MCI diagnosis system based on multi-scale hypergraph network

Abstract

The invention provides an MCI diagnosis system based on a multi-scale hypergraph network, a storage medium and electronic equipment, and relates to the technical field of neural networks. The method comprises the following steps: firstly, preprocessing an fMRI signal of a patient to be diagnosed, and obtaining a BOLD time sequence of each ROI region on a functional image; calculating a phase synchronization relation of the ROI region in a frequency spectrum domain through a phase locking value technology; a graph attention mechanism is utilized to dynamically learn and identify relative importance among different brain regions in a time domain; designing a multi-scale feature fusion mechanism to integrate time domain and frequency spectrum dynamic connection features of the fMRI signal to form a time frequency fusion matrix, and establishing a dynamic hypergraph model based on the matrix; and finally, constructing a detection model for mild cognitive impairment by using a hypergraph neural network, and performing parameter optimization through a grid search strategy to improve the MCI diagnosis effect.

Description

Technical Field

[0001] The present invention relates to the field of neural network technology, and in particular to an MCI diagnosis system, storage medium and electronic equipment based on a multi-scale hypergraph network. Background Art

[0002] Mild cognitive impairment (MCI) is a clinically diagnosed syndrome characterized by objective cognitive decline in one or more cognitive domains. MCI is considered a transitional state between normal aging and dementia, and early diagnosis and intervention are essential.

[0003] Functional Magnetic Resonance Imaging (fMRI), as a non-invasive imaging technique, provides a valuable tool for exploring potential imaging biomarkers of MCI-related neurological complications by measuring the blood oxygen level dependent (BOLD) signal. fMRI technology has been widely used to reveal the clinical phenotype of MCI and provides the possibility for early diagnosis of the disease.

[0004] Although fMRI technology has shown potential in the diagnosis of MCI, the fMRI-based brain network identification and detection of MCI through hypergraph modeling still faces several challenges:

[0005] (1) Limitations of vector distance: Traditional vector distance metrics rely too much on inferred interactions in geometric space, which may not fully capture the dynamic connectivity patterns and complex interactions between brain regions. This dependence limits our ability to understand and analyze the complexity of brain networks.

[0006] (2) Challenges of multi-scale feature fusion: In the process of multi-scale feature fusion, how to effectively align and optimize features of different scales is a technical challenge. This directly affects the hypergraph model’s ability to learn and represent brain network features. Summary of the invention

[0007] 1. Technical issues to be solved

[0008] In view of the deficiencies in the prior art, the present invention provides an MCI diagnosis system, storage medium and electronic device based on a multi-scale hypergraph network, which solves the technical problem of early and accurate diagnosis of MCI.

[0009] (II) Technical solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0011] A MCI diagnosis system based on a multi-scale hypergraph network, comprising:

[0012] A preprocessing module is used to acquire and preprocess the fMRI signals of the patient to be diagnosed, so as to obtain the BOLD time series of each ROI area on the functional image;

[0013] The synchronization analysis module is used to evaluate the phase synchronization between paired ROI time signals using phase-locked value technology and obtain the second-order spectrum matrix;

[0014] The adaptive module is used to calculate the weight coefficients between paired ROI time signals using the graph attention mechanism and obtain the time domain matrix after smoothing;

[0015] A fusion module, used for fusing the second-order spectrum matrix and the time domain matrix to obtain a time-frequency fusion matrix;

[0016] An extraction module, used to use the time-frequency fusion matrix as an input of a hypergraph neural network to obtain a feature representation of a high-order relationship between different ROI regions;

[0017] The classification module is used to predict the MCI diagnosis result of the patient to be diagnosed based on the final output of the hypergraph neural network.

[0018] Preferably, the preprocessing module uses the DPARSF software package to preprocess the fMRI signal, specifically including:

[0019] The original functional images of the first 10 time points of the fMRI signal were removed and slice timing correction was performed; the original functional images with head movement exceeding 2.5 mm or 2.5 degrees were excluded and motion correction was performed; statistical parameter mapping was used to co-register with T1-weighted images, and gray matter, white matter and cerebrospinal fluid were segmented; DARTEL was used to map the functional images to MNI space; a half-maximum Gaussian kernel was used to smooth the functional images by 4 mm full width to reduce registration variability; the functional image sequence was bandpass filtered between 0.01 and 0.1 Hz;

[0020] The acquired BOLD time series is expressed as:

[0021]

[0022] Among them, x i represents the BOLD time series of the i-th ROI region as the ROI time signal; represents a real number set; N represents the number of ROI areas; F de Indicates the length of the signal.

[0023] Preferably, the synchronization analysis module is specifically used for:

[0024] The instantaneous phase of a signal is estimated using the Hilbert transform as follows:

[0025]

[0026] in, represents the Hilbert transform of the signal x(t); PV represents the Cauchy principal value; x(t′) represents the value of the original signal at different times during the integration process; t′ represents the integration variable;

[0027] The analytic signal phase is defined as:

[0028]

[0029] Where φ(t) represents the instantaneous phase of the signal; arctan is the inverse tangent function;

[0030] Calculate the paired ROI time signal x i and x j The PLV values ​​between are as follows:

[0031]

[0032] Where l∈{1,2,...,T} represents a time period in T, and T is the total number of time periods; represents an imaginary unit; Δφ ij (t l )=(φ i (t l )-φ j (t l )) represents the instantaneous phase difference of the ROI time signal, respectively from the ROI time signal x i The data x of the lth period i (t l ) and ROI time signal x j The data x of the lth period j (t l ), calculate φ i (t l ) and φ j (t l ), use Hilbert transform to get the phase value;

[0033] Using ROI time signal x i and x j The synchronous correlation between them forms the second-order spectrum matrix The definition is as follows:

[0034]

[0035] Preferably, the adaptive module is specifically used for:

[0036] Calculate ROI time signal x using graph attention mechanism i and x j The weight coefficient e ij Connection strength as time domain:

[0037]

[0038] Among them, exp is the exponential function; LeakyReLU is the activation function; W e are parameterized vectors and learnable weight matrices respectively; || is a connection operation; if is greater than the preset threshold τ, Represents x i The directly related node set, k is the node index in the node set;

[0039] use Defining the attention matrix Represents the correlation of all ROI time signals in the time domain and minimizes the optimization function Get the best combination of learnable parameters:

[0040]

[0041] Among them, λ≥0 represents the regularization parameter;

[0042] Define the all-ones matrix Smoothed attention matrix M e :

[0043] C TEM =M e +M ones (8)

[0044] Among them, C TEM represents the time domain matrix after smoothing.

[0045] Preferably, the fusion module is specifically used for:

[0046] C TS =C TEM ⊙C SPE (9)

[0047] Among them, C TS represents the time-frequency fusion matrix; C TEM , C SPE represent the time domain matrix and the second-order spectrum matrix respectively; ⊙ represents the Hadamard product.

[0048] Preferably, the extraction module is specifically used for:

[0049] Define a hypergraph using Indicates that represents a node set, n is the total number of nodes; ε={e1,e2,...,e m} represents the hyperedge set, m is the total number of hyperedges; is the weight matrix of the hyperedge; each hyperedge e connects multiple nodes v and is represented by The diagonal matrix weights of ;

[0050] Hypergraph Expressed as The incidence matrix H is defined as follows:

[0051]

[0052] Based on the time-frequency fusion matrix C TS Construct a hyperedge. For the i-th ROI area, in C TS Select the associated regions K with the highest connectivity value from the top_K i ;

[0053] Based on the set top_K i , redefine the incidence matrix H as:

[0054]

[0055] Among them, j represents the set top_K i The index of the associated region in ;

[0056] For the incidence matrix H, the degree matrix of its vertices is defined as use Indicates that diag is a diagonal matrix, d(v i ) represents node v i The degree of ; and the degree matrix of its hyperedge is expressed as in δ(e i ) represents the hyperedge e i degree;

[0057] The hypergraph neural network is used to capture the high-order relationship between different ROI regions for feature representation, wherein the first The convolutional layer is represented as:

[0058]

[0059] The superscript T indicates transposition; X (l) Indicates The convolution output features of the layer, As

[0060] As a filter to extract node features in the hypergraph, represents the learnable hyperedge weight matrix, W(e i ) represents the hyperedge e i The weight coefficient of .

[0061] Preferably, the classification module trains a multilayer perceptron to predict the MCI diagnosis result of the patient to be diagnosed, which is expressed as:

[0062]

[0063] in, represents the MCI diagnosis result of the patient to be diagnosed; MLP represents the multi-layer perceptron; Z represents the final output of the hypergraph neural network.

[0064] Preferably, the hypergraph neural network uses a cross entropy function to estimate the prediction loss during the training phase, expressed as:

[0065]

[0066] Where s represents the index of the subject sample; S is the total number of subject samples; y s , They represent the true label and predicted label of the sth subject sample respectively; log is the logarithmic function.

[0067] A storage medium stores a computer program for MCI diagnosis based on a multi-scale hypergraph network, wherein the computer program enables a computer to control the MCI diagnosis system as described above to perform MCI diagnosis.

[0068] An electronic device, comprising:

[0069] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, so that the one or more processors control the MCI diagnostic system as described above to perform MCI diagnosis.

[0070] (III) Beneficial effects

[0071] The present invention provides an MCI diagnosis system, storage medium and electronic device based on a multi-scale hypergraph network. Compared with the prior art, it has the following beneficial effects:

[0072] In the present invention, the fMRI signals of the patients to be diagnosed are firstly preprocessed to obtain the BOLD time series of each ROI region on the functional image; then the phase synchronization relationship of the ROI region in the spectral domain is calculated by the phase-locked value technology; and the graph attention mechanism is used to dynamically learn and identify the relative importance of different brain regions in the time domain; then a multi-scale feature fusion mechanism is designed to integrate the time domain and spectral dynamic connection features of the fMRI signals to form a time-frequency fusion matrix, and a dynamic hypergraph model is established based on the matrix; finally, a hypergraph neural network is used to construct a detection model for mild cognitive impairment, and the parameters are optimized through a grid search strategy to improve the MCI diagnostic effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0074] Figure 1 A structural block diagram of an MCI diagnosis system based on a multi-scale hypergraph network provided in an embodiment of the present invention;

[0075] Figure 2 A flowchart of an MCI diagnosis method based on a multi-scale hypergraph network is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] The embodiments of the present application solve the technical problem of early and accurate diagnosis of MCI by providing an MCI diagnosis system, storage medium and electronic device based on a multi-scale hypergraph network.

[0078] The technical solution in the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:

[0079] In view of the problem of mild cognitive impairment disease diagnosis, the embodiment of the present invention proposes a multi-scale hypergraph network system based on synchronization and attention fusion. Compared with the existing technology, the accuracy and efficiency of mild cognitive impairment (MCI) detection are significantly improved.

[0080] Specifically, by using phase-locked value (PLV) and advanced attention mechanism, the present invention can accurately capture the phase synchronization relationship between brain regions and the importance of different brain regions in the time domain, and then integrate time-frequency information through multi-scale feature fusion mechanism to build a dynamic hypergraph model. This model not only optimizes the representation of the complex structure and function of the brain network, but also further improves the diagnostic effect through parameter optimization of hypergraph neural network and grid search strategy.

[0081] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0082] Embodiment 1:

[0083] like Figure 1 As shown, an embodiment of the present invention provides an MCI diagnosis system based on a multi-scale hypergraph network, including:

[0084] A preprocessing module is used to acquire and preprocess the fMRI signals of the patient to be diagnosed, so as to obtain the BOLD time series of each ROI area on the functional image;

[0085] The synchronization analysis module is used to evaluate the phase synchronization between paired ROI time signals using phase-locked value technology and obtain the second-order spectrum matrix;

[0086] The adaptive module is used to calculate the weight coefficients between paired ROI time signals using the graph attention mechanism and obtain the time domain matrix after smoothing;

[0087] A fusion module, used for fusing the second-order spectrum matrix and the time domain matrix to obtain a time-frequency fusion matrix;

[0088] An extraction module, used to use the time-frequency fusion matrix as an input of a hypergraph neural network to obtain a feature representation of a high-order relationship between different ROI regions;

[0089] The classification module is used to predict the MCI diagnosis result of the patient to be diagnosed based on the final output of the hypergraph neural network.

[0090] The embodiments of the present invention promote the early detection and intervention of MCI, provide personalized diagnostic support for doctors, accelerate the diagnostic process, and provide new tools for neuroscience research, which has important clinical and scientific research value.

[0091] like Figure 2 As shown, Figure 2 A flowchart of the MCI diagnosis method based on a multi-scale hypergraph network is provided. Figure 2 Detailed introduction to each module of the above solution:

[0092] The preprocessing module is used to acquire and preprocess the fMRI signals of the patient to be diagnosed, so as to obtain the BOLD time series of each ROI region on the functional image.

[0093] This module first obtains the raw image data of fMRI signals. fMRI technology captures brain activity by acquiring magnetic resonance signals from the head of the patient to be diagnosed. The data is usually stored in DICOM (Digital Imaging and Communications) format.

[0094] Since the original fMRI signal is 4-dimensional, it contains 3-dimensional spatial information and 1-dimensional time. In practical applications, brain image time series are more often used for research and analysis, and fMRI signals cannot be used directly for related research. Therefore, this module also needs to preprocess the original data before using it.

[0095] To obtain the BOLD time series of each ROI region on the functional image, one or more regions of interest (ROI) can be defined in the brain region of interest. Common methods include anatomical-based methods (such as template-based ROI) and functional-based methods (such as independent component analysis or seed correlation).

[0096] Exemplarily, this module uses the DPARSF software package to preprocess the fMRI signal, specifically including:

[0097] The original functional images of the first 10 time points of the fMRI signal were removed, and slice timing correction was performed; the original functional images with head movement exceeding 2.5 mm or 2.5 degrees were excluded and motion correction was performed; statistical parametric mapping (SPM) was used to co-register with T1-weighted images, and gray matter, white matter and cerebrospinal fluid were segmented; DARTEL was used to map the functional images to the MNI (Montreal Neurological Institute) space proposed by the Montreal Neurological Institute; the functional images were smoothed with a 4 mm full width at half maximum (FWHM) Gaussian kernel to reduce registration variability; and the functional image sequence was bandpass filtered between 0.01 and 0.1 Hz.

[0098] Finally, the acquired BOLD time series is expressed as:

[0099]

[0100] Among them, x i represents the BOLD time series of the i-th ROI region as the ROI time signal; represents a real number set; N represents the number of ROI areas; F de Indicates the length of the signal.

[0101] As for the synchronization analysis module, it is used to evaluate the phase synchronization degree between paired ROI time signals by using phase-locked value technology to obtain the second-order spectrum matrix.

[0102] To expand, Figure 2 As shown, this module is specifically used for:

[0103] The instantaneous phase of a signal is estimated using the Hilbert transform as follows:

[0104]

[0105] in, represents the Hilbert transform of the signal x(t); PV represents the Cauchy principal value; x(t′) represents the value of the original signal at different times during the integration process; t′ represents the integration variable.

[0106] The analytic signal phase is defined as:

[0107]

[0108] Where φ(t) represents the instantaneous phase of the signal; arctan is the inverse tangent function.

[0109] Calculate the paired ROI time signal x i and x j The PLV values ​​between are as follows:

[0110]

[0111] Where l∈{1,2,...,T} represents a time period in T, and T is the total number of time periods; represents an imaginary unit; Δφ ij (t l )=(φ i (t l )-φ j (t l )) represents the instantaneous phase difference of the ROI time signal, respectively from the ROI time signal x i The data x of the lth period i (t l ) and ROI time signal x j The data x of the lth period j (t l ), calculate φ i (t l ) and φ j (t l ), and use Hilbert transform to get the phase value.

[0112] Using ROI time signal x i and x j The synchronous correlation between them forms the second-order spectrum matrix The definition is as follows:

[0113]

[0114] For the adaptive module, it is used to calculate the weight coefficients between paired ROI time signals using the graph attention mechanism, and obtain the time domain matrix after smoothing.

[0115] To expand, Figure 2 As shown, this module is specifically used for:

[0116] Calculate ROI time signal x using graph attention mechanism i and x j The weight coefficient e ij Connection strength as time domain:

[0117]

[0118] Among them, exp is the exponential function; LeakyReLU is the activation function; W e are parameterized vectors and learnable weight matrices respectively; || is a connection operation; if is greater than the preset threshold τ, Represents x i The directly related node set, k is the node index in the node set.

[0119] use Defining the attention matrix Represents the correlation of all ROI time signals in the time domain and minimizes the optimization function Get the best combination of learnable parameters:

[0120]

[0121] Among them, λ≥0 represents the regularization parameter.

[0122] In order to prevent over-focusing on a specific ROI area, an all-one matrix is ​​then defined Smoothed attention matrix M e :

[0123] C TEM =M e +M ones (8)

[0124] Among them, C TEMrepresents the time domain matrix after smoothing.

[0125] The fusion module is used to fuse the second-order spectrum matrix and the time domain matrix to obtain a time-frequency fusion matrix.

[0126] Taking into account the real changes of actual bioelectric signals in the process of contributing to the brain network, the embodiment of the present invention focuses on studying a multi-scale feature fusion mechanism to fuse the time domain and spectrum dynamic connection features of fMRI.

[0127] In detail, C SPE Provides the synchronization relationship of spectral domain ROI, C TEM Provides the importance relationship of ROI in the time domain. Figure 2 As shown in the figure, this module transforms C TEM and C SPE Combined, they reveal synchronization and dynamic changes between ROI regions:

[0128] C TS =C TEM ⊙C SPE (9)

[0129] Among them, C TS represents the time-frequency fusion matrix; C TEM , C SPE represent the time domain matrix and the second-order spectrum matrix respectively; ⊙ represents the Hadamard product.

[0130] As for the extraction module, it is used to use the time-frequency fusion matrix as the input of the hypergraph neural network to obtain the feature representation of the high-order relationship between different ROI areas.

[0131] like Figure 2 As shown, this module uses the time-frequency fusion matrix C TS Form a hypergraph of feature representation; specifically used for:

[0132] Define a hypergraph using Indicates that represents a node set, n is the total number of nodes; ε={e1,e2,...,e m} represents the hyperedge set, m is the total number of hyperedges; is the weight matrix of the hyperedge; each hyperedge e connects multiple nodes v and is represented by The diagonal matrix weights .

[0133] Hypergraph Expressed as The incidence matrix H is defined as follows:

[0134]

[0135] Based on the time-frequency fusion matrix C TS Construct a hyperedge. For the i-th ROI area, in C TS Select the associated regions K with the highest connectivity value from the top_K i .

[0136] Based on the set top_K i , redefine the incidence matrix H as:

[0137]

[0138] Among them, j represents the set top_K i The index of the associated region in .

[0139] For the incidence matrix H, the degree matrix of its vertices is defined as use Indicates that diag is a diagonal matrix, d(v i ) represents node v i The degree of ; and the degree matrix of its hyperedge is expressed as in δ(e i ) represents the hyperedge e i The degree.

[0140] The hypergraph neural network is used to capture the high-order relationship between different ROI regions for feature representation, where the lth convolutional layer is expressed as:

[0141]

[0142] The superscript T indicates transposition; X (l) represents the convolution output feature of the lth layer, Θ (l) As a filter to extract node features in the hypergraph, represents the learnable hyperedge weight matrix, W(e i ) represents the hyperedge e i The weight coefficient of .

[0143] Finally, after extracting features through the hypergraph neural network, the output Z is used for subsequent graph-level reading and disease classification.

[0144] As for the classification module, it is used to predict the MCI diagnosis result of the patient to be diagnosed based on the final output of the hypergraph neural network.

[0145] To expand, Figure 2 As shown, this module trains a multilayer perceptron to predict the MCI diagnosis result of the patient to be diagnosed, which is expressed as:

[0146]

[0147] in, represents the MCI diagnosis result of the patient to be diagnosed; MLP represents the multi-layer perceptron; Z represents the final output of the hypergraph neural network.

[0148] It should be noted that the hypergraph neural network described in the embodiment of the present invention uses the cross entropy function to estimate the prediction loss during the training phase, which is expressed as:

[0149]

[0150] Where s represents the index of the subject sample; S is the total number of subject samples; y s , They represent the true label and predicted label of the sth subject sample respectively; log is the logarithmic function.

[0151] So far, the embodiment of the present invention has introduced the complete execution process of the MCI diagnosis system based on the multi-scale hypergraph network.

[0152] Embodiment 2:

[0153] An embodiment of the present invention provides a storage medium storing a computer program for MCI diagnosis based on a multi-scale hypergraph network, wherein the computer program enables a computer to control the MCI diagnosis system as described in Example 1 to perform MCI diagnosis.

[0154] Embodiment 3:

[0155] An embodiment of the present invention provides an electronic device, including:

[0156] One or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, so that the one or more processors control the MCI diagnostic system as described in Example 1 to perform MCI diagnosis.

[0157] It can be understood that the storage medium and electronic device provided in the embodiments of the present invention correspond to the MCI diagnostic system based on the multi-scale hypergraph network provided in the embodiments of the present invention. The explanations, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the MCI diagnostic system and will not be repeated here.

[0158] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0159] 1. In an embodiment of the present invention, the fMRI signal of the patient to be diagnosed is first preprocessed to obtain the BOLD time series of each ROI area on the functional image; then the phase synchronization relationship of the ROI area in the spectrum domain is calculated by the phase-locked value technology; and the graph attention mechanism is used to dynamically learn and identify the relative importance of different brain regions in the time domain; then a multi-scale feature fusion mechanism is designed to integrate the time domain and spectrum dynamic connection features of the fMRI signal to form a time-frequency fusion matrix, and a dynamic hypergraph model is established based on the matrix; finally, a hypergraph neural network is used to construct a detection model for mild cognitive impairment, and the parameters are optimized through a grid search strategy to improve the MCI diagnostic effect.

[0160] 2. The embodiments of the present invention promote the early detection and intervention of MCI, provide personalized diagnostic support for doctors, accelerate the diagnostic process, and provide new tools for neuroscience research, which has important clinical and scientific research value.

[0161] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0162] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A MCI diagnosis system based on a multi-scale hypergraph network, characterized in that: include: A preprocessing module is used to acquire and preprocess the fMRI signals of the patient to be diagnosed, so as to obtain the BOLD time series of each ROI area on the functional image; The synchronization analysis module is used to evaluate the phase synchronization between paired ROI time signals using phase-locked value technology and obtain the second-order spectrum matrix; The adaptive module is used to calculate the weight coefficients between paired ROI time signals using the graph attention mechanism and obtain the time domain matrix after smoothing; A fusion module, used for fusing the second-order spectrum matrix and the time domain matrix to obtain a time-frequency fusion matrix; An extraction module, used to use the time-frequency fusion matrix as an input of a hypergraph neural network to obtain a feature representation of a high-order relationship between different ROI regions; The classification module is used to predict the MCI diagnosis result of the patient to be diagnosed based on the final output of the hypergraph neural network.

2. The MCI diagnostic system according to claim 1, wherein: The preprocessing module uses the DPARSF software package to preprocess the fMRI signal, specifically including: The original functional images of the first 10 time points of the fMRI signal were removed and slice timing correction was performed; the original functional images with head movement exceeding 2.5 mm or 2.5 degrees were excluded and motion correction was performed; statistical parameter mapping was used to co-register with T1-weighted images, and gray matter, white matter and cerebrospinal fluid were segmented; DARTEL was used to map the functional images to MNI space; a half-maximum Gaussian kernel was used to smooth the functional images by 4 mm full width to reduce registration variability; the functional image sequence was bandpass filtered between 0.01 and 0.1 Hz; The acquired BOLD time series is expressed as: Among them, x i represents the BOLD time series of the i-th ROI region as the ROI time signal; represents a real number set; N represents the number of ROI areas; F de Indicates the length of the signal.

3. The MCI diagnostic system according to claim 2, wherein: The synchronization analysis module is specifically used for: The instantaneous phase of a signal is estimated using the Hilbert transform as follows: in, represents the Hilbert transform of the signal x(t); PV represents the Cauchy principal value; x(t′) represents the value of the original signal at different times during the integration process; t′ represents the integration variable; The analytic signal phase is defined as: Where φ(t) represents the instantaneous phase of the signal; arctan is the inverse tangent function; Calculate the paired ROI time signal x i and x j The PLV values ​​between are as follows: Where l∈{1,2,...,T} represents a time period in T, and T is the total number of time periods; represents an imaginary unit; Δφ ij (t l )=(φ i (t l )-φ j (t l )) represents the instantaneous phase difference of the ROI time signal, respectively from the ROI time signal x i The data x of the lth period i (t l ) and ROI time signal x j The data x of the lth period j (t l ), calculate φ i (t l ) and φ j (t l ), use Hilbert transform to get the phase value; Using ROI time signal x i and x j The synchronous correlation between them forms the second-order spectrum matrix The definition is as follows:

4. The MCI diagnostic system according to claim 2, wherein: The adaptive module is specifically used for: Calculate ROI time signal x using graph attention mechanism i and x j The weight coefficient e ij Connection strength as time domain: Among them, exp is the exponential function; LeakyReLU is the activation function; W e are parameterized vectors and learnable weight matrices respectively; || is a connection operation; if is greater than the preset threshold τ, Represents x i The directly related node set, k is the node index in the node set; use Defining the attention matrix Represents the correlation of all ROI time signals in the time domain and minimizes the optimization function Get the best combination of learnable parameters: Among them, λ≥0 represents the regularization parameter; Define the all-ones matrix Smoothed attention matrix M e : C TEM =M e +M ones (8) Among them, C TEM represents the time domain matrix after smoothing.

5. The MCI diagnostic system according to claim 1, wherein: The fusion module is specifically used for: C TS =C TEM ⊙C SPE (9) Among them, C TS represents the time-frequency fusion matrix; C TEM , C SPE represent the time domain matrix and the second-order spectrum matrix respectively; ⊙ represents the Hadamard product.

6. The MCI diagnostic system according to claim 5, characterized in that: The extraction module is specifically used for: Define a hypergraph using Indicates that represents a node set, n is the total number of nodes; ε={e1,e2,...,e m } represents the hyperedge set, m is the total number of hyperedges; is the weight matrix of the hyperedge; each hyperedge e connects multiple nodes v and is represented by The diagonal matrix weights of ; Hypergraph Expressed as The incidence matrix H is defined as follows: Based on the time-frequency fusion matrix C TS Construct a hyperedge. For the i-th ROI area, in C TS Select the associated regions K with the highest connectivity value from the top_K i ; Based on the set top_K i , redefine the incidence matrix H as: Among them, j represents the set top_K i The index of the associated region in ; For the incidence matrix H, the degree matrix of its vertices is defined as use Indicates that diag is a diagonal matrix, d(v i ) represents node v i The degree of ; and the degree matrix of its hyperedge is expressed as in δ(e i ) represents the hyperedge e i degree; The hypergraph neural network is used to capture the high-order relationship between different ROI regions for feature representation, where the lth convolutional layer is expressed as: The superscript T indicates transposition; X (l) represents the convolution output feature of the lth layer, Θ (l) As a filter to extract node features in the hypergraph, represents the learnable hyperedge weight matrix, W(e i ) represents the hyperedge e i The weight coefficient of .

7. The MCI diagnostic system according to claim 1, wherein: The classification module trains a multilayer perceptron to predict the MCI diagnosis result of the patient to be diagnosed, which is expressed as: in, represents the MCI diagnosis result of the patient to be diagnosed; MLP represents the multi-layer perceptron; Z represents the final output of the hypergraph neural network.

8. The MCI diagnostic system according to any one of claims 1 to 7, characterized in that: The hypergraph neural network uses the cross entropy function to estimate the prediction loss during the training phase, expressed as: Where s represents the index of the subject sample; S is the total number of subject samples; y s , They represent the true label and predicted label of the sth subject sample respectively; log is the logarithmic function.

9. A storage medium, characterized in that: It stores a computer program for MCI diagnosis based on a multi-scale hypergraph network, wherein the computer program enables a computer to control the MCI diagnosis system as described in any one of claims 1 to 8 to perform MCI diagnosis.

10. An electronic device, characterized in that: include: one or more processors; Memory; and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, so that the one or more processors control the MCI diagnostic system as described in any one of claims 1 to 8 to perform MCI diagnosis.

Images