mutual information between all channel pairs is calculated based on its average power value: ; In the formula, and Indicates channel and The marginal probability distribution of the power value, Describe the joint probability distribution To account for varying marginal entropy and enhance comparability between channel pairs, mutual information values ​​are normalized using self-information: ; Finally, the matrix is ​​symmetric to ensure that the function link matrix is ​​a real symmetric matrix: ; Repeat the above calculation for all 19 frequency sub-bands to obtain a total of 19 frequency band functional connectivity matrices, which are then stacked to form the final tensor. .

[0026] Repeat the above calculation process for all 19 frequency sub-bands, generating one for each frequency band. Functional connection matrix ( (Number of channels) Figure 2 and Figure 3 This study presents 3D MMMIFC matrices across frequencies (1-19 Hz) for AD patients and healthy subjects, as well as 2D functional connectivity matrix diagrams in four representative frequency bands: (1-2 Hz), (5-6 Hz), (9-10 Hz), and (13-14 Hz). Finally, these matrices are stacked along the frequency dimension to form a three-dimensional tensor. ,in, This represents the number of frequency bands. This tensor serves as the input to subsequent deep separable residual convolutional network models, effectively encoding the functional connectivity patterns of EEG signals across multiple frequency bands.

[0027] Step S103: Input the generated multidimensional functional connection tensor into the pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features, and use the output of the deep separable residual convolutional network as the early screening result of Alzheimer's disease.

[0028] Specifically, in this embodiment, after converting the raw EEG signals into a three-dimensional functional connectivity tensor suitable for input to a deep separable residual convolutional network through the aforementioned processing steps, the same operation is performed on all samples from each subject to construct a uniform dataset. To maintain subject independence and avoid data leakage, a subject-based partitioning strategy is adopted, with 80% of the total samples allocated to the training set and the remaining 20% ​​as the validation set. During training, the batch size is set to 16 to balance memory usage and training stability.

[0029] The data is input into a constructed depthwise separable residual convolutional network for training. For example... Figure 4As shown, the main structure of the network includes: an initial convolutional module, which consists of a 3D depthwise separable convolutional layer, batch normalization, and ReLU activation function; followed by a 3D max pooling layer; four sequentially connected residual stages, each containing two improved residual blocks, for a total of eight residual blocks; and a 3D global average pooling layer and a fully connected layer at the end of the network as a classifier.

[0030] The initial convolutional layer adopts a depthwise separable convolutional structure, which decomposes the standard convolution into two steps: first, a channel-wise 3D depthwise convolution is performed to independently extract the spatial features of each input channel, significantly reducing the number of parameters and computational complexity; then, a 1×1×1 pointwise convolution is used to achieve cross-channel information fusion and feature dimension transformation, which retains effective feature representation while reducing redundant computation.

[0031] The main path of each improved residual block consists of two convolutional layers: the first layer uses a 3×3×3 3D depthwise separable convolution, and the second layer uses a standard 3×3×3 convolution. When the input and output dimensions do not match, the skip connection path performs downsampling and channel number adjustment through a 1×1×1 convolution with batch normalization. The outputs of the main path and skip path are fused element-wise and then processed by a ReLU activation function.

[0032] In the first residual stage, the number of input and output channels for both residual blocks is set to 64. The first layer of the main path within each block uses a depthwise separable convolution with a kernel size of 3×3×3, a stride of 1, and padding of 1 to preserve the spatial dimensions of the feature map; the second layer is a standard convolution with the same parameters. No spatial downsampling or channel number changes are performed in this stage, and skip connections do not require additional operations, focusing on extracting shallow local features.

[0033] The second stage increases the number of channels to 128. In the main path of the first residual block, the depthwise separable convolution stride is set to 2 to halve the feature map size and double the number of channels. The skip path is also aligned using a 1×1×1 convolution with a stride of 2. In the second block of this stage, the stride is restored to 1, maintaining 128 channels for feature transformation.

[0034] The third stage further expands the number of channels to 256, and the fourth stage continues to increase it to 512. The structure of these two stages is consistent with that of the second stage: the first residual block of each stage is downsampled and the channels are expanded, and the second block maintains the current resolution to further extract high-level semantic features.

[0035] Batch normalization and ReLU activation are applied after all convolution operations to promote gradient flow, accelerate convergence, and enhance the model's nonlinear expressive power.

[0036] The network ultimately compresses the spatial features of each channel into a single scalar through a 3D global average pooling layer, generating a 512-dimensional feature vector, which is then mapped to the category output via a fully connected layer. This forward process achieves a hierarchical and efficient transformation from the original 3D functional connections to high-level semantic features, maintaining excellent classification performance while significantly reducing the number of parameters.

[0037] Training was performed using the Adam optimizer, with an initial learning rate of 1×10⁻⁶. -4 Every 10 training epochs, the learning rate decays by a factor of 0.9 to achieve finer parameter tuning and improve convergence stability in the later stages of training. This design comprehensively considers the effectiveness of feature extraction, the efficiency of model parameters, and the controllability of the training process, completing efficient hierarchical representation learning from the original signal to the classification result end-to-end.

[0038] In some implementations, this embodiment also integrates powerful visualization and analysis functions, including an interactive graphical user interface, a core visualization panel, a diagnostic results area, a file operation area, a file information area, and a processing log area.

[0039] The interactive graphical user interface uses the ttkbootstrap library to enhance the interface's aesthetics and modern style, supporting dynamic theme switching while enabling maximum window startup. The core visualization panel includes the raw EEG, preprocessed EEG, 3D functional connectivity matrix, time-frequency plot, and power spectral density plot display areas, all organized using Label Frames and equipped with status prompts. The diagnostic results area visually displays the predicted results (AD / HC), using dual-color (red / green) labels and a progress bar to dynamically display the predicted probability; The file operation area integrates core function buttons (load file, preprocess, data analysis, perform prediction), and is equipped with a progress bar and status labels; The file information area displays information related to the loaded EEG file; the processing log area can output system information and status in real time and supports scrolling viewing.

[0040] This embodiment describes an early Alzheimer's disease screening method based on EEG signals. By using multi-band Morlet wavelet transform and mutual information calculation, it constructs a cross-band, cross-brain region functional connectivity matrix, which can effectively capture nonlinear, high-resolution coupling patterns in the global dynamics of the brain, providing a compact and comprehensive representation of the whole-brain functional network. Furthermore, it employs a deep separable residual convolutional network to perform feature learning and classification on the constructed multidimensional functional connectivity tensor, significantly reducing the number of parameters while alleviating the gradient vanishing problem, improving the model's generalization ability, and supporting deployment and application in mobile computing environments. This method can collaboratively achieve early auxiliary diagnosis of Alzheimer's disease.

[0041] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0042] Based on the same inventive concept, and corresponding to any of the above embodiments, the embodiments of this application also provide an early screening system for Alzheimer's disease based on electroencephalogram (EEG) signals.

[0043] like Figure 5 As shown, the Alzheimer's disease early screening system based on electroencephalogram (EEG) signals includes: The data preprocessing module 11 is configured to acquire the subject's resting-state EEG signal as the raw EEG signal and perform data preprocessing on the raw EEG signal; The data analysis module 12 is configured to extract the average power value of the preprocessed EEG signal based on the multi-band Morlet wavelet transform algorithm, construct a functional connectivity matrix based on the extracted average power value, perform multi-dimensional feature fusion, and use the resulting multi-dimensional functional connectivity tensor as the core feature. The classification and detection module 13 is configured to input the generated multidimensional functional connection tensor into a pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features, and use the output of the deep separable residual convolutional network as the early screening result of Alzheimer's disease.

[0044] For ease of description, the above system is described by dividing it into various modules based on their functions. Of course, in implementing the embodiments of this application, the functions of each module can be implemented in one or more software and / or hardware.

[0045] The system described in the above embodiments is used to implement the corresponding method in any of the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0046] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the methods described in any of the above embodiments.

[0047] Figure 6This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0048] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0049] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0050] The input / output interface 1030 is used to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0051] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0052] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0053] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0054] The electronic devices described above are used to implement the corresponding methods in any of the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0055] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium that stores computer instructions for causing the computer to perform the methods described in any of the above embodiments.

[0056] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0057] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to perform the methods described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0058] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.

[0059] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0060] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

Claims

1. A method for early screening of Alzheimer's disease based on electroencephalogram (EEG) signals, characterized in that, include: The subject's resting-state EEG signal was acquired as the raw EEG signal, and the raw EEG signal was preprocessed. Based on the preprocessed EEG signal, the average power value of the signal is extracted using the multi-band Morlet wavelet transform algorithm. A functional connectivity matrix is ​​constructed based on the extracted average power value, and multi-dimensional feature fusion is performed. The resulting multi-dimensional functional connectivity tensor is used as the core feature. The generated multidimensional functional connection tensor is input into a pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features. The output of the deep separable residual convolutional network is used as the early screening result for Alzheimer's disease.

2. The method according to claim 1, characterized in that, The extraction of the signal average power value based on the multi-band Morlet wavelet transform algorithm includes: By using Morlet wavelet transform to decompose a given length of resting-state EEG signal segment into time-frequency components, and dividing it into several continuous sub-bands with a preset step size; Based on each lead electrode and each sub-band, the square of the wavelet coefficient modulus within that band is calculated and then averaged to obtain the average power value of that electrode in that frequency band.

3. The method according to claim 2, characterized in that, The step of constructing a functional connectivity matrix based on the extracted average power value and performing multi-dimensional feature fusion includes: For each sub-band, based on the average power value extracted from each channel, the nonlinear functional connection strength between each pair of all electrode channels is calculated one by one using the mutual information algorithm; The mutual information values ​​are normalized using the self-information of each channel, and the resulting asymmetric matrix is ​​symmetricized to form a symmetric functional connection matrix. The functional connectivity matrices corresponding to all sub-bands are stacked along the frequency dimension to form a multidimensional functional connectivity tensor.

4. The method according to claim 1, characterized in that: The multidimensional functional connectivity tensor includes two spatial dimensions and one frequency dimension. The two spatial dimensions are used to define the functional connectivity strength between electrode pairs, and the frequency dimension is used to characterize the distribution of functional connectivity strength with frequency.

5. The method according to claim 1, characterized in that: The depthwise separable residual convolutional network includes an initial convolutional layer, a max pooling layer, four residual stages, a global adaptive average pooling layer, and a fully connected classifier; wherein, the initial convolutional layer adopts a 3D depthwise separable convolutional structure.

6. The method according to claim 5, characterized in that: Each residual stage contains two improved residual blocks, and the main path of each residual block consists of two convolutional layers: the first layer uses 3D depthwise separable convolution, and the second layer uses standard 3D convolution. When the input and output dimensions do not match, the skip connection path performs downsampling and channel alignment through a 1×1×1 convolution with batch normalization, and finally fuses the outputs of the main path and the skip path by element-wise addition and ReLU activation function.

7. The method according to claim 1, characterized in that: It also includes a visualization interface, the contents of which include an interactive graphical user interface, a core visualization panel, a diagnostic results area, a file operation area, a file information area, and a processing log area.

8. An early screening system for Alzheimer's disease based on electroencephalogram (EEG) signals, characterized in that, include: The data preprocessing module is configured to acquire the subject's resting-state EEG signal as the raw EEG signal and perform data preprocessing on the raw EEG signal; The data analysis module is configured to extract the average power value of the preprocessed EEG signal based on the multi-band Morlet wavelet transform algorithm, construct a functional connectivity matrix based on the extracted average power value, perform multi-dimensional feature fusion, and use the resulting multi-dimensional functional connectivity tensor as the core feature. The classification and detection module is configured to input the generated multidimensional functional connection tensor into a pre-constructed deep separable residual convolutional network to perform end-to-end feature learning and decision classification on the core features, and use the output of the deep separable residual convolutional network as the early screening result for Alzheimer's disease.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that, in, The non-transitory computer-readable storage medium stores computer instructions for causing a computer to perform the method described in any one of claims 1-7.

Images