Early evaluation method, device and system for Alzheimer's disease

Abstract

The invention discloses an early evaluation method, device and system for Alzheimer's disease. The method comprises the following steps: acquiring magnetic resonance imaging data and physiological data of a user; inputting the magnetic resonance imaging data and the physiological data into a cross-modal feature fusion model to obtain fusion features; inputting the fusion features into an expert model to obtain a prediction result; and generating an evaluation report according to the prediction result. According to the method, the magnetic resonance imaging data and the physiological data are processed through the cross-modal feature fusion model, so that the high-level fusion features are obtained, then the expert model is used for processing the fusion features, so that the prediction result is obtained, the evaluation report is obtained, a user can obtain the evaluation report without lumbar puncture and PET imaging detection, and the user experience is improved. The compliance is improved; and the detection cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of early assessment technology for Alzheimer's disease, and particularly to a method, device, and system for early assessment of Alzheimer's disease. Background Technology

[0002] Alzheimer's disease (AD) is a progressive neurodegenerative disease whose pathological cascade can silently begin 15-20 years before the onset of clinical symptoms. Traditional cognitive scales and structural imaging, characterized by functional cognitive impairment and brain structural atrophy, can only detect patients in the middle and late stages of AD.

[0003] Currently, while lumbar puncture to detect CSF (cerebrospinal fluid) amyloid and tau proteins, or PET (positron emission tomography) imaging using 18F-labeled ligands, can detect early-stage Alzheimer's disease, lumbar puncture is highly invasive and difficult for patients to comply with; PET imaging is expensive and limited by radiopharmaceuticals, making large-scale community screening difficult. These factors contribute to poor compliance and high testing costs in the early assessment of Alzheimer's disease.

[0004] Therefore, there is still an urgent need for an early assessment method that can improve compliance and reduce testing costs. Summary of the Invention

[0005] The main objective of this invention is to propose an early assessment method, device, and system for Alzheimer's disease, addressing the shortcomings of existing early assessment methods for Alzheimer's disease, such as poor compliance, high testing costs, and poor timeliness.

[0006] To achieve the above objectives, this invention proposes an early assessment method for Alzheimer's disease, which includes: Acquire the user's magnetic resonance imaging data and physiological data; The magnetic resonance imaging data and the physiological data are input into a cross-modal feature fusion model to obtain fused features; The fused features are input into the expert model to obtain the prediction result; An evaluation report is generated based on the prediction results.

[0007] In some embodiments, acquiring the user's physiological data includes: Detect whether the user is using a VR device and whether the user is wearing an EEG wearable device; If the user is using the VR device and the user is wearing the EEG wearable device, a VR cognitive task is generated and the VR cognitive task is sent to the VR device; Control the VR device to run the VR cognitive task, and control the wearable EEG device to collect the user's EEG data and eye movement data; The electroencephalogram (EEG) data and the eye movement data are identified as the physiological data.

[0008] In some embodiments, inputting the magnetic resonance imaging data and the physiological data into a cross-modal feature fusion model to obtain fused features includes: Feature extraction is performed on the magnetic resonance imaging data and the physiological data respectively to obtain magnetic resonance imaging features and physiological features; The magnetic resonance imaging features and the physiological features are input into the cross-modal feature fusion model; Obtain the fused features output by the cross-modal feature fusion model.

[0009] In some embodiments, the expert model includes a magnetic resonance imaging expert model, a physiological expert model, and a magnetic resonance imaging-physiological fusion expert model; the step of inputting the fusion features into the expert model to obtain the prediction result includes: The fusion features are respectively input into the magnetic resonance imaging expert model, the physiological expert model, and the magnetic resonance imaging-physiology fusion expert model; Obtain the first prediction result output by the magnetic resonance imaging expert model, the second prediction result output by the physiological expert model, and the third prediction result output by the magnetic resonance imaging-physiological fusion expert model; The first prediction result, the second prediction result, and the third prediction result are processed according to a preset voting algorithm to obtain the prediction result.

[0010] In some embodiments, the preset voting algorithm is a weighted voting algorithm; the step of processing the first prediction result, the second prediction result, and the third prediction result according to the preset voting algorithm to obtain the prediction result includes: Obtain the first weight of the magnetic resonance imaging expert model, the second weight of the physiological expert model, and the third weight of the magnetic resonance imaging-physiological fusion expert model, wherein the sum of the first weight, the second weight, and the third weight is 1; The product of the first prediction result and the first weight is calculated as the first product, the product of the second prediction result and the second weight is calculated as the second product, and the product of the third prediction result and the third weight is calculated as the third product. The sum of the first product, the second product, and the third product is calculated as the prediction result.

[0011] In some embodiments, generating an evaluation report based on the prediction results includes: Determine whether the prediction result is greater than a preset threshold; If the prediction result is less than or equal to the preset threshold, an assessment report containing currently undetected abnormal risks is generated. If the predicted result is greater than the preset threshold, an evaluation report is generated that includes a suggestion to increase the target inspection.

[0012] The present invention also proposes an early assessment device for Alzheimer's disease, the early assessment device for Alzheimer's disease comprising: The acquisition unit is used to acquire the user's magnetic resonance imaging data and physiological data; The feature fusion unit is used to input the magnetic resonance imaging data and the physiological data into the cross-modal feature fusion model to obtain fused features; The prediction unit is used to input the fused features into the expert model to obtain the prediction result; A generation unit is used to generate an evaluation report based on the prediction results.

[0013] In some embodiments, the acquisition unit is specifically used for: Detect whether the user is using a VR device and whether the user is wearing an EEG wearable device; If the user is using the VR device and the user is wearing the EEG wearable device, a VR cognitive task is generated and the VR cognitive task is sent to the VR device; Control the VR device to run the VR cognitive task, and control the wearable EEG device to collect the user's EEG data and eye movement data; The electroencephalogram (EEG) data and the eye movement data are identified as the physiological data.

[0014] In some embodiments, the feature fusion unit is specifically used for: Feature extraction is performed on the magnetic resonance imaging data and the physiological data respectively to obtain magnetic resonance imaging features and physiological features; The magnetic resonance imaging features and the physiological features are input into the cross-modal feature fusion model; Obtain the fused features output by the cross-modal feature fusion model.

[0015] The present invention also proposes an early assessment system for Alzheimer's disease, which includes a cloud server, a VR device, and a wearable EEG device, wherein the cloud server includes the early assessment device for Alzheimer's disease described in any one of the above-mentioned methods.

[0016] This invention acquires the user's magnetic resonance imaging (MRI) data and physiological data; inputs the MRI and physiological data into a cross-modal feature fusion model to obtain fused features; inputs the fused features into an expert model to obtain prediction results; generates an evaluation report based on the prediction results; processes the MRI and physiological data through the cross-modal feature fusion model to obtain high-level fused features, and then uses an expert model to process the fused features to obtain prediction results, thereby generating an evaluation report. This allows users to obtain an evaluation report without undergoing lumbar puncture or PET imaging, improving compliance and reducing testing costs. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the early assessment method for Alzheimer's disease in an embodiment of the present invention. Figure 2 This is another flowchart illustrating the early assessment method for Alzheimer's disease in an embodiment of the present invention; Figure 3 This is another flowchart illustrating the early assessment method for Alzheimer's disease in an embodiment of the present invention; Figure 4 This is another flowchart illustrating the early assessment method for Alzheimer's disease in an embodiment of the present invention; Figure 5 This is another flowchart illustrating the early assessment method for Alzheimer's disease in an embodiment of the present invention; Figure 6 This is another flowchart illustrating the early assessment method for Alzheimer's disease in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of the early assessment device for Alzheimer's disease according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of the early assessment system for Alzheimer's disease according to an embodiment of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0021] It should also be noted that when a component is described as "fixed to" or "set on" another component, it can be directly on the other component or there may be an intervening component present. When a component is described as "connected to" another component, it can be directly connected to the other component or there may be an intervening component present.

[0022] Furthermore, the use of terms such as "first" and "second" in this invention is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0023] To achieve the above objectives, this invention proposes an early assessment method for Alzheimer's disease, which includes: Step S110: Acquire the user's magnetic resonance imaging data and physiological data; Step S120: Input magnetic resonance imaging data and physiological data into the cross-modal feature fusion model to obtain fused features; Step S130: Input the fused features into the expert model to obtain the prediction result; Step S140: Generate an evaluation report based on the prediction results.

[0024] In this embodiment, refer to Figure 1 , Figure 7 and Figure 8Early assessment methods for Alzheimer's disease can be applied to early assessment systems for Alzheimer's disease. These systems include cloud servers, VR (Virtual Reality) devices, and wearable EEG devices; cloud servers may include... Figure 7 An early assessment device for Alzheimer's disease. The cloud server is wirelessly connected to the VR device and the wearable EEG device. In this embodiment, the cloud server is the primary entity executing the method steps.

[0025] It is understandable that wearable EEG devices can be EEG headbands, smart glasses, smart goggles, or other wearable devices used to detect users' physiological data. For example, a wearable EEG device may include an adjustable headband, EEG sensors, eye-tracking sensors, a battery compartment, a wireless antenna, and a control module. The adjustable headband ensures the EEG headband can be comfortably and stably worn on users with different head sizes. The EEG sensors, located inside the EEG headband and in contact with the scalp, are used to collect EEG data. The eye-tracking sensors, located near the eyes, collect eye movement data, which can be used to assess the user's eye movement behavior and attention status. The battery compartment provides an independent power source for the EEG headband, supporting extended wireless operation. The wireless antenna enables wireless data transmission between the EEG headband and a cloud server or other receiving devices, ensuring real-time data synchronization. The control module contains the control circuitry and signal processing unit of the wearable EEG device.

[0026] Early assessment systems for Alzheimer's disease may also include magnetic resonance imaging (MRI). An MRI scanner is a large medical imaging device designed based on the principles of nuclear magnetic resonance (NMR). Its core function is to generate tomographic images of the body's internal structures or functions by detecting signal changes in hydrogen protons within human tissue under the influence of strong magnetic fields and radio frequency pulses. The MRI scanner is used to acquire the user's magnetic resonance imaging data. When a user needs early assessment for Alzheimer's disease, they can first use an MRI scanner to acquire magnetic resonance imaging data. After acquiring the data, the MRI scanner can send it to a cloud server. The cloud server can then access the user's magnetic resonance imaging data.

[0027] Alternatively, an early assessment system for Alzheimer's disease could also exclude MRI scanners. When a user needs an early assessment for Alzheimer's disease, they can first go to a hospital equipped with an MRI scanner to have their MRI data collected. The user can then upload the MRI data to a cloud server. The cloud server can then access the user's MRI data.

[0028] Once the cloud server obtains the user's MRI data, and the user is wearing an EEG wearable device, the cloud server can control the wearable device to collect the user's physiological data. After collecting the user's physiological data, the wearable device sends the data to the cloud server. At this point, the cloud server can access the user's physiological data.

[0029] Cloud servers can be pre-configured with cross-modal feature fusion models and expert models. For example, cross-modal feature fusion models can be pre-trained and configured on the cloud server by developers. Expert models can also be pre-trained and configured on the cloud server by developers.

[0030] After acquiring the user's magnetic resonance imaging (MRI) data and physiological data, the cloud server can perform feature fusion between the MRI and physiological data. The cloud server inputs the MRI and physiological data into a cross-modal feature fusion model to obtain fused features. For example, the cloud server can input MRI and physiological data into the cross-modal feature fusion model. Upon receiving the MRI and physiological data, the cross-modal feature fusion model first extracts features from each data separately. After extracting the features corresponding to the MRI and physiological data, it fuses these features to obtain fused features, which are then output. At this point, the cloud server obtains the fused features. The cross-modal feature fusion model can eliminate modal heterogeneity (such as the difference between the spatial features of MRI data and the temporal features of physiological data), mine complementary information from multimodal data, and generate more discriminative fused features.

[0031] After the cloud server obtains the fused features, it can input these features into an expert model to obtain a prediction result. For example, after receiving the fused features, the expert model can predict the user's risk of developing early-stage Alzheimer's disease based on these features, thus obtaining a prediction result, which is then output. The cloud server can then obtain the prediction result. This prediction result can be represented by a probability; the higher the probability, the higher the user's risk of developing early-stage Alzheimer's disease.

[0032] After receiving the prediction results, the cloud server can generate an assessment report based on those results. The early assessment system for Alzheimer's disease can also include a user-side component. After generating the assessment report, the cloud server can send it to the user's client to inform them of the early Alzheimer's disease assessment.

[0033] This embodiment acquires the user's magnetic resonance imaging (MRI) data and physiological data; inputs the MRI data and physiological data into a cross-modal feature fusion model to obtain fused features; inputs the fused features into an expert model to obtain prediction results; generates an evaluation report based on the prediction results; processes the MRI data and physiological data through the cross-modal feature fusion model to obtain high-level fused features, and then uses an expert model to process the fused features to obtain prediction results, thereby obtaining an evaluation report. This method eliminates the need for users to undergo lumbar puncture and PET imaging, improving compliance and reducing testing costs.

[0034] In some embodiments, the aforementioned acquisition of the user's physiological data includes: Step S150: Detect whether the user is using a VR device and whether the user is wearing an EEG wearable device; Step S151: If the user is using a VR device and the user is wearing an EEG wearable device, a VR cognitive task is generated and sent to the VR device. Step S152: Control the VR device to run VR cognitive tasks and control the EEG wearable device to collect the user's EEG data and eye movement data; Step S153: Identify the EEG data and eye movement data as physiological data.

[0035] In this embodiment, refer to Figure 2 When the cloud server acquires the user's physiological data in step S110, it also needs to detect whether the user is using a VR device and whether the user is wearing an EEG wearable device. When a user needs to undergo early assessment for Alzheimer's disease, they can first wear the EEG wearable device, then use the VR device, and then send an assessment start command to the cloud server through the EEG wearable device and the VR device. Specifically, the cloud server detects whether the user is using the VR device and whether the user is wearing the EEG wearable device by checking whether it receives the assessment start command from the EEG wearable device and the VR device. When the cloud server receives the assessment start command from the EEG wearable device and the VR device, it can determine that the user is using the VR device and is wearing the EEG wearable device. Similarly, if the cloud server does not receive the assessment start command from the EEG wearable device and / or the VR device, it can determine that the user is not using the VR device and / or the user is not wearing the EEG wearable device.

[0036] For example, both EEG wearable devices and VR devices can be equipped with a "Start Assessment" button. When a user clicks the "Start Assessment" button on an EEG wearable device, the device triggers an assessment start command and sends it to a cloud server. The cloud server then receives the assessment start command from the EEG wearable device. Similarly, when a user clicks the "Start Assessment" button on a VR device, the device triggers an assessment start command and sends it to a cloud server. The cloud server then receives the assessment start command from the VR device.

[0037] If a user is using a VR device and wearing an EEG wearable device, the cloud server will generate a VR cognitive task and send it to the VR device. This VR cognitive task can include spatial navigation tasks and working memory tasks.

[0038] Spatial navigation tasks, such as virtual spatial maze navigation tasks, can be used to induce subtle deficits in spatial memory and theta rhythm synchronization in early Alzheimer's disease. Theta rhythm is a rhythmic fluctuation with a specific frequency range in electroencephalogram (EEG) signals (EEG data), an important component of the electrical activity of the central nervous system, and closely related to various cognitive functions and physiological states.

[0039] Working memory tasks can be used to assess memory traces using evoked potentials (EPs). Evoked potentials (EPs) are time-locked electrical responses of the nervous system to specific stimuli (such as visual, auditory, or cognitive tasks). Memory traces, also known as engrams, are a core concept in neuroscience describing the physical or biochemical basis of memory storage and representation in the brain.

[0040] Memory traces refer to the persistent changes that occur in the nervous system (especially the brain) after we experience an event or learn information. These changes allow memories to be preserved and retrieved in the future. The formation of memory traces involves three stages: encoding (information storage), storage (information retention), and retrieval (information recall). Each stage is accompanied by changes in the electrical activity of specific brain regions. Evoked potentials, by recording these electrical signals that are "time-locked" to stimuli, can indirectly reflect the "presence," "intensity," and "retrieval efficiency" of memory traces.

[0041] After the cloud server sends the VR cognitive task to the VR device, it can control the VR device to run the VR cognitive task and control the wearable EEG device to collect the user's EEG and eye movement data. For example, after the cloud server sends the VR cognitive task to the VR device, it can control the VR device to run the VR cognitive task. When the VR device is running the VR cognitive task, the user can perform the VR cognitive task through the VR device. And during the user's performance of the VR cognitive task, the cloud server will control the wearable EEG device to continuously collect the user's EEG and eye movement data.

[0042] The cloud server controls the wearable EEG device to collect EEG and eye movement data from the user throughout the VR cognitive task. The collected EEG and eye movement data are then identified as physiological data.

[0043] In some embodiments, the aforementioned input of magnetic resonance imaging data and physiological data into a cross-modal feature fusion model to obtain fused features includes: Step S160: Feature extraction is performed on the magnetic resonance imaging data and physiological data respectively to obtain magnetic resonance imaging features and physiological features; Step S161: Input magnetic resonance imaging features and physiological features into the cross-modal feature fusion model; Step S162: Obtain the fused features output by the cross-modal feature fusion model.

[0044] In this embodiment, refer to Figure 3 When executing step S120, the cloud server can first perform feature extraction, and then input the extracted features into the cross-modal feature fusion model for fusion. The cloud server extracts features from both magnetic resonance imaging (MRI) data and physiological data separately to obtain MRI features and physiological features. For example, the cloud server extracts features from MRI data to obtain MRI features. Similarly, the cloud server extracts features from physiological data to obtain physiological features. By first extracting features from both MRI and physiological data, the cloud server reduces the workload of the cross-modal feature fusion model, allowing it to focus solely on feature fusion.

[0045] After the cloud server obtains the magnetic resonance imaging (MRI) features and physiological features, it can input these features into a cross-modal feature fusion model. Upon receiving the MRI and physiological features, the cross-modal feature fusion model sequentially performs attention fusion, cross-mapping fusion, and graph network fusion on the MRI and physiological features to obtain fused features, which are then input as input. At this point, the cloud server can obtain the fused features output by the cross-modal feature fusion model.

[0046] Attention fusion: The attention intensity of each dimension in the magnetic resonance imaging features (e.g., hippocampal volume attention intensity = 0.8) and the attention intensity of each dimension in the physiological feature vector (e.g., θ-rhythm attention intensity = 0.7) are calculated through the attention layer, and then fused by weight.

[0047] Cross-mapping fusion: Through two sub-networks, magnetic resonance imaging features are mapped to physiological feature space and physiological features are mapped to magnetic resonance imaging feature space respectively. Then, the original features and the mapped features are concatenated (e.g., magnetic resonance imaging features + mapping features of magnetic resonance imaging features to physiological features + physiological features + mapping features of physiological features to magnetic resonance imaging features), forcing intermodal information conversion.

[0048] Graph network fusion: Treating magnetic resonance imaging features and physiological features as nodes in a graph, constructing intramodal and intermodal connections (e.g., the edge weights of the hippocampal volume node and the θ-rhythm node represent their correlation), and learning global correlation features through a graph convolutional network.

[0049] By fusing magnetic resonance imaging features and physiological features through a cross-modal feature fusion model, fused features are obtained. These fused features retain the unique value of different modalities and achieve deep information synergy, providing high-quality input for accurate predictions by subsequent expert models.

[0050] In some embodiments, the expert model includes a magnetic resonance imaging (MRI) expert model, a physiological expert model, and an MRI-physiology fusion expert model; the aforementioned inputting fusion features into the expert model to obtain prediction results includes: Step S170: Input the fusion features into the magnetic resonance imaging expert model, the physiological expert model, and the magnetic resonance imaging-physiological fusion expert model, respectively; Step S171: Obtain the first prediction result output by the magnetic resonance imaging expert model, the second prediction result output by the physiological expert model, and the third prediction result output by the magnetic resonance imaging-physiological fusion expert model; Step S172: Process the first prediction result, the second prediction result, and the third prediction result according to the preset voting algorithm to obtain the prediction result.

[0051] In this embodiment, refer to Figure 4 When the cloud server executes step S130, it can input the fused features into different expert models. These expert models include a magnetic resonance imaging expert model, a physiological expert model, and a magnetic resonance imaging-physiological fusion expert model.

[0052] Magnetic resonance imaging (MRI) expert models are trained using single-modal MRI data (e.g., using a large amount of MRI data to train the identification of brain structural abnormalities) and then fine-tuned using fused features. The structure can employ a multilayer perceptron or random forest. After inputting fused features, the MRI expert model can output prediction results based on MRI information.

[0053] Physiological expert models are trained using single-modal physiological data (e.g., using large amounts of physiological data to train the identification of neurological dysfunction) and then fine-tuned using fused features. The structure can employ long short-term memory networks or gradient boosting trees. After inputting fused features, the physiological expert model can output predictions based on physiological information.

[0054] The MRI-Physiology Fusion Expert Model is trained using MRI-Physiology fusion information. Its structure employs a cross-modal attention mechanism, which forces the model to focus on intermodal correlation features (e.g., the correlation between MRI features and physiological features). After inputting fusion features into the MRI-Physiology Fusion Expert Model, it can output prediction results based on MRI-Physiology fusion information.

[0055] The cloud server can input fused features into three expert models: a magnetic resonance imaging (MRI) expert model, a physiological expert model, and an MRI-physiology fusion expert model. Upon receiving the fused features, the MRI expert model processes them to obtain a first prediction result, which is then output. Similarly, the physiological expert model processes the fused features to obtain a second prediction result, which is also output. Finally, the MRI-physiology fusion expert model processes the fused features to obtain a third prediction result, which is also output.

[0056] After the magnetic resonance imaging expert model outputs the first prediction result, the physiological expert model outputs the second prediction result, and the magnetic resonance imaging-physiological fusion expert model outputs the third prediction result, the cloud server can obtain the first prediction result output by the magnetic resonance imaging expert model, the second prediction result output by the physiological expert model, and the third prediction result output by the magnetic resonance imaging-physiological fusion expert model.

[0057] After the cloud server obtains the first, second, and third prediction results, it can process these results according to a preset voting algorithm to obtain the final prediction. The preset voting algorithm can include majority voting and weighted voting algorithms.

[0058] This embodiment leverages the advantages of different modal expert models in capturing specific information while eliminating the limitations of a single expert model through ensemble learning. This ensures that the final prediction result is not simply a matter of majority rule, but rather an intelligent weighted integration based on the performance of sub-models, which significantly improves the accuracy and reliability of the results.

[0059] In some embodiments, the preset voting algorithm is a weighted voting algorithm; the aforementioned processing of the first prediction result, the second prediction result, and the third prediction result according to the preset voting algorithm to obtain the prediction result includes: Step S180: Obtain the first weight of the magnetic resonance imaging expert model, the second weight of the physiological expert model, and the third weight of the magnetic resonance imaging-physiological fusion expert model, wherein the sum of the first weight, the second weight, and the third weight is 1; Step S181: Calculate the product of the first prediction result and the first weight as the first product, calculate the product of the second prediction result and the second weight as the second product, and calculate the product of the third prediction result and the third weight as the third product. Step S182: Calculate the sum of the first product, the second product, and the third product as the prediction result.

[0060] In this embodiment, refer to Figure 5 When executing step S172, the cloud server uses a weighted voting algorithm. The default voting algorithm is a weighted voting algorithm. The cloud server first obtains the weights of each expert model. These weights can be customized by the developers or set by the cloud server based on actual needs. The cloud server obtains the first weight of the magnetic resonance imaging expert model, the second weight of the physiological expert model, and the third weight of the magnetic resonance imaging-physiology fusion expert model. The sum of the first, second, and third weights is 1.

[0061] After obtaining the first weight of the magnetic resonance imaging expert model, the second weight of the physiological expert model, and the third weight of the magnetic resonance imaging-physiological fusion expert model, the cloud server can calculate the product of the first prediction result and the first weight as the first product, calculate the product of the second prediction result and the second weight as the second product, and calculate the product of the third prediction result and the third weight as the third product.

[0062] After obtaining the first, second, and third products, the cloud server can calculate the sum of the first, second, and third products as the prediction result.

[0063] This embodiment transforms the prediction results of various expert models into a comprehensive and robust final result. It respects the professional advantages of different expert models (such as the ability of the fusion model to capture cross-modal information) and achieves scientific integrated decision-making through quantitative weights, effectively improving the accuracy and reliability of the results.

[0064] In some embodiments, the aforementioned generation of an evaluation report based on the prediction results includes: Step S190: Determine whether the prediction result is greater than a preset threshold; Step S191: If the prediction result is less than or equal to the preset threshold, an assessment report containing currently undetected abnormal risks is generated. Step S192: If the prediction result is greater than the preset threshold, an evaluation report containing suggestions to increase target checks is generated.

[0065] In this embodiment, refer to Figure 6 When the cloud server executes step S140, it can suggest that the user undergo further examination if the predicted result is greater than a preset threshold. The predicted result can be represented by a probability; the higher the probability, the higher the user's risk of developing early-stage Alzheimer's disease. The probability range for the predicted result can be 0%-100%; the preset threshold can be 50%.

[0066] After receiving the prediction results, the cloud server can determine whether the prediction result exceeds a preset threshold. For example, it can determine whether the prediction result is greater than 50%. The early assessment system for Alzheimer's disease may also include a user-side component.

[0067] When the predicted result is less than or equal to a preset threshold, the cloud server can generate an assessment report that includes no currently detected anomalies. After generating the assessment report, the cloud server can also send the report to the user to inform them that no anomalies have been detected.

[0068] When the predicted result exceeds a preset threshold, the cloud server can generate an assessment report that includes suggestions for additional target checks. After generating the assessment report, the cloud server can also send the report to the user to inform them that further target checks are needed. This embodiment enables the assessment of early Alzheimer's disease, recommending further target checks only when the risk is high (predicted result exceeds a preset threshold), thus reducing testing costs for most users.

[0069] This invention acquires the user's magnetic resonance imaging (MRI) data and physiological data; inputs the MRI and physiological data into a cross-modal feature fusion model to obtain fused features; inputs the fused features into an expert model to obtain prediction results; generates an evaluation report based on the prediction results; processes the MRI and physiological data through the cross-modal feature fusion model to obtain high-level fused features, and then uses an expert model to process the fused features to obtain prediction results, thereby generating an evaluation report. This allows users to obtain an evaluation report without undergoing lumbar puncture or PET imaging, improving compliance and reducing testing costs.

[0070] Reference Figure 7 The present invention also proposes an early assessment device 20 for Alzheimer's disease, the early assessment device 20 for Alzheimer's disease comprising: Acquisition unit 201 is used to acquire the user's magnetic resonance imaging data and physiological data; Feature fusion unit 202 is used to input the magnetic resonance imaging data and the physiological data into a cross-modal feature fusion model to obtain fused features; Prediction unit 203 is used to input the fused features into an expert model to obtain prediction results; The generation unit 204 is used to generate an evaluation report based on the prediction results.

[0071] In some embodiments, the acquisition unit 201 is specifically used for: Detect whether the user is using a VR device and whether the user is wearing an EEG wearable device; If the user is using the VR device and the user is wearing the EEG wearable device, a VR cognitive task is generated and the VR cognitive task is sent to the VR device; Control the VR device to run the VR cognitive task, and control the wearable EEG device to collect the user's EEG data and eye movement data; The electroencephalogram (EEG) data and the eye movement data are identified as the physiological data.

[0072] In some embodiments, the feature fusion unit 202 is specifically used for: Feature extraction is performed on the magnetic resonance imaging data and the physiological data respectively to obtain magnetic resonance imaging features and physiological features; The magnetic resonance imaging features and the physiological features are input into the cross-modal feature fusion model; Obtain the fused features output by the cross-modal feature fusion model.

[0073] In some embodiments, the expert model includes a magnetic resonance imaging expert model, a physiological expert model, and a magnetic resonance imaging-physiological fusion expert model; the prediction unit 203 is specifically used for: The fusion features are respectively input into the magnetic resonance imaging expert model, the physiological expert model, and the magnetic resonance imaging-physiology fusion expert model; Obtain the first prediction result output by the magnetic resonance imaging expert model, the second prediction result output by the physiological expert model, and the third prediction result output by the magnetic resonance imaging-physiological fusion expert model; The first prediction result, the second prediction result, and the third prediction result are processed according to a preset voting algorithm to obtain the prediction result.

[0074] In some embodiments, the preset voting algorithm is a weighted voting algorithm; the prediction unit 203 is further specifically used for: Obtain the first weight of the magnetic resonance imaging expert model, the second weight of the physiological expert model, and the third weight of the magnetic resonance imaging-physiological fusion expert model, wherein the sum of the first weight, the second weight, and the third weight is 1; The product of the first prediction result and the first weight is calculated as the first product, the product of the second prediction result and the second weight is calculated as the second product, and the product of the third prediction result and the third weight is calculated as the third product. The sum of the first product, the second product, and the third product is calculated as the prediction result.

[0075] In some embodiments, the generation unit 204 is specifically used for: Determine whether the prediction result is greater than a preset threshold; If the prediction result is less than or equal to the preset threshold, an assessment report containing currently undetected abnormal risks is generated. If the predicted result is greater than the preset threshold, an evaluation report is generated that includes a suggestion to increase the target inspection.

[0076] Reference Figure 8 The present invention also proposes an early assessment system 30 for Alzheimer's disease, which includes a cloud server 301, a VR device 302 and an EEG wearable device 303. The cloud server 301 includes the early assessment device for Alzheimer's disease described in any one of the above.

[0077] In this embodiment, refer to Figure 7 and Figure 8The early assessment system 30 for Alzheimer's disease includes a cloud server 301, a VR device 302, and a wearable EEG device 303; the cloud server 301 may include, for example, Figure 7 The device 20 shown is an early assessment device for Alzheimer's disease. A cloud server 301 is wirelessly connected to a VR device 302 and an EEG wearable device 303.

[0078] The present invention also proposes a computer device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, is capable of performing any of the above-described methods for early assessment of Alzheimer's disease.

[0079] The present invention also proposes a storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, enable the processor to perform any of the above-described methods for early assessment of Alzheimer's disease.

[0080] The above description is only a part or preferred embodiment of the present invention. Neither the text nor the drawings should limit the scope of protection of the present invention. All equivalent structural transformations made using the content of the present invention specification and drawings under the overall concept of the present invention, or direct / indirect applications in other related technical fields, are included within the scope of protection of the present invention.

Claims

1. A method for early assessment of Alzheimer's disease, characterized in that, The methods for early assessment of Alzheimer's disease include: Acquire the user's magnetic resonance imaging data and physiological data; The magnetic resonance imaging data and the physiological data are input into a cross-modal feature fusion model to obtain fused features; The fused features are input into the expert model to obtain the prediction result; An evaluation report is generated based on the prediction results.

2. The method for early assessment of Alzheimer's disease according to claim 1, characterized in that, The acquisition of user physiological data includes: Detect whether the user is using a VR device and whether the user is wearing an EEG wearable device; If the user is using the VR device and the user is wearing the EEG wearable device, a VR cognitive task is generated and the VR cognitive task is sent to the VR device; Control the VR device to run the VR cognitive task, and control the wearable EEG device to collect the user's EEG data and eye movement data; The electroencephalogram (EEG) data and the eye movement data are identified as the physiological data.

3. The method for early assessment of Alzheimer's disease according to claim 1, characterized in that, The step of inputting the magnetic resonance imaging data and the physiological data into a cross-modal feature fusion model to obtain fused features includes: Feature extraction is performed on the magnetic resonance imaging data and the physiological data respectively to obtain magnetic resonance imaging features and physiological features; The magnetic resonance imaging features and the physiological features are input into the cross-modal feature fusion model; Obtain the fused features output by the cross-modal feature fusion model.

4. The method for early assessment of Alzheimer's disease according to claim 1, characterized in that, The expert models include a magnetic resonance imaging expert model, a physiological expert model, and a magnetic resonance imaging-physiological fusion expert model. The step of inputting the fused features into the expert model to obtain the prediction result includes: The fusion features are respectively input into the magnetic resonance imaging expert model, the physiological expert model, and the magnetic resonance imaging-physiology fusion expert model; Obtain the first prediction result output by the magnetic resonance imaging expert model, the second prediction result output by the physiological expert model, and the third prediction result output by the magnetic resonance imaging-physiological fusion expert model; The first prediction result, the second prediction result, and the third prediction result are processed according to a preset voting algorithm to obtain the prediction result.

5. The method for early assessment of Alzheimer's disease according to claim 4, characterized in that, The preset voting algorithm is a weighted voting algorithm; the process of processing the first prediction result, the second prediction result, and the third prediction result according to the preset voting algorithm to obtain the prediction result includes: Obtain the first weight of the magnetic resonance imaging expert model, the second weight of the physiological expert model, and the third weight of the magnetic resonance imaging-physiological fusion expert model, wherein the sum of the first weight, the second weight, and the third weight is 1; The product of the first prediction result and the first weight is calculated as the first product, the product of the second prediction result and the second weight is calculated as the second product, and the product of the third prediction result and the third weight is calculated as the third product. The sum of the first product, the second product, and the third product is calculated as the prediction result.

6. The method for early assessment of Alzheimer's disease according to claim 1, characterized in that, The step of generating an evaluation report based on the prediction results includes: Determine whether the prediction result is greater than a preset threshold; If the prediction result is less than or equal to the preset threshold, an assessment report containing currently undetected abnormal risks is generated. If the predicted result is greater than the preset threshold, an evaluation report is generated that includes a suggestion to increase the target inspection.

7. An early assessment device for Alzheimer's disease, characterized in that, The early assessment device for Alzheimer's disease includes: The acquisition unit is used to acquire the user's magnetic resonance imaging data and physiological data; The feature fusion unit is used to input the magnetic resonance imaging data and the physiological data into the cross-modal feature fusion model to obtain fused features; The prediction unit is used to input the fused features into the expert model to obtain the prediction result; A generation unit is used to generate an evaluation report based on the prediction results.

8. The early assessment device for Alzheimer's disease according to claim 7, characterized in that, The acquisition unit is specifically used for: Detect whether the user is using a VR device and whether the user is wearing an EEG wearable device; If the user is using the VR device and the user is wearing the EEG wearable device, a VR cognitive task is generated and the VR cognitive task is sent to the VR device; Control the VR device to run the VR cognitive task, and control the wearable EEG device to collect the user's EEG data and eye movement data; The electroencephalogram (EEG) data and the eye movement data are identified as the physiological data.

9. The early assessment device for Alzheimer's disease according to claim 7, characterized in that, The feature fusion unit is specifically used for: Feature extraction is performed on the magnetic resonance imaging data and the physiological data respectively to obtain magnetic resonance imaging features and physiological features; The magnetic resonance imaging features and the physiological features are input into the cross-modal feature fusion model; Obtain the fused features output by the cross-modal feature fusion model.

10. An early assessment system for Alzheimer's disease, characterized in that, The early assessment system for Alzheimer's disease includes a cloud server, a VR device, and a wearable EEG device, wherein the cloud server includes the early assessment device for Alzheimer's disease as described in any one of claims 7-9.

Images