Alzheimer disease risk prediction model processing method and device

Abstract

The embodiment of the invention relates to a processing method and device of an Alzheimer disease risk prediction model. The method comprises the steps that DT I-ALPS feature information of dementia crowds and healthy crowds of a specified age group is obtained in a big data collection and volunteer recruitment mode, and a corresponding model data set is constructed and recorded as a first data set; constructing a three-classification prediction model for predicting the risk of the Alzheimer's disease as an Alzheimer's disease risk prediction model corresponding to a specified age group; training an Alzheimer disease risk prediction model based on the first data set; after model training is finished, DT I-ALPS feature information, input by the user, of any tested person of the specified age group is input into the Alzheimer disease risk prediction model for prediction, and a corresponding classification probability vector is obtained and fed back to the current user. According to the invention, prediction accuracy and prediction stability can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and device for processing an Alzheimer's disease risk prediction model. Background Art

[0002] Diffusion Tensor Image Analysis Along the Perivascular Space (DTI-ALPS) is a new technology developed based on Diffusion Tensor Imaging (DTI). This technology can noninvasively observe water diffusion in the perivascular space and assess the functional status of the brain's glymphatic system (GS) based on the ALPS index.

[0003] Dementia is a type of neurocognitive disorder. Mild cognitive impairment (MCI), vascular cognitive impairment (VCI), and Alzheimer's disease (AD) are common subtypes of dementia. MCI can be further divided into non-AD-related MCI and AD-related MCI, the latter of which can progress to AD. Studies have shown that GS functional status is associated with MCI, VCI, and AD, and that the DTI-ALPS index is related to cognitive function. In principle, AD risk can be predicted by analyzing the DTI-ALPS index and its related features. These features refer to the xyz-axis diffusivity of the projection fiber region of interest (ROI) and the xyz-axis diffusivity of the association fiber region of interest, as discussed in the DTI-ALPS technique.

[0004] An investigation of current AD risk prediction mechanisms based on the ALPS index found that while DTI-ALPS technology can generate a relatively accurate DTI-ALPS index and its associated features, predictions still rely on physician experience within a set of pre-defined thresholds. Limited by physician expertise, this traditional prediction mechanism's stability and accuracy need further improvement. Summary of the Invention

[0005] The purpose of the present invention is to address the shortcomings of the existing technology and provide a processing method, device, electronic device and computer-readable storage medium for an Alzheimer's disease risk prediction model. The present invention first obtains DTI-ALPS feature information of dementia patients (including non-AD-derived MCI patients, VCI patients, AD-derived MCI patients, AD patients) and healthy people in a specified age group through big data collection and volunteer recruitment, and thereby constructs a corresponding model data set, namely the first data set; then, non-AD-derived MCI patients, VCI patients and healthy people are regarded as non-AD class objects (i.e., first class objects), AD-derived MCI patients are regarded as second class objects, and AD patients are regarded as third class objects; then, a three-class prediction model for predicting Alzheimer's disease risk is constructed as an Alzheimer's disease risk prediction model; then, the model is trained based on the first data set; and after the training, the prediction model is used to predict the AD risk of the subject in three categories (non-AD class, AD-derived MCI class, AD class) according to the DTI-ALPS feature information. The present invention can improve prediction accuracy and prediction stability through the learning ability of artificial intelligence models and model training with large amounts of data; based on the present invention, adaptation models for multiple age groups can be trained by collecting model data sets for different age groups, thereby achieving the purpose of improving prediction flexibility.

[0006] To achieve the above objectives, a first aspect of an embodiment of the present invention provides a method for processing an Alzheimer's disease risk prediction model, the method comprising:

[0007] The DTI-ALPS feature information of a dementia population and a healthy population of a specified age group is obtained by big data collection and volunteer recruitment, and a corresponding model dataset is constructed therefrom, which is recorded as a first dataset; the dementia population includes a plurality of dementia patients, and the dementia patients include at least patients with mild cognitive impairment not caused by Alzheimer's disease, patients with vascular cognitive impairment, patients with mild cognitive impairment caused by Alzheimer's disease, and patients with Alzheimer's disease; the healthy population consists of a plurality of healthy people who do not suffer from any dementia;

[0008] The patients with mild cognitive impairment not caused by Alzheimer's disease, the patients with vascular cognitive impairment and the healthy subjects are taken as the first category of objects, the patients with mild cognitive impairment caused by Alzheimer's disease are taken as the second category of objects, and the patients with Alzheimer's disease are taken as the third category of objects; and based on the first, second and third category objects, a three-category prediction model for predicting the risk of Alzheimer's disease is constructed as the Alzheimer's disease risk prediction model corresponding to the specified age group; the Alzheimer's disease risk prediction model is used to perform three-category prediction based on the DTI-ALPS feature information input by the model and output a corresponding classification probability vector; the classification probability vector includes first, second and third classification probabilities, the first, second and third classification probabilities correspond one-to-one to the first, second and third category objects, and the sum of the first, second and third classification probabilities is 1;

[0009] Training the Alzheimer's disease risk prediction model based on the first data set;

[0010] After the model training is completed, the DTI-ALPS feature information of any subject in the specified age group input by the user is input into the Alzheimer's disease risk prediction model for prediction to obtain the corresponding classification probability vector and feedback it to the current user.

[0011] Preferably, the DTI-ALPS feature information is composed of a plurality of ALPS features calculated based on the DTI-ALPS technology; the ALPS feature at least includes feature D xproj , Feature D yproj , Feature D zproj , Feature D xassoc , Feature D yassoc , Feature D zassoc and ALPS index characteristics;

[0012] The calculation principle of the ALPS feature is:

[0013] Based on the principle of DTI-ALPS technology, a pair of symmetrical projection fiber regions of interest and a pair of symmetrical association fiber regions of interest were set at the level of the left and right ventricles in the DTI images of the subjects' brains.

[0014] Based on the principle of DTI-ALPS technology, the diffusivity of the left brain projection fiber area of ​​interest along the x, y, and z axes is calculated to obtain the corresponding left brain projection fiber x, y, and z axial diffusivity. The diffusion rates of the right brain projection fiber area of ​​interest along the x, y, and z axes are calculated respectively to obtain the corresponding right brain projection fiber x, y, and z axial diffusion rates.

[0015] Based on the DTI-ALPS technology principle, the diffusivity of the left brain joint fiber area of ​​interest along the x, y, and z axes is calculated to obtain the corresponding left brain joint fiber x, y, and z axial diffusivity. The diffusion rates of the right brain joint fiber region of interest along the x, y, and z axes are calculated respectively to obtain the corresponding right brain joint fiber x, y, and z axial diffusion rates.

[0016] According to the DTI-ALPS index calculation principle, the diffusion rate Calculate the corresponding left brain ALPS index, based on the diffusion rate Calculate the corresponding right brain ALPS index:

[0017]

[0018] mean() is the mean value calculation function;

[0019] And the diffusion rate The average value of the corresponding feature D xproj ;

[0020] And the diffusion rate The average value of the corresponding feature D yproj ;

[0021] And the diffusion rate The average value of the corresponding feature D zproj ;

[0022] And the diffusion rate The average value of the corresponding feature D xassoc ;

[0023] And the diffusion rate The average value of the corresponding feature D yassoc ;

[0024] And the diffusion rate The average value of the corresponding feature D zassoc ;

[0025] The average value of the left and right brain ALPS indices is used as the corresponding ALPS index feature.

[0026] Preferably, the model input end of the Alzheimer's disease risk prediction model is used to receive the DTI-ALPS feature information, and the model output end is used to output the classification probability vector;

[0027] The Alzheimer's disease risk prediction model includes a first feature extraction layer, a channel attention layer, a second feature extraction layer, a linear mapping layer and a prediction output layer;

[0028] The input end of the first feature extraction layer is connected to the input end of the model, and the output end is connected to the input end of the channel attention layer; the output end of the channel attention layer is connected to the input end of the second feature extraction layer; the output end of the second feature extraction layer is connected to the input end of the linear mapping layer; the output end of the linear mapping layer is connected to the input end of the prediction output layer; the output end of the prediction output layer is connected to the output end of the model;

[0029] The first feature extraction layer is used to take the DTI-ALPS feature information as an input vector X with a vector length of 7; and perform feature encoding processing on the input vector X through a fully connected layer to obtain a corresponding feature vector H1 and send it to the channel attention layer;

[0030] The encoding method of the feature vector H1 is:

[0031] H1=ReLU(W1X+b1),

[0032] ReLU() is the ReLU activation function, W1 and b1 are the fully connected layer weight matrix parameters and offset vector parameters of the first feature extraction layer; the shape of the weight matrix parameter W1 is C1×7, the shape of the offset vector parameter b1 is C1×1, C1 is the preset first feature dimension, C1>7; the shape of the feature vector H1 is C1×1, which is composed of C1 feature data Composition, 1≤indexi≤C1;

[0033] The channel attention layer is used to perform global average pooling on the feature vector H1 to obtain a corresponding pooling scalar z; and generate a corresponding channel attention weight vector A based on the pooling scalar z through two fully connected layers; and based on the channel attention weight vector A, the feature vector H1 is weighted to obtain a corresponding feature vector H2 and sent to the second feature extraction layer;

[0034] The global average pooling method of the pooling scalar z is:

[0035]

[0036] The encoding method of the channel attention weight vector A is:

[0037] A=ReLU(W3Sigmoid(W2z+b2)+b3);

[0038] Sigmoid() is a Sigmoid activation function, W2 and b2 are the weight matrix parameters and offset vector parameters of the first fully connected layer of the channel attention layer, W3 and b3 are the weight matrix parameters and offset vector parameters of the second fully connected layer of the channel attention layer; the shape of the weight matrix parameter W2 is C2×1, the shape of the offset vector parameter b2 is C2×1, C2 is the preset second feature dimension, C2<C1; the shape of the weight matrix parameter W3 is C3×C2, the shape of the offset vector parameter b3 is C3×1, C3 is the preset third feature dimension, C3=C1; the output feature shape of Sigmoid(W2z+b2) is C2×1; the shape of the channel attention weight vector A is C3×1=C1×1;

[0039] The encoding method of the feature vector H2 is:

[0040] H2=H1⊙A;

[0041] ⊙ is Hadamard;

[0042] The second feature extraction layer is used to perform feature encoding processing on the feature vector H2 through a fully connected layer to obtain a corresponding feature vector H3 and send it to the linear mapping layer;

[0043] The encoding method of the feature vector H3 is:

[0044] H3=ReLU(W4H2+b4),

[0045] W4 and b4 are the fully connected layer weight matrix parameters and offset vector parameters of the second feature extraction layer; the shape of the weight matrix parameter W4 is C4×C3, and the shape of the offset vector parameter b4 is C4×1, where C4 is the preset fourth feature dimension, C4<C3; the shape of the feature vector H3 is C4×1;

[0046] The linear mapping layer is used to perform feature mapping processing on the feature vector H3 through a linear layer to obtain a corresponding feature vector H4 and send it to the prediction output layer;

[0047] The encoding method of the feature vector H4 is:

[0048] H4=W5H3+b5,

[0049] W5 and b5 are the linear layer weight matrix parameters and offset vector parameters of the linear mapping layer; the shape of the weight matrix parameter W5 is C5×C4, the shape of the offset vector parameter b5 is C5×1, C5 is the preset fifth feature dimension, C5=3; the shape of the feature vector H4 is C5×1;

[0050] The prediction output layer is used to use the Softmax function to perform three-category probability calculation based on the feature vector H4 to obtain the corresponding first, second and third classification probabilities; and the corresponding classification probability vector is composed of the obtained first, second and third classification probabilities and output.

[0051] Preferably, the first data set includes multiple first data records; the first data record includes first training feature information and a first label vector; the first training feature information is one of the DTI-ALPS feature information; the first label vector includes first, second and third label probabilities, and the first, second and third label probabilities correspond one-to-one to the first, second and third categories of objects; only one of the first, second and third label probabilities is 1, and the other two probabilities are 0.

[0052] Preferably, the DTI-ALPS feature information of dementia patients and healthy people of a specified age group is obtained by big data collection and volunteer recruitment, and the corresponding model dataset is constructed therefrom, which is recorded as the first dataset, specifically including:

[0053] Data is collected from the brain DTI images of the patients with mild cognitive impairment not caused by Alzheimer's disease, the patients with vascular cognitive impairment, the patients with mild cognitive impairment caused by Alzheimer's disease, the patients with Alzheimer's disease, and the healthy people in the specified age group through multiple preset big data collection channels; and the patients with mild cognitive impairment not caused by Alzheimer's disease, the patients with vascular cognitive impairment, the patients with mild cognitive impairment caused by Alzheimer's disease, the patients with Alzheimer's disease, and the healthy people in the specified age group are recruited through volunteer recruitment, and brain magnetic resonance imaging examination is performed on each recruited subject to obtain corresponding brain DTI images ; and construct a first image set based on all the obtained brain DTI images and their corresponding patient or healthy person types; wherein the big data acquisition channel includes at least a public medical image data set, a medical institution database that is public or authorized for data use, and a medical image sharing platform that is public or authorized for data use; the first image set includes multiple first image records; the first image record includes a first image and a first object type; the first image is a brain DTI image; the first object type includes a mild cognitive impairment type not originating from Alzheimer's disease, a vascular cognitive impairment type, a mild cognitive impairment type originating from Alzheimer's disease, an Alzheimer's disease type, and a healthy type;

[0054] and taking each of the first image records of the first image set as the corresponding current record; and taking the first image and the first object type of the current record as the corresponding current image and current type; and performing ALPS feature calculation on the current image based on the calculation principle of the ALPS feature to obtain the corresponding DTI-ALPS feature information; and taking the DTI-ALPS feature information obtained this time as the corresponding first training feature information; and identifying the current type; if the current type is the mild cognitive impairment type originating from non-Alzheimer's disease, the vascular cognitive impairment type or the healthy type, setting the corresponding first, second and third label probabilities to 1, 0, 0; if the current type is the mild cognitive impairment type originating from Alzheimer's disease, setting the corresponding first, second and third label probabilities to 0, 1, 0; if the current type is the Alzheimer's disease type, setting the corresponding first, second and third label probabilities to 0, 0, 1; and forming a corresponding first label vector from the first, second and third label probabilities obtained this time; and forming a corresponding first data record from the first training feature information and the first label vector corresponding to the current record;

[0055] All the obtained first data records form the corresponding first data set.

[0056] Preferably, the training of the Alzheimer's disease risk prediction model based on the first data set specifically includes:

[0057] Step 61, initializing the first round counter to 1; and forming a corresponding model parameter set by the weight matrix parameters W1, W3, W3, W4, W5 and the offset vector parameters b1, b2, b3, b4, b5 of the Alzheimer's disease risk prediction model;

[0058] Step 62: randomly split the first data set into two sub-data sets based on a preset first split ratio and record them as a corresponding current training set and a current evaluation set;

[0059] Wherein, the current training set and the current evaluation set are both composed of a plurality of the first data records; the ratio of the total number of records in the current training set and the current evaluation set satisfies the first segmentation ratio;

[0060] Step 63: record the total number of records in the current training set as the current total number of records N tr; and statistically calculate the total number of the first data records corresponding to the first, second, and third class objects in the current training set to obtain the corresponding first class total number N1, second class total number N2, and third class total number N3; and set the corresponding first class sample weight w1, second class sample weight w2, and third class sample weight w3 based on the first, second, and third class total numbers N1, N2, and N3;

[0061] The weights w1, w2, and w3 of the first, second, and third class samples are set as follows:

[0062]

[0063] Step 64, and perform a round of traversal on all the first data records of the current training set; and in this round of traversal, the first data record currently traversed is used as the corresponding current training record R j , 1≤indexj≤N tr ; and record the current training R j The first training feature information is used as the current DTI-ALPS feature information to input into the Alzheimer's disease risk prediction model for prediction, and the classification probability vector output by this prediction is used as the corresponding prediction vector Y j ; and the prediction vector Y j The first, second and third classification probabilities are recorded as the corresponding probabilities y j,k , 1≤index k≤3; and record the current training R j The first label vector is used as the corresponding label vector And the label vector The first, second and third label probabilities are recorded as the corresponding probabilities And the current prediction vector Y j and the label vector Form a corresponding first prediction-label pair

[0064] Step 65, the obtained N tr The first prediction-label pair And the weights w1, w2, and w3 of the first, second, and third class samples are brought into the preset first model loss function L M Calculate and obtain the corresponding first loss value;

[0065] Among them, the first model loss function L M for:

[0066]

[0067] Step 66: Identify whether the first loss value satisfies a preset first loss value range. If so, proceed to step 67. If not, perform a round of parameter modulation on the model parameter set based on a preset first model optimizer. Return to step 63 to continue training after this round of modulation is completed.

[0068] Wherein, the first model optimizer includes at least an Adam optimizer and an SGD optimizer;

[0069] Step 67, perform a round of traversal on all the first data records of the current evaluation set; and during this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the first training feature information of the current evaluation record as the current DTI-ALPS feature information into the Alzheimer's disease risk prediction model for prediction, and use the classification probability vector output by this prediction as the corresponding current prediction vector; and form a corresponding second prediction-label pair by the current prediction vector and the first label vector of the current evaluation record; and at the end of this round of traversal, calculate the accuracy, precision, recall and F1 score based on all the obtained second prediction-label pairs to obtain the corresponding current accuracy, current precision, current recall and current F1 score;

[0070] Step 68: Identify whether the current accuracy, the current precision, the current recall, and the current F1 score all meet their corresponding first accuracy range, first precision range, first recall range, and first F1 score range; if so, proceed to step 69; if not, return to step 63 to continue training;

[0071] In step 69, it is determined whether the first round counter is greater than a preset training round threshold; if not, the first round counter is incremented by 1 and the process returns to step 62 to continue the next round of training; if so, the training is stopped and the model training is confirmed to be complete.

[0072] A second aspect of an embodiment of the present invention provides a device for implementing the processing method of the Alzheimer's disease risk prediction model described in the first aspect, the device comprising: a data set preparation module, a model construction module, a model training module, and a model application module;

[0073] The data set preparation module is used to obtain DTI-ALPS feature information of a dementia population and a healthy population of a specified age group through big data collection and volunteer recruitment, and thereby construct a corresponding model data set recorded as a first data set; the dementia population includes a plurality of dementia patients, and the dementia patients include at least patients with mild cognitive impairment of a non-Alzheimer's disease origin, patients with vascular cognitive impairment, patients with mild cognitive impairment of Alzheimer's disease origin, and patients with Alzheimer's disease; the healthy population consists of a plurality of healthy people who do not suffer from any dementia;

[0074] The model construction module is used to take the patients with mild cognitive impairment of non-Alzheimer's disease origin, the patients with vascular cognitive impairment and the healthy subjects as the first category of objects, the patients with mild cognitive impairment of Alzheimer's disease origin as the second category of objects, and the patients with Alzheimer's disease as the third category of objects; and construct a three-category prediction model for predicting the risk of Alzheimer's disease based on the first, second and third category objects as the Alzheimer's disease risk prediction model corresponding to the specified age group; the Alzheimer's disease risk prediction model is used to perform three-category prediction based on the DTI-ALPS feature information input by the model and output a corresponding classification probability vector; the classification probability vector includes first, second and third classification probabilities, the first, second and third classification probabilities correspond one-to-one to the first, second and third category objects, and the sum of the first, second and third classification probabilities is 1;

[0075] The model training module is used to train the Alzheimer's disease risk prediction model based on the first data set;

[0076] The model application module is used to input the DTI-ALPS feature information of any subject in the specified age group input by the user into the Alzheimer's disease risk prediction model after the model training is completed, to obtain the corresponding classification probability vector and feed it back to the current user.

[0077] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0078] The processor is configured to be coupled to the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;

[0079] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0080] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a computer, the computer executes the instructions of the method described in the first aspect above.

[0081] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for an Alzheimer's disease risk prediction model. As can be seen from the above content, the embodiment of the present invention first obtains the DTI-ALPS feature information of dementia population (including non-AD-derived MCI patients, VCI patients, AD-derived MCI patients, AD patients) and healthy people in a specified age group through big data collection and volunteer recruitment, and thereby constructs a corresponding model data set, i.e., the first data set; then, non-AD-derived MCI patients, VCI patients and healthy people are regarded as non-AD objects (i.e., first-class objects), AD-derived MCI patients are regarded as second-class objects, and AD patients are regarded as third-class objects; then, a three-class prediction model for predicting the risk of Alzheimer's disease is constructed as an Alzheimer's disease risk prediction model; then, the model is trained based on the first data set; and after the training, the prediction model is used to perform a three-class prediction of the AD risk of the subject according to the DTI-ALPS feature information (non-AD class, AD-derived MCI class, AD class). The embodiments of the present invention improve the prediction accuracy and stability of AD risk prediction through the learning ability of the artificial intelligence model and model training with large amounts of data; based on the embodiments of the present invention, model data sets of different age groups can be collected to train adapted model versions for multiple age groups, thereby improving the prediction flexibility of AD risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A schematic diagram of a processing method for an Alzheimer's disease risk prediction model provided in Example 1 of the present invention;

[0083] Figure 2 This is a schematic diagram of the modules of the Alzheimer's disease risk prediction model provided in Example 1 of the present invention;

[0084] Figure 3 A module structure diagram of a processing device for an Alzheimer's disease risk prediction model provided in Example 2 of the present invention;

[0085] Figure 4 This is a structural diagram of an electronic device provided in Example 3 of the present invention. DETAILED DESCRIPTION

[0086] To make the objectives, technical solutions, and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the embodiments described herein are merely some, rather than all, of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0087] The first embodiment of the present invention provides a method for processing an Alzheimer's disease risk prediction model, such as Figure 1 As shown in the schematic diagram of a processing method for an Alzheimer's disease risk prediction model provided in Example 1 of the present invention, the method mainly includes the following steps:

[0088] Step 1: Obtain DTI-ALPS feature information of dementia patients and healthy people in a specified age group through big data collection and volunteer recruitment, and construct a corresponding model dataset based on this information, which is recorded as the first dataset.

[0089] Here, the designated age group in the embodiment of the present invention is a preset age group information. The dementia population in the embodiment of the present invention includes a plurality of dementia patients; wherein, the dementia patients include at least patients with mild cognitive impairment of non-Alzheimer's disease origin, patients with vascular cognitive impairment, patients with mild cognitive impairment of Alzheimer's disease origin and patients with Alzheimer's disease; the healthy population in the embodiment of the present invention is composed of a plurality of healthy people who do not suffer from any dementia. The DTI-ALPS feature information in the embodiment of the present invention is composed of a plurality of ALPS features calculated based on the DTI-ALPS technology; wherein, the ALPS feature includes at least feature D xproj , Feature D yproj , Feature D zproj , Feature D xassoc , Feature D yassoc , Feature D zassoc and ALPS index characteristics.

[0090] The principles of DTI-ALPS technology can be understood by referring to the public technical document "Evaluation of glymphatic system activity with the diffusion MR technique: diffusion tensor image analysis along the perivascular space (DTI-ALPS) in Alzheimer's disease cases." All feature data in the ALPS feature of the present invention are calculated based on the principles of DTI-ALPS technology. The specific calculation principle of the ALPS feature is as follows:

[0091] Step A1: Based on the principle of DTI-ALPS technology, a pair of symmetrical projection fiber regions of interest (ROIs) and a pair of symmetrical association fiber regions of interest (ROIs) are set at the level of the left and right ventricles in the DTI images of the subject's brain, respectively.

[0092] Step A2, based on the principle of DTI-ALPS technology, the diffusivity of the left brain projection fiber area of ​​interest along the x, y, and z axes is calculated to obtain the corresponding left brain projection fiber x, y, and z axial diffusivity. The diffusion rates of the right brain projection fiber area of ​​interest along the x, y, and z axes are calculated respectively to obtain the corresponding x, y, and z axial diffusion rates of the right brain projection fiber.

[0093] Step A3, based on the principle of DTI-ALPS technology, the diffusivity of the left brain association fiber region of interest along the x, y, and z axes is calculated to obtain the corresponding left brain association fiber x, y, and z axial diffusivity. The diffusion rates of the right brain association fiber area of ​​interest along the x, y, and z axes are calculated respectively to obtain the corresponding right brain association fiber x, y, and z axial diffusion rates.

[0094] Step A4, and calculate the diffusion rate according to the DTI-ALPS index principle Calculate the corresponding left brain ALPS index, based on the diffusion rate Calculate the corresponding right brain ALPS index;

[0095] Here, the left and right brain ALPS indexes are specifically:

[0096]

[0097] Among them, mean() is the average value calculation function;

[0098] Step A5, and the diffusion rate The average value of the corresponding feature D xproj ; and the diffusion rate The average value of the corresponding feature D yproj ; and the diffusion rate The average value of the corresponding feature D zproj ; and the diffusion rate The average value of the corresponding feature D xassoc ; and the diffusion rate The average value of the corresponding feature D yassoc ; and the diffusion rate The average value of the corresponding feature D zassoc ; and the average value of the left and right ALPS indices was used as the corresponding ALPS index feature.

[0099] Step 1 of the embodiment of the present invention specifically includes:

[0100] Step 11: collecting brain DTI images of patients with mild cognitive impairment (NCI) not caused by Alzheimer's disease, patients with vascular cognitive impairment, patients with mild cognitive impairment (NCI) caused by Alzheimer's disease, patients with Alzheimer's disease, and healthy subjects in a specified age group through multiple preset big data collection channels; recruiting patients with mild cognitive impairment (NCI) not caused by Alzheimer's disease, patients with vascular cognitive impairment, patients with mild cognitive impairment (NCI) caused by Alzheimer's disease, patients with Alzheimer's disease, and healthy subjects in a specified age group through volunteer recruitment, and performing brain magnetic resonance imaging on each recruited subject to obtain corresponding brain DTI images; and constructing a first image set based on all the obtained brain DTI images and their corresponding patient or healthy subject types;

[0101] The big data collection channels include at least public medical imaging data sets, public or authorized medical institution databases, and public or authorized medical imaging sharing platforms; the first image set includes multiple first image records; the first image record includes a first image and a first object type; the first image is a brain DTI image; the first object type includes a mild cognitive impairment type not caused by Alzheimer's disease, a vascular cognitive impairment type, a mild cognitive impairment type caused by Alzheimer's disease, an Alzheimer's disease type, and a healthy type;

[0102] Step 12, and use each first image record of the first image set as the corresponding current record; and use the first image and the first object type of the current record as the corresponding current image and current type; and based on the calculation principle of ALPS features, perform ALPS feature calculation according to the current image to obtain corresponding DTI-ALPS feature information; and use the DTI-ALPS feature information obtained this time as a corresponding first training feature information; and identify the current type; if the current type is a mild cognitive impairment type originating from non-Alzheimer's disease, a vascular cognitive impairment type, or a healthy type, then set the corresponding first, second, and third label probabilities to 1, 0, 0; if the current type is a mild cognitive impairment type originating from Alzheimer's disease, then set the corresponding first, second, and third label probabilities to 0, 1, 0; if the current type is an Alzheimer's disease type, then set the corresponding first, second, and third label probabilities to 0, 0, 1; and form a corresponding first label vector from the first, second, and third label probabilities obtained this time; and form a corresponding first data record from the first training feature information and the first label vector corresponding to the current record;

[0103] In step 13, all the obtained first data records are combined into a corresponding first data set.

[0104] The first data set of an embodiment of the present invention includes multiple first data records; wherein, the first data record includes first training feature information and a first label vector; the first training feature information is a DTI-ALPS feature information; the first label vector includes first, second and third label probabilities, and the first, second and third label probabilities correspond one-to-one to the first, second and third categories of objects; only one of the first, second and third label probabilities is 1, and the other two probabilities are 0.

[0105] Step 2: patients with mild cognitive impairment not caused by Alzheimer's disease, patients with vascular cognitive impairment and healthy people are regarded as the first category of objects, patients with mild cognitive impairment caused by Alzheimer's disease are regarded as the second category of objects, and patients with Alzheimer's disease are regarded as the third category of objects; and based on the first, second and third category objects, a three-category prediction model for predicting the risk of Alzheimer's disease is constructed as the Alzheimer's disease risk prediction model corresponding to the specified age group.

[0106] Here, the Alzheimer's disease risk prediction model of an embodiment of the present invention is used to perform three-category prediction based on the DTI-ALPS feature information input by the model and output the corresponding classification probability vector; wherein the classification probability vector includes first, second and third classification probabilities, the first, second and third classification probabilities correspond one-to-one to the first, second and third categories of objects, and the sum of the first, second and third classification probabilities is 1.

[0107] like Figure 2 As shown in the module diagram of the Alzheimer's disease risk prediction model provided in Example 1 of the present invention, the model input end of the Alzheimer's disease risk prediction model is used to receive DTI-ALPS feature information, and the model output end is used to output a classification probability vector.

[0108] like Figure 2 As shown, the model components of the Alzheimer's disease risk prediction model include: a first feature extraction layer, a channel attention layer, a second feature extraction layer, a linear mapping layer, and a prediction output layer.

[0109] like Figure 2 As shown, the connection relationship of the model components of the Alzheimer's disease risk prediction model is as follows: the input end of the first feature extraction layer is connected to the model input end, and the output end is connected to the input end of the channel attention layer; the output end of the channel attention layer is connected to the input end of the second feature extraction layer; the output end of the second feature extraction layer is connected to the input end of the linear mapping layer; the output end of the linear mapping layer is connected to the input end of the prediction output layer; and the output end of the prediction output layer is connected to the model output end.

[0110] The functions of the model components of the Alzheimer's disease risk prediction model are shown below.

[0111] 1) First feature extraction layer:

[0112] The first feature extraction layer of the embodiment of the present invention is used to treat the DTI-ALPS feature information as an input vector X with a vector length of 7; and perform feature encoding processing on the input vector X through a fully connected layer to obtain the corresponding feature vector H1 and send it to the channel attention layer.

[0113] Here, the encoding method of the feature vector H1 is:

[0114] H1=ReLU(W1X+b1);

[0115] Among them, ReLU() is the ReLU activation function, W1 and b1 are the fully connected layer weight matrix parameters and offset vector parameters of the first feature extraction layer; the shape of the input vector X can be regarded as 7×1; the shape of the weight matrix parameter W1 is C1×7, the shape of the offset vector parameter b1 is C1×1, C1 is the preset first feature dimension, C1>7, for example, C1=64; the shape of the feature vector H1 is C1×1, which consists of C1 feature data Composition, 1≤index i≤C1.

[0116] 2) Channel Attention Layer:

[0117] The channel attention layer of an embodiment of the present invention is used to perform global average pooling processing on the feature vector H1 to obtain the corresponding pooling scalar z; and generate a corresponding channel attention weight vector A based on the pooling scalar z through two fully connected layers; and based on the channel attention weight vector A, the feature vector H1 is weighted to obtain the corresponding feature vector H2 and sent to the second feature extraction layer.

[0118] Here, the global average pooling method of the pooling scalar z is:

[0119] The encoding method of the channel attention weight vector A is:

[0120] A=ReLU(W3Sigmoid(W2z+b2)+b3);

[0121] Among them, Sigmoid() is the Sigmoid activation function; W2 and b2 are the weight matrix parameters and offset vector parameters of the first fully connected layer of the channel attention layer, and W3 and b3 are the weight matrix parameters and offset vector parameters of the second fully connected layer of the channel attention layer; the shape of the weight matrix parameter W2 is C2×1, the shape of the offset vector parameter b2 is C2×1, C2 is the preset second feature dimension, C2<C1, for example, C2=8; the shape of the weight matrix parameter W3 is C3×C2, the shape of the offset vector parameter b3 is C3×1, C3 is the preset third feature dimension, C3=C1; the output feature shape of Sigmoid(W2z+b2) is C2×1; the shape of the channel attention weight vector A is C3×1=C1×1;

[0122] The encoding method of the eigenvector H2 is: H2 = H1⊙A; ⊙ is the Hadamard product.

[0123] 3) Second feature extraction layer:

[0124] The second feature extraction layer in the embodiment of the present invention is used to perform feature encoding processing on the feature vector H2 through a fully connected layer to obtain a corresponding feature vector H3 and send it to the linear mapping layer.

[0125] Here, the encoding of the feature vector H3 is:

[0126] H3=ReLU(W4H2+b4);

[0127] Among them, W4 and b4 are the fully connected layer weight matrix parameters and offset vector parameters of the second feature extraction layer; the shape of the weight matrix parameter W4 is C4×C3, and the shape of the offset vector parameter b4 is C4×1, where C4 is the preset fourth feature dimension, C4<C3, for example, C4=32; the shape of the feature vector H3 is C4×1.

[0128] 4) Linear mapping layer:

[0129] The linear mapping layer in the embodiment of the present invention is used to perform feature mapping processing on the feature vector H3 through a linear layer to obtain a corresponding feature vector H4 and send it to the prediction output layer.

[0130] Here, the encoding of the feature vector H4 is:

[0131] H4=W5H3+b5;

[0132] Among them, W5 and b5 are the linear layer weight matrix parameters and offset vector parameters of the linear mapping layer; the shape of the weight matrix parameter W5 is C5×C4, the shape of the offset vector parameter b5 is C5×1, C5 is the preset fifth feature dimension and C5=3; the shape of the feature vector H4 is C5×1.

[0133] 5) Prediction output layer:

[0134] The prediction output layer of the embodiment of the present invention is used to use the Softmax function to perform three-category probability calculation based on the feature vector H4 to obtain the corresponding first, second and third classification probabilities; and the corresponding classification probability vector is composed of the obtained first, second and third classification probabilities and output.

[0135] Step 3: training an Alzheimer's disease risk prediction model based on the first data set;

[0136] Specifically, the steps include: step 31, initializing the first round counter to 1; and forming a corresponding model parameter set by the weight matrix parameters W1, W3, W3, W4, W5 and the offset vector parameters b1, b2, b3, b4, b5 of the Alzheimer's disease risk prediction model;

[0137] Step 32: randomly split the first data set into two sub-data sets based on a preset first split ratio and record them as a corresponding current training set and a current evaluation set;

[0138] The first split ratio is a preset ratio parameter, such as 8:2; the current training set and the current evaluation set both consist of a plurality of first data records; the ratio of the total number of records in the current training set to the total number of records in the current evaluation set satisfies the first split ratio;

[0139] Step 33: record the total number of records in the current training set as the current total number of records N tr ; And count the total number of first data records corresponding to the first, second, and third class objects in the current training set to obtain the corresponding first class total number N1, second class total number N2, and third class total number N3; and set the corresponding first class sample weight w1, second class sample weight w2, and third class sample weight w3 based on the first, second, and third class total numbers N1, N2, and N3;

[0140] Here, the weights w1, w2, and w3 of the first, second, and third class samples are set as follows:

[0141]

[0142]

[0143] Step 34, perform a round of traversal on all first data records of the current training set; and in this round of traversal, the first data record currently traversed is used as the corresponding current training record R j , 1≤indexj≤N tr ; and record the current training R j The first training feature information is used as the current DTI-ALPS feature information to input into the Alzheimer's disease risk prediction model for prediction, and the classification probability vector output by this prediction is used as the corresponding prediction vector Y j ; and the predicted vector Y j The first, second and third classification probabilities are recorded as the corresponding probabilities y j,k , 1≤index k≤3; and record the current training record R j The first label vector is the corresponding label vector And the label vector The first, second and third label probabilities are recorded as the corresponding probabilities And by the current prediction vector Y j and label vector Form a corresponding first prediction-label pair

[0144] Step 35, the obtained N tr First prediction-label pair And the weights w1, w2, and w3 of the first, second, and third class samples are brought into the preset first model loss function L M Calculate and obtain the corresponding first loss value;

[0145] Among them, the first model loss function L M for:

[0146]

[0147] Step 36: Identify whether the first loss value satisfies a preset first loss value range. If so, proceed to step 37. If not, perform a round of parameter modulation on the model parameter set based on a preset first model optimizer. Return to step 33 to continue training after this round of modulation is completed.

[0148] Here, the first loss value range is a preset numerical range; the first model optimizer includes at least an Adam optimizer and an SGD optimizer;

[0149] Step 37: Perform a round of traversal on all first data records of the current evaluation set; and during this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the first training feature information of the current evaluation record as the current DTI-ALPS feature information into the Alzheimer's disease risk prediction model for prediction, and use the classification probability vector output by this prediction as the corresponding current prediction vector; and form a corresponding second prediction-label pair by using the current prediction vector and the first label vector of the current evaluation record; and at the end of this round of traversal, calculate the accuracy, precision, recall, and F1 score based on all the obtained second prediction-label pairs to obtain the corresponding current accuracy, current precision, current recall, and current F1 score;

[0150] Step 38: Identify whether the current accuracy, current precision, current recall, and current F1 score all meet their corresponding first accuracy range, first precision range, first recall range, and first F1 score range; if so, go to step 39; if not, return to step 33 and continue training;

[0151] Here, the first accuracy range, the first precision range, the first recall range, and the first F1 score range are four preset value ranges;

[0152] Step 39, identify whether the first round counter is greater than the preset training round threshold; if not, add 1 to the first round counter and return to step 32 to continue the next round of training; if so, stop training and confirm the end of model training.

[0153] Here, the training round threshold is a preset positive integer.

[0154] Step 4: After the model training is completed, the DTI-ALPS feature information of any subject in the specified age group input by the user is input into the Alzheimer's disease risk prediction model for prediction to obtain the corresponding classification probability vector and feedback it to the current user.

[0155] It should be noted that through steps 1-4, a version of the Alzheimer's disease risk prediction model for a specified age group can be obtained; if you want to customize the prediction model version for multiple age groups, you only need to take each customized age group as a specified age group and execute steps 1-4 under the conditions of the specified age group to obtain a version of the Alzheimer's disease risk prediction model corresponding to the current age group.

[0156] Figure 3 This is a module structure diagram of a processing device for an Alzheimer's disease risk prediction model provided in the second embodiment of the present invention. The device is a terminal device or server that implements the aforementioned method embodiment, or can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiment. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 3 As shown, the device includes: a data set preparation module 201, a model construction module 202, a model training module 203 and a model application module 204.

[0157] The data set preparation module 201 is used to obtain DTI-ALPS feature information of dementia people and healthy people in a specified age group through big data collection and volunteer recruitment, and thereby construct a corresponding model data set recorded as the first data set; the dementia population includes multiple dementia patients, and the dementia patients include at least patients with mild cognitive impairment of non-Alzheimer's disease origin, patients with vascular cognitive impairment, patients with mild cognitive impairment of Alzheimer's disease origin, and patients with Alzheimer's disease; the healthy population is composed of multiple healthy people who do not suffer from any dementia.

[0158] The model construction module 202 is used to treat patients with mild cognitive impairment of non-Alzheimer's disease origin, patients with vascular cognitive impairment and healthy people as the first category of objects, patients with mild cognitive impairment of Alzheimer's disease origin as the second category of objects, and patients with Alzheimer's disease as the third category of objects; and based on the first, second and third category objects, a three-category prediction model for predicting the risk of Alzheimer's disease is constructed as the Alzheimer's disease risk prediction model corresponding to the specified age group; the Alzheimer's disease risk prediction model is used to perform three-category prediction based on the DTI-ALPS feature information input by the model and output the corresponding classification probability vector; the classification probability vector includes the first, second and third classification probabilities, the first, second and third classification probabilities correspond one-to-one to the first, second and third category objects, and the sum of the first, second and third classification probabilities is 1.

[0159] The model training module 203 is used to train the Alzheimer's disease risk prediction model based on the first data set.

[0160] The model application module 204 is used to input the DTI-ALPS feature information of any subject in the specified age group input by the user into the Alzheimer's disease risk prediction model after the model training is completed, to obtain the corresponding classification probability vector and feed it back to the current user.

[0161] An embodiment of the present invention provides a processing device for an Alzheimer's disease risk prediction model, which can execute the method steps in the above method embodiment. Its implementation principles and technical effects are similar and will not be repeated here.

[0162] It should be noted that it should be understood that the division of the various modules of the above device is merely a division of logical functions. In actual implementation, they can be fully or partially integrated into a physical entity, or they can be physically separated. Moreover, these modules can all be implemented in the form of software called by a processing element; or they can all be implemented in the form of hardware; or some modules can be implemented in the form of software called by a processing element, and some modules can be implemented in the form of hardware. For example, the data set preparation module can be a separately established processing element, or it can be integrated into a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called by a processing element of the above device to perform the functions of the above-mentioned determination module. The implementation of other modules is similar. In addition, these modules can all or partly be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. During implementation, each step of the above method or each of the above modules can be completed by hardware integrated logic circuits in the processor element or instructions in the form of software.

[0163] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code through a processing element, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0164] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the above method embodiments are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) means. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, hard disk, tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0165] Figure 4 This is a schematic diagram of the structure of an electronic device provided in the third embodiment of the present invention. The electronic device can be a terminal device or server that implements the method of the aforementioned embodiment, or it can be a terminal device or server that implements the method of the aforementioned embodiment connected to the aforementioned terminal device or server. Figure 4As shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver 303's transceiver actions. Various instructions may be stored in the memory 302 for completing various processing functions and implementing the processing steps described in the aforementioned embodiment method. Preferably, the electronic device involved in the embodiment of the present invention further includes: a power supply 304, a system bus 305, and a communication port 306. The system bus 305 is used to realize communication connections between components. The above-mentioned communication port 306 is used for connecting and communicating between the electronic device and other peripherals.

[0166] exist Figure 4 The system bus 305 mentioned in the figure can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM) and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk storage.

[0167] The above-mentioned processors can be general-purpose processors, including central processing units (CPUs), network processors (NPs), graphics processing units (GPUs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0168] It should be noted that an embodiment of the present invention further provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, it enables the computer to execute the methods and processing procedures provided in the above embodiments.

[0169] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for an Alzheimer's disease risk prediction model. As can be seen from the above content, the embodiment of the present invention first obtains the DTI-ALPS feature information of dementia population (including non-AD-derived MCI patients, VCI patients, AD-derived MCI patients, AD patients) and healthy people in a specified age group through big data collection and volunteer recruitment, and thereby constructs a corresponding model data set, i.e., the first data set; then, non-AD-derived MCI patients, VCI patients and healthy people are regarded as non-AD objects (i.e., first-class objects), AD-derived MCI patients are regarded as second-class objects, and AD patients are regarded as third-class objects; then, a three-class prediction model for predicting the risk of Alzheimer's disease is constructed as an Alzheimer's disease risk prediction model; then, the model is trained based on the first data set; and after the training, the prediction model is used to perform a three-class prediction of the AD risk of the subject according to the DTI-ALPS feature information (non-AD class, AD-derived MCI class, AD class). The embodiments of the present invention improve the prediction accuracy and stability of AD risk prediction through the learning ability of the artificial intelligence model and model training with large amounts of data; based on the embodiments of the present invention, model data sets of different age groups can be collected to train adapted model versions for multiple age groups, thereby improving the prediction flexibility of AD risk prediction.

[0170] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0171] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing an Alzheimer's disease risk prediction model, characterized in that: The method comprises: The DTI-ALPS feature information of a dementia population and a healthy population of a specified age group is obtained by big data collection and volunteer recruitment, and a corresponding model dataset is constructed therefrom, which is recorded as a first dataset; the dementia population includes a plurality of dementia patients, and the dementia patients include at least patients with mild cognitive impairment not caused by Alzheimer's disease, patients with vascular cognitive impairment, patients with mild cognitive impairment caused by Alzheimer's disease, and patients with Alzheimer's disease; the healthy population consists of a plurality of healthy people who do not suffer from any dementia; The patients with mild cognitive impairment not caused by Alzheimer's disease, the patients with vascular cognitive impairment and the healthy subjects are taken as the first category of objects, the patients with mild cognitive impairment caused by Alzheimer's disease are taken as the second category of objects, and the patients with Alzheimer's disease are taken as the third category of objects; and based on the first, second and third category objects, a three-category prediction model for predicting the risk of Alzheimer's disease is constructed as the Alzheimer's disease risk prediction model corresponding to the specified age group; the Alzheimer's disease risk prediction model is used to perform three-category prediction based on the DTI-ALPS feature information input by the model and output a corresponding classification probability vector; the classification probability vector includes first, second and third classification probabilities, the first, second and third classification probabilities correspond one-to-one to the first, second and third category objects, and the sum of the first, second and third classification probabilities is 1; Training the Alzheimer's disease risk prediction model based on the first data set; After the model training is completed, the DTI-ALPS feature information of any subject in the specified age group input by the user is input into the Alzheimer's disease risk prediction model for prediction to obtain the corresponding classification probability vector and feedback it to the current user.

2. The method for processing the Alzheimer's disease risk prediction model according to claim 1, characterized in that: The DTI-ALPS feature information is composed of multiple ALPS features calculated based on the DTI-ALPS technology; the ALPS feature at least includes feature D xproj , Feature D yproj , Feature D zproj , Feature D xassoc , Feature D yassoc , Feature D zassoc and ALPS index characteristics; The calculation principle of the ALPS feature is: Based on the principle of DTI-ALPS technology, a pair of symmetrical projection fiber regions of interest and a pair of symmetrical association fiber regions of interest were set at the level of the left and right ventricles in the DTI images of the subjects' brains. Based on the principle of DTI-ALPS technology, the diffusivity of the left brain projection fiber area of ​​interest along the x, y, and z axes is calculated to obtain the corresponding left brain projection fiber x, y, and z axial diffusivity. The diffusion rates of the right brain projection fiber area of ​​interest along the x, y, and z axes are calculated respectively to obtain the corresponding right brain projection fiber x, y, and z axial diffusion rates. Based on the DTI-ALPS technology principle, the diffusivity of the left brain joint fiber area of ​​interest along the x, y, and z axes is calculated to obtain the corresponding left brain joint fiber x, y, and z axial diffusivity. The diffusion rates of the right brain joint fiber region of interest along the x, y, and z axes are calculated respectively to obtain the corresponding right brain joint fiber x, y, and z axial diffusion rates. According to the DTI-ALPS index calculation principle, the diffusion rate Calculate the corresponding left brain ALPS index, based on the diffusion rate Calculate the corresponding right brain ALPS index: mean() is the mean value calculation function; And the diffusion rate The average value of the corresponding feature D xproj ; And the diffusion rate The average value of the corresponding feature D yproj ; And the diffusion rate The average value of the corresponding feature D zproj ; And the diffusion rate The average value of the corresponding feature D xassoc ; And the diffusion rate The average value of the corresponding feature D yassoc ; And the diffusion rate The average value of the corresponding feature D zassoc ; The average value of the left and right brain ALPS indices is used as the corresponding ALPS index feature.

3. The method for processing the Alzheimer's disease risk prediction model according to claim 2, characterized in that: The model input end of the Alzheimer's disease risk prediction model is used to receive the DTI-ALPS feature information, and the model output end is used to output the classification probability vector; The Alzheimer's disease risk prediction model includes a first feature extraction layer, a channel attention layer, a second feature extraction layer, a linear mapping layer and a prediction output layer; The input end of the first feature extraction layer is connected to the input end of the model, and the output end is connected to the input end of the channel attention layer; the output end of the channel attention layer is connected to the input end of the second feature extraction layer; the output end of the second feature extraction layer is connected to the input end of the linear mapping layer; the output end of the linear mapping layer is connected to the input end of the prediction output layer; the output end of the prediction output layer is connected to the output end of the model; The first feature extraction layer is used to take the DTI-ALPS feature information as an input vector X with a vector length of 7; and perform feature encoding processing on the input vector X through a fully connected layer to obtain a corresponding feature vector H1 and send it to the channel attention layer; The encoding method of the feature vector H1 is: H1=ReLU(W1X+b1), ReLU() is the ReLU activation function, W1 and b1 are the fully connected layer weight matrix parameters and offset vector parameters of the first feature extraction layer; the shape of the weight matrix parameter W1 is C1×7, the shape of the offset vector parameter b1 is C1×1, C1 is the preset first feature dimension, C1>7; the shape of the feature vector H1 is C1×1, which is composed of C1 feature data Composition, 1≤indexi≤C1; The channel attention layer is used to perform global average pooling on the feature vector H1 to obtain a corresponding pooling scalar z; and generate a corresponding channel attention weight vector A based on the pooling scalar z through two fully connected layers; and based on the channel attention weight vector A, the feature vector H1 is weighted to obtain a corresponding feature vector H2 and sent to the second feature extraction layer; The global average pooling method of the pooling scalar z is: The encoding method of the channel attention weight vector A is: A=ReLU(W3Sigmoid(W2z+b2)+b3); Sigmoid() is a Sigmoid activation function, W2 and b2 are the weight matrix parameters and offset vector parameters of the first fully connected layer of the channel attention layer, W3 and b3 are the weight matrix parameters and offset vector parameters of the second fully connected layer of the channel attention layer; the shape of the weight matrix parameter W2 is C2×1, the shape of the offset vector parameter b2 is C2×1, C2 is the preset second feature dimension, C2<C1; the shape of the weight matrix parameter W3 is C3×C2, the shape of the offset vector parameter b3 is C3×1, C3 is the preset third feature dimension, C3=C1; the output feature shape of Sigmoid(W2z+b2) is C2×1; the shape of the channel attention weight vector A is C3×1=C1×1; The encoding method of the feature vector H2 is: H2=H1⊙A; ⊙ is Hadamard; The second feature extraction layer is used to perform feature encoding processing on the feature vector H2 through a fully connected layer to obtain a corresponding feature vector H3 and send it to the linear mapping layer; The encoding method of the feature vector H3 is: H3=ReLU(W4H2+b4), W4 and b4 are the fully connected layer weight matrix parameters and offset vector parameters of the second feature extraction layer; the shape of the weight matrix parameter W4 is C4×C3, and the shape of the offset vector parameter b4 is C4×1, where C4 is the preset fourth feature dimension, C4<C3; the shape of the feature vector H3 is C4×1; The linear mapping layer is used to perform feature mapping processing on the feature vector H3 through a linear layer to obtain a corresponding feature vector H4 and send it to the prediction output layer; The encoding method of the feature vector H4 is: H4=W5H3+b5, W5 and b5 are the linear layer weight matrix parameters and offset vector parameters of the linear mapping layer; the shape of the weight matrix parameter W5 is C5×C4, the shape of the offset vector parameter b5 is C5×1, C5 is the preset fifth feature dimension, C5=3; the shape of the feature vector H4 is C5×1; The prediction output layer is used to use the Softmax function to perform three-category probability calculation based on the feature vector H4 to obtain the corresponding first, second and third classification probabilities; and the corresponding classification probability vector is composed of the obtained first, second and third classification probabilities and output.

4. The method for processing the Alzheimer's disease risk prediction model according to claim 2, characterized in that: The first data set includes multiple first data records; the first data record includes first training feature information and a first label vector; the first training feature information is a DTI-ALPS feature information; the first label vector includes first, second and third label probabilities, and the first, second and third label probabilities correspond one-to-one to the first, second and third categories of objects; only one of the first, second and third label probabilities is 1, and the other two probabilities are 0.

5. The method for processing the Alzheimer's disease risk prediction model according to claim 4, characterized in that: The DTI-ALPS feature information of dementia patients and healthy people in a specified age group is obtained through big data collection and volunteer recruitment, and the corresponding model dataset is constructed based on this information, which is recorded as the first dataset, specifically including: Data is collected from the brain DTI images of the patients with mild cognitive impairment not caused by Alzheimer's disease, the patients with vascular cognitive impairment, the patients with mild cognitive impairment caused by Alzheimer's disease, the patients with Alzheimer's disease, and the healthy people in the specified age group through multiple preset big data collection channels; and the patients with mild cognitive impairment not caused by Alzheimer's disease, the patients with vascular cognitive impairment, the patients with mild cognitive impairment caused by Alzheimer's disease, the patients with Alzheimer's disease, and the healthy people in the specified age group are recruited through volunteer recruitment, and brain magnetic resonance imaging examination is performed on each recruited subject to obtain corresponding brain DTI images ; and construct a first image set based on all the obtained brain DTI images and their corresponding patient or healthy person types; wherein the big data acquisition channel includes at least a public medical image data set, a medical institution database that is public or authorized for data use, and a medical image sharing platform that is public or authorized for data use; the first image set includes multiple first image records; the first image record includes a first image and a first object type; the first image is a brain DTI image; the first object type includes a mild cognitive impairment type not originating from Alzheimer's disease, a vascular cognitive impairment type, a mild cognitive impairment type originating from Alzheimer's disease, an Alzheimer's disease type, and a healthy type; and taking each of the first image records of the first image set as the corresponding current record; and taking the first image and the first object type of the current record as the corresponding current image and current type; and performing ALPS feature calculation on the current image based on the calculation principle of the ALPS feature to obtain the corresponding DTI-ALPS feature information; and taking the DTI-ALPS feature information obtained this time as the corresponding first training feature information; and identifying the current type; if the current type is the mild cognitive impairment type originating from non-Alzheimer's disease, the vascular cognitive impairment type or the healthy type, setting the corresponding first, second and third label probabilities to 1, 0, 0; if the current type is the mild cognitive impairment type originating from Alzheimer's disease, setting the corresponding first, second and third label probabilities to 0, 1, 0; if the current type is the Alzheimer's disease type, setting the corresponding first, second and third label probabilities to 0, 0, 1; and forming a corresponding first label vector from the first, second and third label probabilities obtained this time; and forming a corresponding first data record from the first training feature information and the first label vector corresponding to the current record; All the obtained first data records form the corresponding first data set.

6. The method for processing the Alzheimer's disease risk prediction model according to claim 4, characterized in that: The training of the Alzheimer's disease risk prediction model based on the first data set specifically includes: Step 61, initializing the first round counter to 1; and forming a corresponding model parameter set by the weight matrix parameters W1, W3, W3, W4, W5 and the offset vector parameters b1, b2, b3, b4, b5 of the Alzheimer's disease risk prediction model; Step 62: randomly split the first data set into two sub-data sets based on a preset first split ratio and record them as a corresponding current training set and a current evaluation set; Wherein, the current training set and the current evaluation set are both composed of a plurality of the first data records; the ratio of the total number of records in the current training set and the current evaluation set satisfies the first segmentation ratio; Step 63: record the total number of records in the current training set as the current total number of records N tr ; and statistically calculate the total number of the first data records corresponding to the first, second, and third class objects in the current training set to obtain the corresponding first class total number N1, second class total number N2, and third class total number N3; and set the corresponding first class sample weight w1, second class sample weight w2, and third class sample weight w3 based on the first, second, and third class total numbers N1, N2, and N3; The weights w1, w2, and w3 of the first, second, and third class samples are set as follows: Step 64, perform a round of traversal on all the first data records of the current training set; and in this round of traversal, use the first data record currently traversed as the corresponding current training record R j , 1≤indexj≤N tr ; and record the current training R j The first training feature information is used as the current DTI-ALPS feature information to input into the Alzheimer's disease risk prediction model for prediction, and the classification probability vector output by this prediction is used as the corresponding prediction vector Y j ; and the prediction vector Y j The first, second and third classification probabilities are recorded as the corresponding probabilities y j,k , 1≤index k≤3; and record the current training R j The first label vector is used as the corresponding label vector And the label vector The first, second and third label probabilities are recorded as the corresponding probabilities And the current prediction vector Y j and the label vector Form a corresponding first prediction-label pair Step 65, the obtained N tr The first prediction-label pair And the weights w1, w2, and w3 of the first, second, and third class samples are brought into the preset first model loss function L M Calculate and obtain the corresponding first loss value; Among them, the first model loss function L M for: Step 66: Identify whether the first loss value satisfies a preset first loss value range. If so, proceed to step 67. If not, perform a round of parameter modulation on the model parameter set based on a preset first model optimizer. Return to step 63 to continue training after this round of modulation is completed. Wherein, the first model optimizer includes at least an Adam optimizer and an SGD optimizer; Step 67, perform a round of traversal on all the first data records of the current evaluation set; and during this round of traversal, use the first data record currently traversed as the corresponding current evaluation record; and input the first training feature information of the current evaluation record as the current DTI-ALPS feature information into the Alzheimer's disease risk prediction model for prediction, and use the classification probability vector output by this prediction as the corresponding current prediction vector; and form a corresponding second prediction-label pair by the current prediction vector and the first label vector of the current evaluation record; and at the end of this round of traversal, calculate the accuracy, precision, recall and F1 score based on all the obtained second prediction-label pairs to obtain the corresponding current accuracy, current precision, current recall and current F1 score; Step 68: Identify whether the current accuracy, the current precision, the current recall, and the current F1 score all meet their corresponding first accuracy range, first precision range, first recall range, and first F1 score range; if so, proceed to step 69; if not, return to step 63 to continue training; In step 69, it is determined whether the first round counter is greater than a preset training round threshold; if not, the first round counter is incremented by 1 and the process returns to step 62 to continue the next round of training; if so, the training is stopped and the model training is confirmed to be complete.

7. A device for executing the method for processing the Alzheimer's disease risk prediction model according to any one of claims 1 to 6, characterized in that: The device includes: a data set preparation module, a model construction module, a model training module and a model application module; The data set preparation module is used to obtain DTI-ALPS feature information of a dementia population and a healthy population of a specified age group through big data collection and volunteer recruitment, and thereby construct a corresponding model data set recorded as a first data set; the dementia population includes a plurality of dementia patients, and the dementia patients include at least patients with mild cognitive impairment of a non-Alzheimer's disease origin, patients with vascular cognitive impairment, patients with mild cognitive impairment of Alzheimer's disease origin, and patients with Alzheimer's disease; the healthy population consists of a plurality of healthy people who do not suffer from any dementia; The model construction module is used to take the patients with mild cognitive impairment of non-Alzheimer's disease origin, the patients with vascular cognitive impairment and the healthy subjects as the first category of objects, the patients with mild cognitive impairment of Alzheimer's disease origin as the second category of objects, and the patients with Alzheimer's disease as the third category of objects; and construct a three-category prediction model for predicting the risk of Alzheimer's disease based on the first, second and third category objects as the Alzheimer's disease risk prediction model corresponding to the specified age group; the Alzheimer's disease risk prediction model is used to perform three-category prediction based on the DTI-ALPS feature information input by the model and output a corresponding classification probability vector; the classification probability vector includes first, second and third classification probabilities, the first, second and third classification probabilities correspond one-to-one to the first, second and third category objects, and the sum of the first, second and third classification probabilities is 1; The model training module is used to train the Alzheimer's disease risk prediction model based on the first data set; The model application module is used to input the DTI-ALPS feature information of any subject in the specified age group input by the user into the Alzheimer's disease risk prediction model after the model training is completed, to obtain the corresponding classification probability vector and feed it back to the current user.

8. An electronic device, characterized in that: include: memory, processors, and transceivers; The processor is configured to be coupled to the memory, read and execute instructions in the memory, so as to implement the method according to any one of claims 1 to 6; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer is caused to execute the method according to any one of claims 1 to 6.

Images