An alzheimer's disease efficacy evaluation method based on multi-modal data

Abstract

This application belongs to the field of biomedicine, specifically relating to a method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data. It includes: constructing a personalized brain network model based on multimodal magnetic resonance imaging to locate hippocampal targets; implementing time-interventional electrical stimulation intervention; simultaneously collecting data from neuropsychological scales, electroencephalograms, event-related potentials, and blood biomarkers at four time points; constructing an efficacy evaluation system by integrating four-dimensional indicators: subjective cognition, objective electrophysiology, molecular pathology, and brain network regulation; and quantifying the intervention effect through principal component analysis and a linear mixed-effects model, and establishing a logistic regression classifier based on baseline features to predict individual response probabilities. This method achieves precise targeted intervention and multi-dimensional dynamic efficacy evaluation, improving the objectivity of the evaluation, the depth of mechanism analysis, and the ability for individualized clinical decision-making.

Description

Technical Field

[0001] This application belongs to the field of biomedicine, specifically relating to a method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data. Background Technology

[0002] With the accelerating aging of the population, Alzheimer's disease (AD), as the most common neurodegenerative disease, has become a major challenge to global public health systems. AD is characterized by progressive cognitive decline, initially manifesting as episodic memory impairment, subsequently affecting multiple cognitive domains such as language, executive function, spatial orientation, and emotion regulation, ultimately leading to complete loss of self-care ability. Current clinical treatments primarily rely on cholinesterase inhibitors and NMDA receptor antagonists, which only provide temporary symptom relief and cannot block or reverse disease progression. While emerging anti-Aβ monoclonal antibodies have made progress in pathological clearance, their widespread clinical application is limited by low blood-brain barrier penetration efficiency, high treatment costs, and potential serious adverse reactions.

[0003] Non-invasive neuromodulation techniques have been actively explored for AD intervention in recent years due to their safety and reproducibility. However, traditional methods such as repetitive transcranial magnetic stimulation (rTMS) and transcranial direct current stimulation (tDCS) are limited by physical penetration depth and spatial focusing ability, making it difficult to effectively target deep key memory structures such as the hippocampus, resulting in limited modulation effects. Although deep brain stimulation (DBS) has the potential for precise modulation, its invasiveness brings significant surgical risks, making it unsuitable for the long-term intervention needs of early-stage AD patients. Temporal Interference Stimulation (TIS), as an emerging non-invasive technique, forms a low-frequency envelope in deep brain regions through high-frequency electric field interference, theoretically enabling non-invasive and precise modulation of regions such as the hippocampus. However, existing research is mostly focused on healthy individuals or animal models, lacking a systematic approach for individualized target localization and dynamic efficacy evaluation for AD patients.

[0004] Current technologies also have significant limitations in assessing the efficacy of Alzheimer's disease (AD). Traditional methods relying on single neuropsychological scales are highly subjective and have low sensitivity, making it difficult to capture subtle cognitive changes. While multimodal biomarkers (such as MRI, EEG, and blood indicators) can provide objective data, an integrated assessment framework dynamically coupled with neural regulatory parameters has not yet been established. Especially in the context of TIS intervention, how to customize stimulation targets based on individual hippocampal structural-functional characteristics and simultaneously quantify neural circuit responses and cognitive improvements in real time using multimodal data remains a pressing technical challenge. Therefore, there is an urgent need for an Alzheimer's disease efficacy assessment method that integrates personalized anatomical-functional imaging guidance, multimodal dynamic monitoring, and closed-loop feedback mechanisms. Summary of the Invention

[0005] The purpose of this invention is to provide a method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data, which can effectively solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data includes the following steps: S1. Acquire multimodal magnetic resonance imaging data of patients with early Alzheimer's disease, construct an individualized brain network model based on the multimodal magnetic resonance imaging data, calculate hippocampal functional integrity score to identify effective targets, and output individualized hippocampal target localization data. S2, based on the individualized hippocampal target localization data, electrodes are placed on the patient's scalp surface, and time-interference electrical stimulation is applied to intervene, so that two high-frequency electric fields interfere in the bilateral hippocampal regions to form a low-frequency envelope electric field. S3, simultaneously collects neuropsychological scale scores, resting-state electroencephalogram signals, event-related potentials and peripheral blood biomarker concentration data at multiple preset time points; S4. Based on the data collected in step S3, construct a multimodal efficacy evaluation index system that includes subjective cognitive function indicators, objective neurophysiological indicators, molecular pathological burden indicators, and brain network regulation indicators. S5. Principal component analysis is used to reduce the dimensionality of the multimodal efficacy evaluation index system and integrate it to establish a linear mixed-effects model to quantify the effect of time-interference stimulation. Based on the baseline multimodal features, a logistic regression classifier is constructed to predict the individual treatment response probability and output the efficacy evaluation results.

[0007] Preferably, the construction of the individualized brain network model in step S1 includes: Input structural magnetic resonance imaging, functional magnetic resonance imaging, and diffusion tensor imaging data; The processing method is as follows: Based on structural magnetic resonance imaging data, the bilateral hippocampus was segmented and normalized to obtain the normalized hippocampal volume Z-fraction. Based on functional magnetic resonance imaging data, the average functional connectivity strength between the bilateral hippocampus and key nodes of the preset brain network was calculated. Based on diffusion tensor imaging data, white matter fiber bundles in the hippocampus-entorhinal cortex pathway were extracted, and their standardized fractional anisotropy Z-scores were calculated. The hippocampal functional integrity score was calculated using a weighted summation formula.

[0008] in, To standardize the hippocampal volume Z-fraction, For average functional connection strength, For standardized anisotropic Z-scores, , , These are weighting coefficients, and their sum is 1; If the score is lower than the preset threshold, the corresponding target point is excluded, and the spatial coordinates and electrode parameters of the valid hippocampal target point are output.

[0009] Preferably, the application of time-interference electrical stimulation in step S2 includes: Input the individualized hippocampal target localization data obtained in step S1; The treatment involves placing two pairs of electrodes on the scalp surface and applying first and second high-frequency alternating currents to generate a low-frequency envelope electric field through electric field interference. The total stimulation intensity does not exceed a safe threshold, and the intervention is repeated continuously. Output the status of intervention completion.

[0010] Preferably, the synchronous data acquisition in step S3 includes: Enter the preset time point information; The treatment involved collecting neuropsychological scale scores, resting-state electroencephalogram (EEG) signals, event-related potential (P300) components, and blood biomarker concentrations before the intervention, after the intervention, at the first follow-up time, and at the second follow-up time. Output the longitudinal multimodal dataset.

[0011] Preferably, the construction of the multimodal efficacy evaluation index system in step S4 includes: Input the neuropsychological scale scores, electroencephalogram signals, event-related potentials, and blood biomarker data collected in step S3; The treatment approach was to use the scale scores as subjective cognitive function indicators, the power spectral density and microstate parameters of electroencephalograms as objective neurophysiological indicators, changes in blood biomarker concentrations as molecular pathological burden indicators, and changes in the functional connectivity strength of brain networks in functional magnetic resonance imaging as brain network regulation indicators. Output data for the four-dimensional indicator system.

[0012] Preferably, the output efficacy evaluation results in step S5 include: Input the multimodal efficacy evaluation index system constructed in step S4; The approach involves using principal component analysis to reduce dimensionality and extract principal components, establishing a linear mixed-effects model to analyze the short-term and sustained effects of time-interference stimuli, and training a logistic regression classifier based on baseline features to predict response probabilities. Output quantitative effect values ​​and individualized prediction results.

[0013] Compared with the prior art, the present invention has the following beneficial effects: By fusing structural, functional, and diffusion-weighted magnetic resonance imaging (DMRI) data, this invention enables the customization of precise stimulation targets in the bilateral hippocampus and its functionally associated brain regions for each early-stage Alzheimer's disease patient. This overcomes the inherent limitation of traditional non-invasive neuromodulation techniques, which cannot effectively target deep memory core structures due to insufficient penetration depth. Finite element electric field modeling ensures that stimulation energy is highly focused on the target region while minimizing non-specific activation of the overlying cortex, thereby significantly improving the targeting and safety of the intervention.

[0014] This invention abandons the traditional assessment model that relies solely on subjective scales, and innovatively integrates four-dimensional data from neuropsychology, neuroelectrophysiology, molecular biology, and brain network imaging to construct a dynamic assessment framework that complements subjective and objective perspectives and combines macroscopic and microscopic data. This system can not only sensitively capture subtle improvements in cognitive function, but also reveal the mechanism of action of time-interference stimuli from multiple levels, such as neural circuits, synaptic plasticity, and pathological burden, providing a comprehensive and objective chain of evidence for efficacy assessment.

[0015] By employing a linear mixed-effects model, this invention can quantify the short-term effects (end of intervention) and the sustained effects (follow-up time points) of time-intervention stimuli, clarifying the time window for clinical benefit. More importantly, the logistic regression classifier constructed based on baseline multimodal features can predict the treatment response probability of individual patients in advance, providing hierarchical guidance for clinical decision-making and promoting the transformation of Alzheimer's disease intervention from a "one-size-fits-all" approach to a precise, individualized approach.

[0016] This invention incorporates a rigorous safety monitoring process, employing both adverse event recording and high-sensitivity magnetic resonance imaging (magnetic susceptibility-weighted imaging) for dual verification to monitor potential risks in real time. Combined with a standardized operating environment and emergency response plan, it ensures that subject safety is prioritized while exploring novel physical intervention methods, laying a solid foundation for subsequent large-scale clinical applications. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall process of an Alzheimer's disease efficacy evaluation method based on multimodal data provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0020] A multimodal data-based method for evaluating the efficacy of Alzheimer's disease treatment includes the following S: S1: Obtain personalized hippocampal target localization data: Perform multimodal magnetic resonance imaging examinations on patients with early Alzheimer's disease, including structural magnetic resonance imaging, functional magnetic resonance imaging, and diffusion tensor imaging. Based on the imaging data, construct a personalized brain network model, identify the spatial coordinates of the bilateral hippocampus and its functionally closely related brain regions, and calculate the optimal electrode placement position and current intensity parameters. In S1 above, firstly, patients with early-stage Alzheimer's disease who meet the inclusion criteria undergo multimodal magnetic resonance imaging (MRI) to obtain structural MRI, functional MRI, and diffusion tensor imaging.

[0021] After completing the multimodal data acquisition described above, the next stage is the construction of the personalized brain network model. This stage is executed by a medical image processing software platform, and the specific operations are divided into three sub-S: S11: Based on structural magnetic resonance imaging data, the bilateral hippocampus is precisely segmented using automated segmentation tools such as FSL or FreeSurfer, and the degree of cortical thickness atrophy is quantified. The segmentation results are output in spatial coordinates in millimeters and registered with the standard MNI152 template space to eliminate individual anatomical differences.

[0022] S12: Based on functional magnetic resonance imaging data, conventional preprocessing is first performed, including temporal correction, head motion correction, spatial normalization and Gaussian smoothing. Then, using the segmented bilateral hippocampus as seed regions, the functional connectivity strength between it and key nodes of the default network, central executive network and salience network is calculated.

[0023] The key nodes of the default network include the posterior cingulate cortex and the medial prefrontal cortex. Key nodes of the central execution network include the left and right dorsolateral prefrontal cortex and the posterior parietal cortex; The key nodes of the salient network include the left and right anterior islands and the anterior cingulate cortex.

[0024] Functional connectivity strength is quantified by calculating the time-series Pearson correlation coefficient of the blood oxygen level-dependent signal between the seed region and each key node.

[0025] S13: Based on diffusion tensor imaging data, a deterministic fiber tract tracking algorithm was used to extract white matter fiber tracts in the hippocampus-entorhinal cortex pathway and calculate the average fractional anisotropy value of this pathway. The fractional anisotropy value is a key indicator for measuring the integrity of white matter fibers, and its value ranges from 0 to 1. The higher the value, the more intact the fiber tract.

[0026] By combining the results of the three sub-S values—hippocampal volume, functional connectivity strength between the hippocampus and the three core networks, and fractional anisotropy of the hippocampus-entorhinal cortex pathway—a hippocampal functional integrity score is calculated through a weighted summation. This rating The calculation formula is as follows:

[0027] in, Z-score represents the hippocampal volume standardized to the mean of an age-matched healthy control group. This represents the average connection strength between the hippocampus and all critical functional nodes. Z-scores are the hippocampal-entorhinal cortex pathway FA values ​​standardized to the mean of age-matched healthy controls. Weighting coefficients. , , Determined based on prior knowledge or machine learning optimization, the sum is 1.

[0028] like If the value is below a preset threshold (e.g., -1.5), the hippocampus on that side is considered to be severely impaired, and its corresponding spatial coordinates will be excluded from the stimulation target. Only the hippocampus on that side will be stimulated. Targeted intervention is performed on hippocampal regions that exceed the threshold.

[0029] After determining the effective spatial coordinates of the bilateral hippocampal target points, the system uses finite element modeling technology to simulate the distribution of intracranial electric fields under different scalp electrode configurations. The model is constructed based on the patient's individualized structural magnetic resonance imaging data and includes different tissue layers such as scalp, skull, cerebrospinal fluid, gray matter, and white matter. Each tissue is assigned a specific conductivity parameter. Through iterative optimization algorithms, the optimal electrode placement and current intensity parameters are calculated to maximize the electric field envelope in the target hippocampal region while minimizing exposure in the overlying cortex region.

[0030] S2: Based on the individualized hippocampal target localization data: two pairs of electrodes are placed on the patient's scalp surface, and alternating currents with frequencies of the first high frequency and the second high frequency are applied to make the two electric fields interfere in the bilateral hippocampal regions to form a low-frequency envelope electric field. The total stimulation intensity does not exceed the preset safety threshold. Repeated continuous stimulation is performed once a day for a predetermined duration, and the intervention is continued for a predetermined number of days. Specifically, based on the individualized hippocampal target localization data and optimal electrode parameters calculated by S1, two pairs of circular gel electrodes are precisely placed on the patient's scalp surface to ensure good contact with the scalp. In this embodiment, the Neurolux™ series time-interference (TI) neuromodulation system provided by Hangzhou Ruierweikang Technology Co., Ltd. is used. A first high-frequency alternating current with a frequency of 2000Hz is applied to the first pair of electrodes, and a second high-frequency alternating current with a frequency of 2006Hz is applied to the second pair of electrodes. Since the two high-frequency electric fields are superimposed in the intracranial space, according to the principle of sum-to-product of trigonometric functions, a low-frequency modulated electric field envelope with a frequency of 6Hz (i.e., |2006-2000|=6Hz) is generated in the bilateral hippocampal regions. This envelope frequency is in the theta band (4-8Hz), which matches the natural oscillation frequency of the hippocampus during the memory encoding process.

[0031] Throughout the stimulation process, the sum of the current intensities of the two pairs of electrodes was strictly controlled within a preset safety threshold of no more than 4mA. The specific current intensity distribution ratio was determined by the finite element modeling results in S1 to ensure that the energy was focused on the target area. The stimulation mode employed repetitive continuous stimulation, with the treatment plan consisting of two consecutive weeks, six days a week, once a day, each session lasting 30 minutes.

[0032] In this embodiment, during each 30-minute stimulation session, the patient must remain resting with their eyes open, seated in a soundproof, softly lit dedicated intervention room. The ambient noise in the intervention room is controlled below 45 decibels, and the room temperature is maintained within the range of 24±1 degrees Celsius to minimize the impact of external interference on the stimulation effect. The operation procedure for the sham stimulation control group is exactly the same, but the stimulation device stops outputting effective current after the initial few seconds, retaining only electrode contact sensation to maintain the blinded design.

[0033] S3: Collect multidimensional baseline and follow-up data: At four time points, including before intervention, within a specific time period after intervention, the first follow-up time point after intervention, and the second follow-up time point, neuropsychological scale scores, resting-state electroencephalogram signals, time-related evoked potentials, and peripheral blood biomarker concentrations, were collected simultaneously. Specifically, four types of data were collected simultaneously at four pre-set time points: before intervention, within 24 hours after all interventions ended, 1 month after intervention, and 3 months after intervention. These four types of data constituted a complete longitudinal assessment dataset.

[0034] The first category of data consists of neuropsychological scale scores. These were completed by blinded assessors who were uniformly trained and unaware of the group assignments, in a standardized, quiet assessment room. The scales used included: the 13-item Alzheimer's Disease Assessment Scale - Cognitive Subscale, the Mini-Mental State Examination, the Montreal Cognitive Assessment Scale, the Auditory Word Learning Test, the 24-item Alzheimer's Disease Collaborative Study - Activities of Daily Living Scale, the Neuropsychiatric Questionnaire, the 17-item Hamilton Depression Rating Scale, and the Hamilton Anxiety Rating Scale.

[0035] The administration order, instructions, and scoring criteria for all scales were strictly followed in accordance with the manual to ensure the reliability and validity of the data.

[0036] The second type of data is resting-state electroencephalogram (EEG) signals. Before signal acquisition, ensure that the impedance of all electrodes is below 5kΩ. The sampling frequency is set to 1000Hz, and the recording duration is no less than 8 minutes. During the recording period, subjects are required to close their eyes, relax, remain awake, and avoid any conscious thought activity. The raw EEG data, after filtering and removal of electrooculography (EOG) and electromyography (EMG) artifacts, is used for subsequent analysis of the power spectral density of each frequency band and the frequency and duration of microstates.

[0037] The third type of data is time-related evoked potentials. The classic oddball paradigm was used to evoke the P300 component. The stimuli consisted of two types of pure tones: a high-frequency target stimulus (2000 Hz) and a low-frequency non-target stimulus (1000 Hz). The probability of presentation of the target stimulus was 20%, and the probability of presentation of the non-target stimulus was 80%. The stimulus interval (ISI) was randomly distributed between 1000 and 1500 milliseconds. Subjects were required to press a key as quickly as possible upon hearing the target stimulus. The latency (time from stimulus onset to peak) and amplitude (voltage difference from baseline to peak) of the P300 component were recorded, with particular attention paid to data from midline electrode sites such as Pz, Cz, and Fz.

[0038] The fourth category of data is the concentration of peripheral blood biomarkers. The biomarkers to be detected include: β-amyloid protein 42 / 40 ratio, total tau protein, phosphorylated tau protein 217, neurofilament light chain protein, glial fibrillary acidic protein, interleukin-6, tumor necrosis factor-α, and brain-derived neurotrophic factor.

[0039] S4: Construct a multimodal efficacy evaluation index system: The neuropsychological scale score is used as a subjective cognitive function index, the power spectral density of each frequency band of the resting-state electroencephalogram, the frequency of microstate occurrence, and the latency and amplitude of event-related potential P300 are used as objective neurophysiological indicators, the changes in the concentration of blood biomarkers are used as molecular pathological burden indicators, and the changes in the functional connectivity strength of the default network, central executive network, and salience network in functional magnetic resonance imaging are used as brain network regulation indicators; Specifically, the multi-dimensional data collected by S3 is structured and integrated to form four categories of mutually corroborating evaluation indicators.

[0040] Subjective cognitive function indicators were directly derived from neuropsychological scale scores. Among them, changes in the scores of the 13 Alzheimer's Disease Assessment Scales – Cognitive Subscale served as primary efficacy indicators, reflecting the degree of improvement or deterioration in overall cognitive function. Changes in scores from scales such as the Mini-Mental State Examination, the Montreal Cognitive Assessment, and the Auditory-Verbal Learning Test served as secondary indicators, reflecting the functional status of specific cognitive domains such as global cognition, memory, and attention, respectively. Scales such as the 24-item Alzheimer's Disease Collaborative Study – Activities of Daily Living Scale, the Neuropsychiatric Questionnaire, the 17-item Hamilton Depression Rating Scale, and the Hamilton Anxiety Rating Scale were used to assess changes in daily living abilities, psychosocial symptoms, and depressive and anxious moods.

[0041] Objective neurophysiological parameters are derived from resting-state electroencephalography (EEG) and time-related evoked potential (TLP) data. Specifically, these include: changes in the power spectral density of the theta band (4-8 Hz) and alpha band (8-13 Hz) in the resting state, which are closely related to memory and attention functions; changes in the frequency and average duration of microstates in the resting-state EEG, which reflect the transient dynamics of the brain's large-scale functional networks; and shortened latency and increased amplitude of the P300 component in TLPs, which are generally considered electrophysiological evidence of improved cognitive processing speed and resource allocation efficiency.

[0042] Molecular pathological burden indicators are derived from changes in the concentration of blood biomarkers. An increased β-amyloid 42 / 40 ratio and decreased concentrations of total tau protein and phosphorylated tau protein 217 are interpreted as peripheral evidence of reduced pathological burden of intracerebral amyloid plaques and neurofibrillary tangles. Decreased concentrations of neurofilament light chain proteins reflect a reduction in axonal injury, while increased concentrations of brain-derived neurotrophic factor (BDNF) may indicate enhanced neurotrophic support and synaptic plasticity. Changes in inflammatory markers such as glial fibrillary acidic protein, interleukin-6, and tumor necrosis factor-α are used to assess alterations in neuroinflammatory levels.

[0043] Brain network modulation indices were derived from functional magnetic resonance imaging (fMRI) data. The specific calculation method was as follows: based on the seed points defined in S1, a sliding window dynamic functional connectivity analysis (TMI) technique was used. For each window, the average functional connectivity strength between nodes within the default network, central executive network, and salience network was calculated, as well as the average functional connectivity strength between each pair of the three networks. Finally, these connectivity strengths at each time point after intervention were compared with the baseline period, and their rate of change was calculated as the core indicator for measuring the TSI's effect on large-scale brain network functional reorganization and modulation.

[0044] S5: Integrate and analyze the efficacy evaluation results: Principal component analysis is used to reduce the dimensionality of the multi-dimensional indicators and fuse them to establish a linear mixed-effects model to quantify the short-term and sustained effects of time-intervention stimuli on cognitive function. Based on the baseline multimodal features, a logistic regression classifier is constructed to predict the probability of individual treatment response.

[0045] Specifically, firstly, principal component analysis (PCA) was used to reduce the dimensionality of indicators across all dimensions. Key variables from subjective cognitive function indicators, objective neurophysiological indicators, molecular pathological burden indicators, and brain network regulation indicators were collectively incorporated into PCA analysis. Principal components with eigenvalues ​​greater than 1 were extracted, ensuring that the cumulative variance contribution rate of the first few extracted principal components was greater than 85%. These principal components represent independent information dimensions inherent in the multimodal data, effectively avoiding the one-sidedness of a single indicator.

[0046] Secondly, a linear mixed-effects model was established to quantify the short-term and persistent effects of time-based intervention stimuli on cognitive function. In this model, principal component scores extracted by PCA or core indicators such as ADAS-cog-13 were used as dependent variables, time points as fixed effects, and individual subjects as random effects to capture intra-individual repetitive measures correlations. Covariates included age, gender, years of education, and baseline cognitive scores to control for confounding factors. This model allows for accurate estimation of the short-term effect size at the end of the intervention, as well as the decay or enhancement of the persistent effect at 1-month and 3-month follow-ups.

[0047] Finally, based on the multimodal features at baseline, a logistic regression classifier was constructed to predict the probability of individual treatment response. Treatment response was defined as an improvement of ≥3 points in the 13-item Alzheimer's Disease Assessment Scale - Cognitive Subscale score compared to baseline at the 3-month follow-up after intervention. The input features of the logistic regression model included: baseline hippocampal volume, resting-state theta-band power, blood Aβ42 / 40 ratio, and the strength of functional connectivity within the default network. The model output was a probability value between 0 and 1, which, based on a preset cutoff value, categorized patients into a binary classification of "high response probability" or "low response probability." This classifier provides a quantitative tool for individualized treatment decisions in clinical practice.

[0048] Furthermore, this invention includes a built-in safety monitoring module. After each intervention, researchers meticulously record any adverse events, including headaches, dizziness, localized skin discomfort, and mood swings, categorizing them by severity (mild, moderate, severe) and relevance to the intervention (definite, possible, questionable, irrelevant). At four time points—baseline, within 24 hours after intervention, 1-month follow-up, and 3-month follow-up—magnetic susceptibility-weighted imaging and T2-weighted imaging scans are performed to highly sensitively assess the presence of structural risks such as intracranial microbleeds or cerebral edema. If a serious adverse event (such as a seizure) occurs during the intervention or imaging reveals new microbleeds, cerebral edema, or other abnormalities, the intervention is immediately terminated, and a pre-set emergency plan is activated, including emergency medical treatment, unblinding, and reporting to the ethics committee.

[0049] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data, characterized in that, Includes the following steps: S1. Acquire multimodal magnetic resonance imaging data of patients with early Alzheimer's disease, construct an individualized brain network model based on the multimodal magnetic resonance imaging data, calculate hippocampal functional integrity score to identify effective targets, and output individualized hippocampal target localization data. S2, based on the individualized hippocampal target localization data, electrodes are placed on the patient's scalp surface, and time-interference electrical stimulation is applied to intervene, so that two high-frequency electric fields interfere in the bilateral hippocampal regions to form a low-frequency envelope electric field. S3, simultaneously collects neuropsychological scale scores, resting-state electroencephalogram signals, event-related potentials and peripheral blood biomarker concentration data at multiple preset time points; S4. Based on the data collected in step S3, construct a multimodal efficacy evaluation index system that includes subjective cognitive function indicators, objective neurophysiological indicators, molecular pathological burden indicators, and brain network regulation indicators. S5. Principal component analysis is used to reduce the dimensionality of the multimodal efficacy evaluation index system and integrate it to establish a linear mixed-effects model to quantify the effect of time-interference stimulation. Based on the baseline multimodal features, a logistic regression classifier is constructed to predict the individual treatment response probability and output the efficacy evaluation results.

2. The method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data according to claim 1, characterized in that, The construction of the individualized brain network model in step S1 includes: Input structural magnetic resonance imaging, functional magnetic resonance imaging, and diffusion tensor imaging data; The processing method is as follows: Based on structural magnetic resonance imaging data, the bilateral hippocampus was segmented and normalized to obtain the normalized hippocampal volume Z-fraction. Based on functional magnetic resonance imaging data, the average functional connectivity strength between the bilateral hippocampus and key nodes of the preset brain network was calculated. Based on diffusion tensor imaging data, white matter fiber bundles in the hippocampus-entorhinal cortex pathway were extracted, and their standardized fractional anisotropy Z-scores were calculated. The hippocampal functional integrity score was calculated using a weighted summation formula. ， in, To standardize the hippocampal volume Z-fraction, For average functional connection strength, For standardized anisotropic Z-scores, , , These are weighting coefficients, and their sum is 1; If the score is lower than the preset threshold, the corresponding target point is excluded, and the spatial coordinates and electrode parameters of the valid hippocampal target point are output.

3. The method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data according to claim 1, characterized in that, The application of time-interference electrical stimulation in step S2 includes: Input the individualized hippocampal target localization data obtained in step S1; The treatment involves placing two pairs of electrodes on the scalp surface and applying first and second high-frequency alternating currents to generate a low-frequency envelope electric field through electric field interference. The total stimulation intensity does not exceed a safe threshold, and the intervention is repeated continuously. Output the status of intervention completion.

4. The method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data according to claim 1, characterized in that, The synchronous data acquisition mentioned in step S3 includes: Enter the preset time point information; The treatment involved collecting neuropsychological scale scores, resting-state electroencephalogram (EEG) signals, event-related potential (P300) components, and blood biomarker concentrations before the intervention, after the intervention, at the first follow-up time, and at the second follow-up time. Output the longitudinal multimodal dataset.

5. The method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data according to claim 1, characterized in that, The construction of the multimodal efficacy evaluation index system in step S4 includes: Input the neuropsychological scale scores, electroencephalogram signals, event-related potentials, and blood biomarker data collected in step S3; The treatment approach was to use the scale scores as subjective cognitive function indicators, the power spectral density and microstate parameters of electroencephalograms as objective neurophysiological indicators, changes in blood biomarker concentrations as molecular pathological burden indicators, and changes in the functional connectivity strength of brain networks in functional magnetic resonance imaging as brain network regulation indicators. Output data for the four-dimensional indicator system.

6. The method for evaluating the efficacy of Alzheimer's disease treatment based on multimodal data according to claim 1, characterized in that, The output of the efficacy assessment results in step S5 includes: Input the multimodal efficacy evaluation index system constructed in step S4; The approach involves using principal component analysis to reduce dimensionality and extract principal components, establishing a linear mixed-effects model to analyze the short-term and sustained effects of time-interference stimuli, and training a logistic regression classifier based on baseline features to predict response probabilities. Output quantitative effect values ​​and individualized prediction results.

Images