Intervention device and method for brain health screening and rehabilitation training closed-loop linkage

Abstract

The application provides an intervention device and method for brain health screening and rehabilitation training closed-loop linkage, relates to the technical field of data processing, and comprises the following steps: obtaining a risk type and a risk level output by a brain health screening process, encapsulating the risk type and the risk level as a strategy mapping reference; performing a strategy matching input strategy mapping operation by taking the strategy mapping reference as an input, obtaining an initial rehabilitation intervention parameter set, and converting the initial rehabilitation intervention parameter set into a training driving instruction; taking the training driving instruction as a terminal control input, synchronously executing a rehabilitation training operation and collecting an original electroencephalogram feedback signal; performing frequency domain space mapping on the original electroencephalogram feedback signal, dynamically extracting a neuroelectrophysiological topology node corresponding to a specific brain area phase synchronization extreme value; and obtaining a spatial coordinate sequence of the neuroelectrophysiological topology node. The application can realize high-fidelity space reconstruction of brain function state data.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intervention device and method that links brain health screening and rehabilitation training in a closed loop. Background Technology

[0002] In medical institutions and community health management, brain health screening and rehabilitation training have been gradually deployed digitally, forming a preliminary health data flow and training intervention link through standardized assessment processes and physiological signal collection.

[0003] However, existing data processing systems have some technical limitations in the conversion of screening data into rehabilitation parameters and in the real-time signal analysis during the training process. The core defects are concentrated in inaccurate data mapping and lack of dynamic analysis capabilities.

[0004] Specifically, existing solutions mostly use linear rules or static lookup tables to directly map screening risk types and levels to fixed rehabilitation parameters. They lack a computational mechanism for strategically encapsulating multidimensional screening data and adapting it to spatial topology, resulting in quantitative deviations between the initial intervention parameter set and the individual's basic brain function state.

[0005] During the rehabilitation training phase, the system typically only performs routine time-frequency domain statistics or global power spectrum analysis on the raw EEG signals, failing to extract precise spatial coordinates for phase synchronization extrema in specific brain regions, and thus unable to construct a whole-brain electrophysiological topology that conforms to the actual anatomical distribution of the cortex.

[0006] Due to the lack of in-depth quantitative analysis of changes in the spatial morphology and curvature field of EEG signals, existing systems cannot calculate the morphological volume projection ratio and morphological compactness index, which characterize the dynamic evolution of neural activity. As a result, brain functional state data during training only remain at the surface statistical level and are difficult to transform into quantitative dynamic tuples that can drive parameter correction. Summary of the Invention

[0007] This invention provides an intervention device and method that links brain health screening and rehabilitation training in a closed loop, which can achieve high-fidelity spatial reconstruction of brain functional state data.

[0008] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: The first aspect is an intervention device that links brain health screening and rehabilitation training in a closed loop, including: The screening data processing module is used to obtain the risk type and risk level output by the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark. The scheme mapping module is used to perform scheme mapping operations with the strategy mapping benchmark as the strategy matching input to obtain the initial rehabilitation intervention parameter set, and convert the initial rehabilitation intervention parameter set into training driving instructions. The rehabilitation training and signal acquisition module is used to take training drive commands as terminal control inputs, synchronously execute rehabilitation training operations and acquire raw EEG feedback signals; The EEG signal processing module is used to perform frequency domain spatial mapping on the raw EEG feedback signal and dynamically extract the neurophysiological topological nodes corresponding to the phase synchronization extrema of specific brain regions; based on the spatial coordinate sequence of the neurophysiological topological nodes, a whole-brain electrophysiological topological shell that fits the topological distribution of the cortex is constructed. The topology analysis module is used to analyze the contour projection trajectory and surface curvature field of the whole brain electrophysiological topology shell, calculate the morphological volume projection ratio and morphological compactness index, and integrate them into a neuromorphic dynamics tuple. The dynamic control module is used to compare the neuromorphic dynamic tuple as the dynamic control input with the preset benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated by the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set. The closed-loop iteration module is used to take the updated rehabilitation intervention parameter set as the input for periodic iteration, forming a closed-loop intervention process until the preset termination conditions are met.

[0009] Secondly, an intervention method that links brain health screening and rehabilitation training in a closed loop, applied in the aforementioned device, includes: Step 1: Obtain the risk type and risk level output from the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark; Step 2: Use the strategy mapping benchmark as the strategy matching input to perform the scheme mapping operation to obtain the initial rehabilitation intervention parameter set, and convert the initial rehabilitation intervention parameter set into training driving instructions; Step 3: Use training-driven instructions as terminal control inputs to synchronously execute rehabilitation training operations and collect raw EEG feedback signals; perform frequency domain spatial mapping on the raw EEG feedback signals to dynamically extract neurophysiological topological nodes corresponding to phase synchronization extrema of specific brain regions; construct a whole-brain electrophysiological topological shell that conforms to the topological distribution of the cortex based on the spatial coordinate sequence of the neurophysiological topological nodes. Step 4: Analyze the contour projection trajectory and surface curvature field of the whole brain electrophysiological topological shell, calculate the morphological volume projection ratio and morphological compactness index, and integrate them into a neuromorphic dynamics tuple. Step 5: The neuromorphic dynamics tuple is used as the dynamic control input and the deviation is compared with the preset target benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated through the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set. Step 6: Use the updated rehabilitation intervention parameter set as the input for periodic iteration, return to the terminal control input step of step 3 to form a closed-loop intervention process until the preset termination conditions are met.

[0010] Thirdly, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0011] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0012] The above-described solution of the present invention has at least the following beneficial effects: This application encapsulates risk types and levels into a strategy mapping benchmark through a screening data processing module, and performs high-precision scheme mapping calculations in conjunction with a scheme mapping module, effectively overcoming the problem of initial parameter quantification deviation caused by traditional linear mapping.

[0013] Based on this, the EEG signal processing module performs frequency domain spatial mapping on the original EEG feedback signal, accurately extracts the neurophysiological topological nodes corresponding to the phase synchronization extrema of specific brain regions, and constructs a whole-brain electrophysiological topological shell that fits the topological distribution of the cortex. This breaks through the spatial representation limitations of conventional global statistical analysis and realizes high-fidelity spatial reconstruction of brain functional state data.

[0014] The topology analysis module analyzes the contour projection trajectory and surface curvature field of the whole brain electrophysiological topology shell, quantitatively calculates the morphological volume projection ratio and morphological compactness index, and integrates them into a neuromorphic dynamics tuple. This transforms discrete EEG waveform data into quantitative dynamic features with clear physical meaning, solving the problem of state representation distortion caused by coarse dynamic analysis.

[0015] The dynamic control module compares the tuple with the preset benchmark to calculate the deviation, automatically generates difficulty adjustment instructions and parameter correction instructions, and overwrites and updates the initial parameter set. This eliminates the flow gap between the screening benchmark and the training feedback data, and achieves real-time, accurate, and dynamic adaptation of intervention parameters. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of an intervention method for closed-loop linkage of brain health screening and rehabilitation training provided in an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of an intervention device that links brain health screening and rehabilitation training in a closed loop, provided in an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0019] like Figure 1 As shown, embodiments of the present invention propose an intervention device that links brain health screening and rehabilitation training in a closed loop, comprising: The screening data processing module is used to obtain the risk type and risk level output by the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark. The scheme mapping module is used to perform scheme mapping operations with the strategy mapping benchmark as the strategy matching input to obtain the initial rehabilitation intervention parameter set, and convert the initial rehabilitation intervention parameter set into training driving instructions. The rehabilitation training and signal acquisition module is used to take training drive commands as terminal control inputs, synchronously execute rehabilitation training operations and acquire raw EEG feedback signals; The EEG signal processing module is used to perform frequency domain spatial mapping on the raw EEG feedback signal and dynamically extract the neurophysiological topological nodes corresponding to the phase synchronization extrema of specific brain regions; based on the spatial coordinate sequence of the neurophysiological topological nodes, a whole-brain electrophysiological topological shell that fits the topological distribution of the cortex is constructed. The topology analysis module is used to analyze the contour projection trajectory and surface curvature field of the whole brain electrophysiological topology shell, calculate the morphological volume projection ratio and morphological compactness index, and integrate them into a neuromorphic dynamics tuple. The dynamic control module is used to compare the neuromorphic dynamic tuple as the dynamic control input with the preset benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated by the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set. The closed-loop iteration module is used to take the updated rehabilitation intervention parameter set as the input for periodic iteration, forming a closed-loop intervention process until the preset termination conditions are met.

[0020] In this embodiment of the invention, the risk type and level are encapsulated into a strategy mapping benchmark by the screening data processing module, and high-precision scheme mapping calculations are performed in conjunction with the scheme mapping module. This effectively overcomes the initial parameter quantification deviation problem caused by traditional linear mapping, providing precise training drive instructions that fit the individual's basic state for rehabilitation training. Based on this, the EEG signal processing module performs frequency domain spatial mapping on the raw EEG feedback signal, accurately extracting the neurophysiological topological nodes corresponding to the phase synchronization extrema of specific brain regions, and constructing a whole-brain electrophysiological topological shell that fits the cortical topological distribution. This breaks through the spatial representation limitations of conventional global statistical analysis, achieving high-fidelity spatial reconstruction of brain functional state data. Furthermore, the topology analysis module analyzes the contour projection trajectory and surface curvature field of the whole-brain electrophysiological topological shell, quantitatively calculates the morphological volume projection ratio and morphological compactness index, and integrates them into a neuromorphic dynamics tuple. This transforms discrete EEG waveform data into quantitative dynamic features with clear physical meaning, solving the problem of state representation distortion caused by coarse dynamic analysis. The dynamic adjustment module compares the tuple with the preset benchmark, automatically generates difficulty adjustment and parameter correction instructions, and updates the initial parameter set. This eliminates the flow gap between the screening benchmark and training feedback data, achieving real-time, accurate, and dynamic adaptation of intervention parameters. The closed-loop iteration module uses the updated parameter set as periodic iterative input, forming a complete closed loop of bidirectional data flow verification and continuous optimization. This ensures that the entire intervention process maintains high accuracy in data analysis and high matching degree in parameter control, meeting the requirements for data processing accuracy, dynamic response speed, and closed-loop control stability in brain health intervention scenarios.

[0021] like Figure 2 As shown, embodiments of the present invention also provide an intervention method for closed-loop linkage of brain health screening and rehabilitation training, which is applied in the device and includes: Step 1: Obtain the risk type and risk level output from the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark; Step 2: Use the strategy mapping benchmark as the strategy matching input to perform the scheme mapping operation to obtain the initial rehabilitation intervention parameter set, and convert the initial rehabilitation intervention parameter set into training driving instructions; Step 3: Use training-driven instructions as terminal control inputs to synchronously execute rehabilitation training operations and collect raw EEG feedback signals; perform frequency domain spatial mapping on the raw EEG feedback signals to dynamically extract neurophysiological topological nodes corresponding to phase synchronization extrema of specific brain regions; construct a whole-brain electrophysiological topological shell that conforms to the topological distribution of the cortex based on the spatial coordinate sequence of the neurophysiological topological nodes. Step 4: Analyze the contour projection trajectory and surface curvature field of the whole brain electrophysiological topological shell, calculate the morphological volume projection ratio and morphological compactness index, and integrate them into a neuromorphic dynamics tuple. Step 5: The neuromorphic dynamics tuple is used as the dynamic control input and the deviation is compared with the preset target benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated through the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set. Step 6: Use the updated rehabilitation intervention parameter set as the input for periodic iteration, return to the terminal control input step of step 3 to form a closed-loop intervention process until the preset termination conditions are met.

[0022] It should be noted that this device is a device corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0023] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Receive raw output data from the brain health screening process, and parse the raw output data to separate the risk type identifier and risk level quantification value. Specifically, this includes: parsing multiple data fields from the raw output data of the brain health screening process according to a predefined JSON or XML data exchange format; the predefined data exchange format specifies the field name, field type, and nesting structure, including the risk type field and the risk level field; extracting key identifiers from the risk type field using a set string matching rule, which adopts a fuzzy matching calculation method based on a keyword dictionary: first, the input text in the risk type field is segmented and filtered for stop words, then the edit distance or semantic similarity between the filtered words and each standard term in the dictionary is calculated, and the standard term with the highest similarity and exceeding a set threshold is selected as the matching result, thereby generating a standardized risk type identifier.

[0024] This risk type identifier is used to uniquely distinguish various pathological subtypes, including early Alzheimer's disease, vascular cognitive impairment, frontotemporal dementia, Lewy body dementia, cognitive impairment caused by Parkinson's disease, and mild cognitive impairment. The original risk level quantification value is read from the risk level field through preset numerical extraction rules. The numerical extraction rules include type verification of field content, outlier detection, and numerical range truncation to ensure that the output risk level quantification value is a continuous value within a valid range, used to characterize the severity of the screened cognitive impairment.

[0025] Step 1.2 involves mapping the separated risk type identifiers to preset intervention strategy index codes, and simultaneously performing dimensionless normalization on the separated risk level quantification values ​​to obtain a relative intensity scalar. Specifically, this includes using the risk type identifier as a query key to perform a key-value matching operation in a preset intervention strategy mapping table. The preset intervention strategy mapping table is a lookup structure with the risk type identifier as the key and the intervention strategy index code as the value, where each risk type identifier uniquely corresponds to an intervention strategy index code. The intervention strategy index code is a fixed-length numeric or string code used to abstractly represent a specific type of rehabilitation intervention. The strategy combination includes, for example, the type of cognitive training, the mode of neural modulation, and the selection of targets. After locating the intervention strategy index code that uniquely corresponds to this risk type identifier, the obtained risk level quantification value is simultaneously sent to the normalization processing unit. This normalization processing unit is a logic module that specifically performs numerical scaling. The normalization processing unit uses the extreme value normalization method, taking the minimum and maximum observation values ​​of the same screening scale in the historical database as the upper and lower boundaries, to perform a linear transformation on the current risk level quantification value, thereby eliminating the dimensional differences between different screening scales and obtaining a relative intensity scalar with a value range between 0 and 1.

[0026] Step 1.3 involves combining the preset intervention strategy index code and the relative intensity scalar into a two-dimensional strategy mapping tuple, using this tuple as the strategy mapping benchmark. Specifically, this includes: combining and encapsulating the obtained intervention strategy index code and relative intensity scalar according to a preset two-dimensional tuple data structure. The preset two-dimensional tuple data structure is an ordered data container with a fixed length of 2, where the first position is defined as storing the intervention strategy index code and the second position is defined as storing the relative intensity scalar. For example, if the intervention strategy index code output in step 1.2 is CT01, corresponding to the cognitive rehabilitation and neuromodulation scheme direction for mild cognitive impairment, and the output relative intensity scalar is 0.62, then the two-dimensional strategy mapping tuple formed after combination and encapsulation is (CT01, 0.62).

[0027] The combination process is as follows: First, an empty two-dimensional tuple instance is created. The intervention strategy index code is written into the slot with index 0 of this tuple, and then the relative intensity scalar is written into the slot with index 1, thus forming a complete key-value pair tuple object. The intervention strategy index code, as the first element of the tuple, is used to indicate the type and direction of the rehabilitation program to be selected in the subsequent program mapping operation. For example, CT01 points to a cognitive rehabilitation and low-frequency repetitive transcranial magnetic stimulation program with memory strategy training as the core. The relative intensity scalar, as the second element of the tuple, is used to quantify the initial intensity level of the rehabilitation intervention. For example, 0.62 represents a moderate to high intervention intensity, corresponding to the initial difficulty level of the training task and the setting basis of the stimulation pulse intensity. After the combination is completed, this two-dimensional strategy mapping tuple is formally determined as the strategy mapping benchmark and output in the form of a structured data object.

[0028] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1 involves unpacking the two-dimensional policy mapping tuple to separate the intervention policy index code and the relative strength scalar. Specifically, this includes receiving the output two-dimensional policy mapping tuple and performing an unpacking operation on it. The unpacking operation involves extracting the two elements from the tuple according to their storage order, where the first element is extracted as the intervention policy index code and the second element is extracted as the relative strength scalar. After unpacking, the intervention policy index code and the relative strength scalar are provided as two independent variables for use in subsequent steps.

[0029] Step 2.2: Using the intervention strategy index code as the key, retrieve the preset rehabilitation program database and match the corresponding cognitive rehabilitation basic parameter template and neuromodulation basic parameter template. Specifically, this includes: using the separated intervention strategy index code as the key, performing precise matching and retrieval in the preset rehabilitation program database; the preset rehabilitation program database is a pre-constructed relational storage structure, with the intervention strategy index code as the primary key, and each record corresponding to a complete set of rehabilitation program parameter combinations; this database is organized according to the correspondence between risk type and intervention strategy, for example, differentiated rehabilitation program entries are preset for different pathological subtypes such as mild cognitive impairment, early Alzheimer's disease, and vascular cognitive impairment; each entry contains two core components: a cognitive rehabilitation basic parameter template and a neuromodulation basic parameter template.

[0030] The cognitive rehabilitation basic parameter template is a structured set of parameters, specifically including: a training type field, used to specify different cognitive domains such as working memory training, executive function training, attention training, or processing speed training; a task difficulty baseline field, used to set the initial task difficulty level, usually represented by an integer from 1 to 10; a stimulus presentation time field, used to specify the display duration of visual or auditory stimuli in each training trial; a feedback delay field, used to set the waiting time for the intervention system to provide feedback after the user makes a response; and a training duration field, used to specify the total duration of a single rehabilitation training session.

[0031] The neuromodulation basic parameter template is another structured set of parameters, specifically including: a stimulation mode field, used to specify different modulation methods such as repetitive transcranial magnetic stimulation, transcranial direct current stimulation, or transcranial alternating current stimulation; a stimulation intensity baseline field, used to set the initial stimulation intensity, usually expressed in milliamperes or as a percentage of the maximum output intensity; a stimulation frequency field, used to specify the pulse repetition frequency; a target coordinate field, used to store three-dimensional coordinates based on the standard brain space coordinate system to accurately locate the brain region to be modulated, such as the dorsolateral prefrontal cortex, the temporoparietal junction, or the motor cortex; and a pulse train duration and interval field, used to set the duration of each stimulation sequence and the resting interval between sequences. After successful retrieval, the matched cognitive rehabilitation basic parameter template and neuromodulation basic parameter template are read from the database and temporarily stored in the memory of the intervention system.

[0032] Step 2.3: Input the intervention strategy index code and relative intensity scalar into the preset topological response field model; the preset topological response field model calls the corresponding cortical region activation map according to the index code, and performs nonlinear modulation on the activation gradient field in the map according to the intensity scalar to obtain an initial rehabilitation intervention parameter set that matches the cognitive rehabilitation basic parameter template and the neuromodulation basic parameter template. Specifically, the preset topological response field model is a generative neural network model built on a conditional variational autoencoder architecture. This architecture can generate a distribution consistent with the real brain functional state based on the input conditions by modeling the latent space of the high-dimensional neural response distribution. The topological response field is constructed as follows: The encoder consists of multiple convolutional and fully connected layers stacked together, which is responsible for compressing the high-dimensional cortical activation distribution map into a mean vector and a log-variance vector in a low-dimensional latent space. The latent space is set to 32 dimensions to capture the core features of cortical activation patterns under different intervention strategies. The decoder consists of multiple deconvolutional and fully connected layers, which are responsible for reconstructing the complete cortical activation distribution map from the latent variables and conditional inputs. The conditional input is a concatenation vector of the one-hot encoding of the intervention strategy index code and the relative intensity scalar, which is injected into the intermediate layers of the encoder and decoder in the form of conditional affine transformations.

[0033] The specific construction process of this model is as follows: First, historical brain health screening and rehabilitation training data are collected, including intervention strategy index codes for patients with different risk types, corresponding relative intensity scalars, and synchronously recorded multi-lead EEG signals. The collected EEG signals are processed for source localization. A standardized low-resolution EEG algorithm is used to inversely map the scalp electrode potentials onto approximately 15,000 dipole source points on the cortical surface, reconstructing the activation intensity value of each source point to form an activation intensity distribution map of the cortical surface, which serves as the training target for the model. Then, for each training sample, the corresponding intervention strategy index code is converted into a one-hot vector and concatenated with the relative intensity scalar to form a conditional input tensor. The encoder receives the cortical activation distribution map as input, and after downsampling in a convolutional layer and mapping in a fully connected layer, outputs the mean vector and logarithmic variance vector of the latent space. The decoder randomly samples latent variables from the latent space, and together with the conditional input tensor, it is upsampled in a deconvolutional layer to reconstruct a cortical activation distribution map of the same size as the input.

[0034] The model training process uses the lower bound of evidence as an optimization criterion, which consists of two parts: the first is the reconstruction consistency constraint, which is the pixel-level difference between the original cortical activation distribution map and the distribution map reconstructed by the decoder, measured by mean squared error; the second is the latent space regularization constraint, which is the relative entropy between the latent distribution output by the encoder and the standard Gaussian distribution, used to prevent overfitting and ensure the continuity of the latent space. During training, historical datasets are input into the model in batches, and the network weight parameters of the encoder and decoder are iteratively adjusted through the backpropagation algorithm until the sum of the two constraints converges to below the preset threshold. At this point, the model can generate cortical region activation maps that conform to the true distribution pattern based on the input intervention strategy index code and relative intensity scalar. In a specific application scenario, when the intervention strategy index code corresponds to the memory training program for patients with mild cognitive impairment, and the relative strength scalar is 0.6, the model generates an activation map mainly composed of the hippocampus and medial temporal lobe, with the activation intensity increasing as the scalar increases. Compared with traditional linear interpolation or lookup table mapping methods, this model has the following advantages: it can capture the nonlinear coupling relationship of cortical activation under different intervention strategies; the generated activation map has smooth transition characteristics with individual differences; and it has the ability to generalize to unseen risk types and intensity combinations.

[0035] After construction and training, the model is deployed. During the real-time inference phase, the model first calls the hidden code of the corresponding cortical region activation map from the internal latent space based on the input intervention strategy index code. Then, it modulates the hidden code according to the relative intensity scalar and decodes it to generate the corresponding activation gradient field. The modulation adopts the conditional affine transformation mechanism in the conditional variational autoencoder, so that the activation gradient field also evolves smoothly when the intensity scalar changes continuously. After modulation, an initial rehabilitation intervention parameter set is output that matches the cognitive rehabilitation basic parameter template and the neuromodulation basic parameter template in step 2.2. This parameter set contains specific values ​​after modulation, such as training difficulty level, stimulus pulse intensity, task switching frequency, etc.

[0036] Step 2.4 involves encoding each parameter item in the initial rehabilitation intervention parameter set into a training-driven instruction sequence that can be directly parsed by the rehabilitation training terminal, ending with the training-driven instruction. Specifically, this includes: encoding each parameter item in the obtained initial rehabilitation intervention parameter set into an instruction; the instruction encoding refers to converting the name and value of each parameter item into corresponding instruction fields according to a pre-agreed communication protocol of the rehabilitation training terminal, including an instruction header, parameter identifier, parameter value, and check bit; concatenating the converted instruction fields of all parameter items into a complete training-driven instruction sequence according to the execution order; this training-driven instruction sequence can be directly parsed and executed by the rehabilitation training terminal, ending with the training-driven instruction.

[0037] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1 involves parsing the training-driven instruction sequence into real-time control parameters, and synchronously driving the rehabilitation training terminal to perform cognitive rehabilitation operations and neuromodulation operations based on the real-time control parameters. Specifically, this includes: receiving the training-driven instruction sequence output in step 2.4, and parsing each instruction frame in the sequence according to the communication protocol agreed upon by the rehabilitation training terminal; during parsing, first reading the frame header identifier to confirm the frame start position, extracting the parameter identifier and parameter value fields, then reading the check bit and verifying the data integrity through XOR operation; after the verification is passed, writing the corresponding parameter value into the specified offset address of the corresponding control register in the terminal according to the parameter identifier. The specified offset address is a fixed location pre-allocated in the terminal's internal memory for each control parameter. For example, offset address 0x00 is used to store the training difficulty level, and 0x02 is used to store the stimulus pulse intensity, thereby restoring the complete set of real-time control parameters.

[0038] This real-time control parameter set includes two main categories: cognitive rehabilitation parameters and neuromodulation parameters. Based on this set, the cognitive rehabilitation module and the neuromodulation module are simultaneously driven by a synchronous clock signal, enabling them to work collaboratively under the same time reference. The functional connection between the cognitive rehabilitation module and the neuromodulation module is as follows: the task presentation sequence of the cognitive rehabilitation module and the pulse output sequence of the neuromodulation module are synchronously triggered by the same clock source, ensuring that the modulation stimulus and the cognitive task are precisely aligned in time. For example, transcranial magnetic stimulation pulses targeting the hippocampus are applied simultaneously while the user performs a memory encoding task. Cognitive rehabilitation operations include presenting visual stimuli on the display device, playing audio stimuli, collecting the user's key or touch responses, and recording reaction time and accuracy. Neuromodulation operations include outputting a precise timing pulse sequence to the corresponding brain region through stimulation coils or electrodes according to the preset stimulation mode, stimulation intensity, stimulation frequency, and target coordinates.

[0039] Step 3.2: During the rehabilitation training operation, raw EEG feedback signals from the user's scalp are continuously acquired at a preset sampling rate. The raw EEG feedback signals are then filtered and denoised to obtain preprocessed EEG signals. Specifically, this includes: during the rehabilitation training operation, raw EEG feedback signals from multiple electrode locations on the user's scalp are continuously acquired at a preset sampling rate; the preset sampling rate is set to 1000 Hz based on the frequency band range of the EEG signals to meet the Nyquist sampling theorem's requirement for acquiring the highest frequency components; and a 32-channel EEG sensor is used in accordance with the international 10 to 20 standard system for acquisition. Alternatively, a 64-electrode cap is used, with each electrode maintaining an impedance of less than 5 kiloohms between itself and the scalp. The obtained raw EEG feedback signal is first input into a hardware bandpass filter to filter out baseline drift components below 0.5 Hz and high-frequency noise components above 70 Hz. Then, blind source separation is performed on the filtered signal to identify and remove independent components corresponding to electrooculography (EOG), electromyography (EMG), and power frequency interference. The remaining components are reconstructed into a clean EEG signal through inverse transformation, which is the preprocessed EEG signal. This preprocessed EEG signal fully preserves the waveform characteristics of the delta, theta, alpha, and beta bands that are closely related to cognitive activities.

[0040] Step 3.3 involves performing time-frequency conversion on the preprocessed EEG signal to map the time-domain waveform to the frequency-domain space, and calculating the phase synchronization value of each lead signal within a preset frequency band in the frequency-domain space. Specifically, this includes: performing time-frequency conversion on the output preprocessed EEG signal; specifically, using a sliding window-based time-frequency analysis method, setting a Hamming window as the window shape, with a window length corresponding to a time span of 256 sampling points, and an overlap ratio of 50% between adjacent windows, sliding frame by frame with a fixed step size; performing frequency domain decomposition on the signal segment within each window, mapping the time-domain waveform signal of each lead to the frequency-time joint plane, and obtaining the complex amplitude and instantaneous phase information of each frequency component within each time window; in the frequency-domain space, extracting the instantaneous phase sequence of any two leads within the same frequency band for each of the four preset frequency bands, namely the delta band, theta band, alpha band, and beta band.

[0041] The phase synchronization value between two leads is calculated using a phase consistency assessment method. The specific process is as follows: First, the instantaneous phase values ​​of the two leads within the same time window and frequency band are subtracted to obtain the phase difference value at each time point. This difference value lies within the range of -180° to +180° or -π to +π. For each phase difference value, it is considered as an angle on a unit circle, i.e., a point on a circle with the origin as its center and a radius of 1. The rectangular coordinates of this point are obtained by calculating the sine and cosine values ​​of this angle, thus converting each phase difference value into a pair of planar coordinate values. The planar coordinate values ​​corresponding to all time points within the window are vector-superimposed: the x-coordinates and y-coordinates of all points are summed to obtain a composite vector with the sum of the x-coordinates and the sum of the y-coordinates; the magnitude of this composite vector, which is the square root of the sum of the squares of the x-coordinates and the sum of the squares of the y-coordinates, is the phase lock value; the phase lock value ranges from 0 to 1, the closer the value is to 1, the higher the phase synchronization degree of the two leads in this frequency band, and the closer the value is to 0, the more random the phase relationship is and the no synchronization is; after completing this calculation, the phase synchronization value of the lead pair in the current frequency band is obtained.

[0042] Step 3.4: Extract spatial points in each brain region where the phase synchronization value reaches a preset extreme threshold, mark these spatial points as neurophysiological topological nodes, and record the three-dimensional spatial coordinates of each neurophysiological topological node. Specifically, this includes: statistically analyzing all leads in each brain region based on the calculated phase synchronization values ​​of each lead pair in each frequency band; the brain regions are divided into frontal lobe, parietal lobe, temporal lobe, occipital lobe, and central region according to their anatomical location, and each brain region contains several lead locations; for each brain region, first extract the phase synchronization values ​​of all possible lead pairs in that region under four preset frequency bands, and calculate the maximum value under each frequency band; compare the maximum value with the preset extreme threshold corresponding to that brain region; the preset extreme threshold is determined by the distribution of phase synchronization values ​​of healthy control groups under the same type of cognitive task in an offline statistical historical database, specifically taking the 85th percentile of the distribution as the threshold, and setting independent thresholds for different brain regions and different frequency bands.

[0043] When the phase synchronization value of a certain lead pair reaches or exceeds the preset extreme threshold of the corresponding frequency band in a certain brain region, the spatial points of both leads of the lead pair are marked as candidate nodes. After traversing all lead pairs and all frequency bands, the candidate nodes are deduplicated, that is, the same spatial point is kept only once, and the remaining spatial points after deduplication are officially marked as neurophysiological topological nodes. For each marked node, its corresponding three-dimensional spatial coordinates, including X-axis, Y-axis and Z-axis coordinates, are read by the electrodes in the pre-calibrated coordinates in the standardized brain space and recorded in millimeters. At the same time, the brain region information to which each node belongs and the frequency band label that triggered the marking are attached to each node for use in subsequent topology reconstruction.

[0044] Step 3.5: Arrange the three-dimensional spatial coordinates of all neurophysiological topological nodes into a spatial coordinate sequence according to the cortical anatomical order. Perform topological reconstruction of the spatial coordinate sequence using the convex boundary envelope of the spatial extremum points to construct a closed whole-brain electrophysiological topological shell that conforms to the cortical topological distribution. Specifically, this includes arranging the three-dimensional spatial coordinates of all recorded neurophysiological topological nodes according to the cortical anatomical order, starting from the prefrontal lobe polar region, and sequentially passing through the superior frontal gyrus, middle frontal gyrus, inferior frontal gyrus, precentral gyrus, postcentral gyrus, superior parietal gyrus, inferior parietal gyrus, and superior temporal lobe. The system consists of the gyrus, middle temporal gyrus, inferior temporal gyrus, superior occipital gyrus, inferior occipital gyrus, and so on, up to the occipital lobe polar region. Simultaneously, it is organized alternately from the left hemisphere to the right hemisphere, forming a spatial coordinate sequence with a definite physiological order. This spatial coordinate sequence is then subjected to topological reconstruction of the convex boundary envelope of spatial extrema points. Specifically, this includes the following sub-operations: First, scan the three-dimensional coordinates of all nodes to find the minimum and maximum points in the X-axis direction, the Y-axis direction, and the Z-axis direction. After deduplication, 3 to 6 extrema points are obtained, which are used as the initial envelope reference point set.

[0045] Based on this initial set of reference points, an initial convex envelope surface is constructed. The initial surface is a minimal polyhedron containing these reference points. An incremental envelope expansion algorithm is used to traverse each remaining node in turn: that is, to determine whether the current node is inside or on the boundary of an existing envelope surface. If it is inside or on the boundary, it is skipped; if it is outside, the node is added to the envelope surface. Based on the visibility relationship between the node and the existing vertices of the envelope surface, the new envelope surface boundary is recalculated, the occluded old facets are deleted, and new facets connecting the new node are added, so that the envelope surface expands to include the node.

[0046] Repeat the above process until all nodes are contained within the envelope surface, and finally construct a closed convex polyhedron composed of multiple triangular facets. Each facet of this polyhedron is located on the outermost layer of all nodes, and its interior does not contain any nodes. The closed convex polyhedron is the whole-brain electrophysiological topological shell, whose shape closely conforms to the cortical topological distribution and is used to characterize the overall spatial morphology and connectivity of the brain functional network under the current rehabilitation training state.

[0047] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Extract the contour projection trajectory of the outer surface from the whole-brain electrophysiological topological shell, and project the contour projection trajectory onto three orthogonal coordinate planes in three-dimensional space to obtain three sets of projected closed curves. Specifically, this includes: extracting the contour projection trajectory of the outer surface from the constructed whole-brain electrophysiological topological shell; the shell is a closed convex polyhedron composed of multiple triangular facets, and its outer surface contour is formed by connecting all the edges located on the boundary end to end. Specifically, the edges on the shell that are used by only one triangular facet are extracted, and the set of these edges is the outer surface contour line; the contour trajectory is then projected onto three orthogonal coordinate planes in three-dimensional space to obtain three sets of projected closed curves. The projection operation involves projecting onto three orthogonal coordinate planes in three-dimensional space: the XY plane, the XZ plane, and the YZ plane. The specific steps are as follows: For each spatial point on the contour, retain the two coordinate components parallel to the projection plane and set the coordinate components perpendicular to the projection plane to zero. For example, when projecting onto the XY plane, set the Z coordinate of each point to zero and retain the X and Y coordinates; when projecting onto the XZ plane, set the Y coordinate to zero; and when projecting onto the YZ plane, set the X coordinate to zero. After projection, each spatial contour line forms a closed curve on the corresponding plane, composed of sequentially connected two-dimensional points, thus obtaining three sets of projected closed curves.

[0048] Step 4.2: Calculate the projected area enclosed by each set of projected closed curves to obtain three sets of projected areas; simultaneously calculate the total volume of the whole-brain electrophysiological topological shell, and perform ratio calculations between this total volume and the three sets of projected areas to obtain the morphological volume projection ratios in three directions. Specifically, this includes: calculating the projected area enclosed by each of the three sets of projected closed curves; for any curve formed by... N A closed curve whose vertices are arranged in clockwise or counterclockwise order has an area of... A The calculation uses the Gaussian area formula, which is expressed as follows: ; in, For the first on the projection curve i The planar coordinates of the vertices, when hour, Take the first vertex The absolute value sign ensures that the area is positive; applying this formula to the projection curves on the XY, XZ, and YZ planes respectively yields three projected areas, denoted as... 、 and Simultaneously, the total volume of the whole-brain electrophysiological topological shell was calculated. ; because the shell is made of M It is composed of three triangular facets, and the three vertices of each facet form a tetrahedron in space (with the origin as the reference point). The volume is calculated using the following formula: ; in, 、 、 The first j The spatial position vectors of the three vertices of a triangular facet are used. The dot sign indicates the dot product, and the cross sign indicates the cross product. The sum of these vectors is then taken, and the absolute value is divided by 6 to obtain the total volume. After calculating the total volume and the three projected areas, the total volume is... V The ratios are calculated by comparing the projected areas with the three projected areas to obtain the shape-volume projection ratios in the three directions. The calculation formula is as follows: ; In the formula This represents the ratio of the shape and volume projection perpendicular to the XY plane. Represents the ratio perpendicular to the XZ plane. These represent ratios perpendicular to the YZ plane; these ratios reflect the extent of the shell's extension and volumetric efficiency in different directions.

[0049] Step 4.3: Extract the surface curvature field from the whole-brain electrophysiological topological shell, calculate the global root mean square value of the Gaussian curvature and the mean curvature of the surface curvature field, and use the global root mean square value as a morphological compactness index. Specifically, this includes: extracting the surface curvature field from the whole-brain electrophysiological topological shell; calculating the curvature at the vertex of each triangular facet on the shell surface by analyzing the change in the normal vector of all triangular facets in the neighborhood of that vertex; specifically, for each vertex, first collect all triangular facets adjacent to that vertex, calculate the unit normal vector of each facet; and calculate the Gaussian curvature at that vertex using the cotangent formula in discrete differential geometry. and mean curvature Gaussian curvature reflects the local degree of curvature of a surface, while mean curvature reflects the local average curvature direction of the surface. After calculating the curvature of all vertices, the global root mean square value of the Gaussian curvature is calculated for each vertex. and the global root mean square value of the mean curvature The calculation formula is: ; in The total number of vertices in the shell. and are respectively the first v The squares of the Gaussian curvature and the mean curvature at each vertex; and The product of and is used as an indicator of morphological compactness. The calculation formula is: ; The index integrates global fluctuation information of Gaussian curvature and average curvature. The larger the value, the more complex the shell surface, the more wrinkles, and the tighter the shape; the smaller the value, the flatter the shell surface and the more expansive the shape.

[0050] Step 4.4: Concatenate the morphological volume projection ratio and the morphological compactness index according to a preset dimensional order to form a multidimensional numerical group. This multidimensional numerical group is defined as the neuromorphic dynamics tuple. Specifically, this includes: calculating the three morphological volume projection ratios... 、 、 Compared with the calculated morphological compactness index The data is assembled according to a preset dimensional order to form a four-dimensional numerical group; the preset dimensional order is as follows: the first dimension stores the projection ratio perpendicular to the XY plane. The second dimension stores the projection ratio perpendicular to the XZ plane. The third dimension stores the projection ratio perpendicular to the YZ plane. The fourth dimension is the compactness index of storage form. The concatenation operation specifically involves creating a one-dimensional array with four elements, assigning values ​​sequentially as described above, and formally defining this four-dimensional numerical group as the neuromorphic dynamics tuple, denoted as . This neuromorphic dynamics tuple serves as the input for subsequent dynamic regulation steps, used to perform deviation comparison calculations with the pre-stored benchmark tuple, thereby driving the iterative update of rehabilitation intervention parameters.

[0051] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Call the pre-stored benchmark tuple, wherein the pre-stored benchmark tuple has the same dimensional structure and dimensions as the neuromorphic dynamics tuple. Specifically, this includes: calling the pre-set benchmark tuple; this benchmark tuple is a standard reference value calculated by offline collection of EEG data from a large number of healthy individuals under the same cognitive rehabilitation task, and following the same method as steps 4.1 to 4.4; the benchmark tuple has the same dimensional structure as the neuromorphic dynamics tuple output in step 4.4, i.e., both are four-dimensional numerical groups, and the dimensions of each dimension are consistent to ensure that subsequent difference calculations have physical meaning; the four values ​​in the benchmark tuple represent three standard values ​​of morphological volume projection ratios and one standard value of morphological compactness index corresponding to healthy individuals or ideal rehabilitation states; the benchmark tuple is denoted as... ,in 、、 These represent the standard projection ratios perpendicular to the XY plane, XZ plane, and YZ plane, respectively. This indicates the standard tightness index.

[0052] Step 5.2 involves performing a dimension-by-dimensional difference operation between the neuromorphic dynamics tuple and the benchmark tuple to obtain the deviation vector. Specifically, this includes: comparing the obtained neuromorphic dynamics tuple with the invoked benchmark tuple. Perform a dimension-by-dimensional difference operation; a dimension-by-dimensional difference operation refers to subtracting the values ​​in the same dimension from the values ​​in the two tuples, that is, subtracting the values ​​in the first dimension, the second dimension, the third dimension, and the fourth dimension, thus obtaining a four-dimensional difference vector; this difference vector is denoted as . The calculation formula is as follows: ; For ease of subsequent processing and description, the four components of the deviation vector are denoted as follows: ,Right now: ; Each component The sign of the value indicates the direction of deviation of the current measured value from the standard value. A positive value indicates that the measured value is higher than the standard value, and a negative value indicates that the measured value is lower than the standard value. The magnitude of the absolute value indicates the degree of deviation.

[0053] Step 5.3 involves norm normalization of the deviation vector to obtain the comprehensive deviation degree; simultaneously, it analyzes the sign and magnitude of deviation for each dimension of the deviation vector to identify the direction of deviation and the dominant dimension of deviation. Specifically, this includes: processing the obtained deviation vector... After norm normalization, the comprehensive deviation is obtained. Norm normalization refers to first calculating the Euclidean norm of the deviation vector, which is the square root of the sum of squares of its components; then dividing this norm by a preset normalization factor. This compresses the overall deviation into a standardized range, and its calculation formula is as follows: ; in This is the maximum possible norm value obtained statistically from the deviation vector norm distribution of a large number of samples in the historical database. Its purpose is to make... The value range is limited to between 0 and 1, which facilitates the unified processing of subsequent adjustment coefficients; at the same time, the deviation vector... The analysis extracts the sign and deviation of each dimension. Specifically, for each component... (in Record its positive or negative sign: If This is recorded as a positive deviation. This is recorded as a negative deviation. This is recorded as zero deviation; simultaneously, the absolute deviation magnitude of this component is calculated. By comparing the magnitudes of the four absolute deviations, the dimension with the largest deviation is identified, and its index is denoted as... (Values ​​can be 1, 2, 3 or 4), and record the deviation sign (positive or negative) corresponding to that dimension; Known as the deviation from the dominant dimension, it indicates the most significant source of deviation of the current neuromorphic dynamics tuple from the benchmark.

[0054] Step 5.4: Generate a difficulty adjustment instruction based on the comprehensive deviation degree, and generate a parameter correction instruction based on the deviation direction and the deviation dominant dimension. Specifically, this includes: generating a parameter correction instruction based on the obtained comprehensive deviation degree. The core of the difficulty adjustment command is a difficulty adjustment coefficient. This coefficient is used to scale the rehabilitation training intensity parameters subsequently. The calculation uses a linear mapping method to convert the overall deviation into the range of difficulty adjustment. The calculation formula is as follows: ; in This is a preset adjustment step size coefficient, the value of which is pre-set based on clinical experience and system response speed, typically ranging from 0.2 to 0.5; when hour, , indicating that no difficulty adjustment is needed; when hour, This indicates that the training difficulty needs to be increased to address the deviation between the current state and the standard value; theoretically... It can also be less than 1 (if the formula is changed to...) (And limits the lower limit), but by default in this system, a larger deviation indicates that the difficulty needs to be increased, therefore Simultaneously, based on the deviations from the dominant dimension identified in step 5.3... Its deviation sign, and generate parameter correction instructions; the parameter correction instructions contain a four-dimensional positive vector. Only the components corresponding to the dominant dimension are included. The component is assigned a non-zero value, while the other components are set to zero. The specific value is determined based on the deviation sign: if the deviation direction is positive, then set... If the deviation from the direction is negative, then set... ;in The preset correction coefficient, with a value range between 0.1 and 0.3, is used to control the step size of parameter correction in each iteration to avoid over-adjustment. The sign design of the correction vector ensures that the correction direction is always opposite to the deviation direction, that is, the parameter is decreased when there is a positive deviation and increased when there is a negative deviation, thereby driving the system to converge toward the benchmark.

[0055] Step 5.5: Using the difficulty adjustment instruction and parameter correction instruction as the basis for overwriting, replace or weight and fuse the adjustable parameter items corresponding to the initial rehabilitation intervention parameter set to form an updated rehabilitation intervention parameter set. Specifically, this includes updating the adjustable parameter items corresponding to the obtained initial rehabilitation intervention parameter set using the generated difficulty adjustment instruction and parameter correction instruction as the basis for overwriting. Specifically, the difficulty adjustment coefficient in the difficulty adjustment instruction... Used to override parameters related to intervention intensity, such as training difficulty level and stimulus intensity baseline, by multiplying the original parameter values ​​by... Rounding or truncating to the valid range; correction vector in parameter correction instructions. This is used to overwrite parameters directly related to morpho-volume projection ratio and compactness indices, such as target coordinate offsets and fine-tuning of stimulation frequency. Two update methods are employed: for discrete parameters, direct replacement is used, meaning the newly calculated parameter value overwrites the original value; for continuous parameters, weighted fusion is used, meaning the value is adjusted according to... Update, among which The preset fusion weights, ranging from 0.3 to 0.7, are used to balance the contributions between the original parameters and the correction instructions, avoiding abrupt changes. The original parameter value. The adjusted values ​​are calculated based on the correction instructions; after updating all adjustable parameter items, an updated rehabilitation intervention parameter set is formed for use in step 6.

[0056] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1, temporarily storing the updated rehabilitation intervention parameter set as the valid intervention parameter set for the current iteration cycle, specifically includes: storing the formed updated rehabilitation intervention parameter set into a pre-allocated dedicated storage area for the current iteration cycle in memory; the dedicated storage area for the current iteration cycle is organized according to the iteration rounds, with each round independently storing a complete parameter set; when storing, the current iteration round number and timestamp are automatically recorded, and the parameter set is marked as the valid intervention parameter set for the current iteration cycle, while overwriting the old parameter set stored in the previous round to save storage space; this valid intervention parameter set contains all rehabilitation training control parameters after difficulty adjustment and parameter correction, such as training difficulty level, stimulus pulse intensity, task switching frequency, target coordinate offset, etc., and its format is completely consistent with the initial rehabilitation intervention parameter set output in step 2.3, which facilitates unified processing in subsequent steps.

[0057] Step 6.2: Read the preset termination condition and determine whether the effective intervention parameter set of the current iteration cycle meets the preset termination condition. Specifically, this includes: reading the preset termination condition from the configuration file and comparing the effective intervention parameter set of the current iteration cycle with the termination condition item by item. The preset termination condition includes the following three types: First, comprehensive deviation. If the value is less than the preset convergence threshold, such as 0.05, it indicates that the current recovery status is close to the target benchmark and no further adjustment is needed; second, the current iteration round has reached the maximum allowed number of iterations, such as 10, to prevent infinite loops; third, a termination command is received from the user or medical staff. When making this judgment, the comprehensive deviation is first extracted from or recalculated from the set of effective intervention parameters. ,like If not stored in step 5.3, recalculate, simultaneously obtain the current number of completed iterations, and check for any external termination instruction flags; finally, compare these metrics with the corresponding thresholds: if... If the number of iterations is less than the convergence threshold, or the maximum number of iterations is reached, or the termination instruction flag is true, then the termination condition is considered met; otherwise, it is considered not met.

[0058] Step 6.3: If the preset termination conditions are met, the closed-loop intervention process is terminated, and the rehabilitation training completion status is obtained. Specifically, if the judgment result of step 6.2 is that the effective intervention parameter set of the current iteration cycle meets any one of the preset termination conditions, the closed-loop intervention process is terminated immediately. Upon termination, a stop command is first sent to the rehabilitation training terminal to immediately stop the stimulus output and cognitive task presentation, ensuring user safety. The effective intervention parameter set of the current iteration cycle, the last constructed whole-brain electrophysiological topology shell, and all intermediate data (such as neuromorphic dynamic tuples and deviation vectors of each round) are packaged and saved to non-volatile memory (wherein non-volatile memory is a storage medium that does not lose data after power failure, such as flash memory or solid-state storage devices, used to store the history of this rehabilitation training for a long time) as the history of this rehabilitation training. Finally, a rehabilitation training completion status identifier is generated, which includes the termination reason, such as convergence reaching the target, reaching the maximum number of iterations, or user-initiated termination, and is output to the user or medical staff through the user interface or communication interface.

[0059] Step 6.4: If the preset termination condition is not met, the effective intervention parameter set of the current iteration cycle is converted back into a new round of training drive instructions. Using the new round of training drive instructions as input, the processes of training drive instruction parsing, rehabilitation training terminal synchronization, EEG feedback signal acquisition, and whole-brain electrophysiological topology shell construction are re-executed to begin the next round of closed-loop iteration. Specifically, if the judgment result of step 6.2 indicates that the effective intervention parameter set of the current iteration cycle does not meet any preset termination condition, the next round of closed-loop iteration is initiated. Specifically, the effective intervention parameter set of the current iteration cycle is first read from memory, and each parameter item in the parameter set is converted into the instruction frame format agreed upon by the rehabilitation training terminal using the same instruction encoding method as in step 2.4, including frame header, parameter identifier, parameter value, and check bit. All instruction frames are then... The training drive instruction sequence is assembled into a new training drive instruction sequence according to the execution order. Using this new training drive instruction sequence as input, the program execution flow jumps to step 3.1. Starting from step 3.1, the following operations are executed again in sequence: training drive instruction parsing, real-time control parameter extraction, rehabilitation training terminal synchronous driving, continuous acquisition of EEG feedback signals, signal filtering and denoising preprocessing, time-frequency conversion and phase synchronization value calculation, neurophysiological topology node extraction, whole-brain electrophysiological topology shell construction, contour projection and curvature field analysis, morphological volume projection ratio and morphological compactness index calculation, neuromorphic dynamics tuple generation, deviation comparison with the benchmark, generation of difficulty adjustment instructions and parameter correction instructions, and parameter overwriting and updating, thereby forming a complete next round of closed-loop iteration. The above process is repeated until the preset termination condition is met in step 6.2 of a certain iteration.

[0060] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0061] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0062] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An intervention device that links brain health screening and rehabilitation training in a closed loop, characterized in that, include: The screening data processing module is used to obtain the risk type and risk level output by the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark. The scheme mapping module is used to perform scheme mapping operations with the strategy mapping benchmark as the strategy matching input to obtain the initial rehabilitation intervention parameter set, and convert the initial rehabilitation intervention parameter set into training driving instructions. The rehabilitation training and signal acquisition module is used to take training drive commands as terminal control inputs, synchronously execute rehabilitation training operations and acquire raw EEG feedback signals; The EEG signal processing module is used to perform frequency domain spatial mapping on the raw EEG feedback signal and dynamically extract the neurophysiological topological nodes corresponding to the phase synchronization extrema of specific brain regions; based on the spatial coordinate sequence of the neurophysiological topological nodes, a whole-brain electrophysiological topological shell that fits the topological distribution of the cortex is constructed. The topology analysis module is used to analyze the contour projection trajectory and surface curvature field of the whole brain electrophysiological topology shell, calculate the morphological volume projection ratio and morphological compactness index, and integrate them into a neuromorphic dynamics tuple. The dynamic control module is used to compare the neuromorphic dynamic tuple as the dynamic control input with the preset benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated by the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set. The closed-loop iteration module is used to take the updated rehabilitation intervention parameter set as the input for periodic iteration, forming a closed-loop intervention process until the preset termination conditions are met.

2. The intervention device for closed-loop linkage of brain health screening and rehabilitation training according to claim 1, characterized in that, Obtain the risk type and risk level output from the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark, including: The system receives raw output data from the brain health screening process, performs field parsing on the raw output data, and separates the risk type identifier and the risk level quantification value. The separated risk type identifiers are mapped to preset intervention strategy index codes, and the separated risk level quantification values ​​are subjected to dimensionless normalization to obtain the relative intensity scalar. The preset intervention strategy index code and the relative intensity scalar are combined into a two-dimensional strategy mapping tuple, and the two-dimensional strategy mapping tuple is used as the strategy mapping benchmark.

3. The intervention device for closed-loop linkage of brain health screening and rehabilitation training according to claim 2, characterized in that, The strategy mapping benchmark is used as the input for strategy matching to perform scheme mapping operations, resulting in an initial rehabilitation intervention parameter set. This initial rehabilitation intervention parameter set is then converted into training-driven instructions, including: The two-dimensional policy mapping tuple is unpacked to separate the intervention policy index code and the relative strength scalar. Using the intervention strategy index code as the key, a preset rehabilitation program database is retrieved, and the corresponding cognitive rehabilitation basic parameter template and neuromodulation basic parameter template are matched. The intervention strategy index code and relative intensity scalar are input into a preset topological response field model; the preset topological response field model calls the corresponding cortical region activation map according to the index code, and performs nonlinear modulation on the activation gradient field in the map according to the intensity scalar to obtain an initial rehabilitation intervention parameter set that matches the cognitive rehabilitation basic parameter template and the neuromodulation basic parameter template. Each parameter in the initial rehabilitation intervention parameter set is encoded into instructions and converted into a training-driven instruction sequence that can be directly parsed by the rehabilitation training terminal, ending with a training-driven instruction.

4. The intervention device for closed-loop linkage of brain health screening and rehabilitation training according to claim 3, characterized in that, The training-driven instructions are used as terminal control inputs to synchronously execute rehabilitation training operations and collect raw EEG feedback signals; the raw EEG feedback signals are spatially mapped in the frequency domain to dynamically extract the neurophysiological topological nodes corresponding to the phase synchronization extrema of specific brain regions. Based on the spatial coordinate sequence of neurophysiological topological nodes, a whole-brain electrophysiological topological shell conforming to the cortical topological distribution is constructed, including: The training drive instruction sequence is parsed into real-time control parameters, and the rehabilitation training terminal is synchronously driven to perform cognitive rehabilitation operations and neuromodulation operations based on the real-time control parameters. During the rehabilitation training process, raw EEG feedback signals from the user's scalp are continuously collected at a preset sampling rate. The raw EEG feedback signals are then filtered and denoised to obtain preprocessed EEG signals. The preprocessed EEG signal is converted from time to frequency to map the time-domain waveform to the frequency domain, and the phase synchronization value of each lead signal in the preset frequency band is calculated in the frequency domain. Extract spatial points in each brain region where the phase synchronization value reaches a preset extreme threshold, mark these spatial points as neurophysiological topological nodes, and record the three-dimensional spatial coordinates of each neurophysiological topological node. The three-dimensional spatial coordinates of all neurophysiological topological nodes are arranged into a spatial coordinate sequence according to the anatomical order of the cortex. The spatial coordinate sequence is then subjected to topological reconstruction of the convex boundary envelope of the spatial extremum points to construct a closed whole-brain electrophysiological topological shell that conforms to the topological distribution of the cortex.

5. The intervention device for closed-loop linkage of brain health screening and rehabilitation training according to claim 4, characterized in that, The contour projection trajectory and surface curvature field of the whole-brain electrophysiological topological shell are analyzed, and the morphovolume projection ratio and morphological compactness index are calculated for integration into a neuromorphodynamic tuple, including: The contour projection trajectory of the outer surface of the whole brain electrophysiological topological shell is extracted, and the contour projection trajectory is projected onto three orthogonal coordinate planes in three-dimensional space to obtain three sets of projection closed curves. Calculate the projected area enclosed by each set of projected closed curves to obtain three sets of projected areas; at the same time, calculate the total volume of the whole brain electrophysiological topological shell, and perform ratio calculations on this total volume with the three sets of projected areas to obtain the morphological volume projection ratios in three directions. The surface curvature field is extracted from the whole brain electrophysiological topological shell, and the global root mean square value of the Gaussian curvature and the mean curvature of the surface curvature field is calculated. The global root mean square value is used as the morphological compactness index. The morphological volume projection ratio and morphological compactness index are concatenated according to a preset dimensional order to form a multidimensional numerical group, which is then identified as the neuromorphic dynamics tuple.

6. The intervention device for closed-loop linkage of brain health screening and rehabilitation training according to claim 5, characterized in that, The neuromorphic dynamics tuple is used as the dynamic control input, and the deviation is compared with the preset target benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated using the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set, including: Call the pre-stored benchmark tuple, wherein the pre-stored benchmark tuple has the same dimensional structure and dimensions as the neuromorphic dynamics tuple; The deviation vector is obtained by performing a dimension-by-dimensional difference operation between the neuromorphic dynamics tuple and the benchmark tuple. The deviation vector is normalized to obtain the comprehensive deviation degree; at the same time, the positive and negative signs and deviation magnitudes of each dimension in the deviation vector are analyzed to identify the deviation direction and the dominant deviation dimension. Difficulty adjustment instructions are generated based on the comprehensive deviation degree, and parameter correction instructions are generated based on the deviation direction and the deviation dominant dimension. Based on the difficulty adjustment instructions and parameter correction instructions, the adjustable parameter items corresponding to the initial rehabilitation intervention parameter set are replaced or weighted and merged to form an updated rehabilitation intervention parameter set.

7. The intervention device for closed-loop linkage of brain health screening and rehabilitation training according to claim 6, characterized in that, The updated rehabilitation intervention parameter set is used as input for periodic iterations to form a closed-loop intervention process until preset termination conditions are met, including: The updated rehabilitation intervention parameter set is temporarily stored as the valid intervention parameter set for the current iteration cycle; Read the preset termination condition and determine whether the effective intervention parameter set of the current iteration cycle meets the preset termination condition; If the preset termination conditions are met, the closed-loop intervention process will be terminated, and the rehabilitation training completion status will be obtained. If the preset termination condition is not met, the effective intervention parameter set of the current iteration cycle will be converted back into a new round of training drive instructions. The new round of training drive instructions will be used as input to re-execute the process of training drive instruction parsing, rehabilitation training terminal synchronous drive, EEG feedback signal acquisition and whole brain electrophysiological topology shell construction, and start the next round of closed-loop iteration.

8. An intervention method that links brain health screening and rehabilitation training in a closed loop, wherein the method is applied to the device as described in any one of claims 1 to 7, characterized in that, include: Step 1: Obtain the risk type and risk level output from the brain health screening process, and encapsulate the risk type and risk level into a strategy mapping benchmark; Step 2: Use the strategy mapping benchmark as the strategy matching input to perform the scheme mapping operation to obtain the initial rehabilitation intervention parameter set, and convert the initial rehabilitation intervention parameter set into training driving instructions; Step 3: Use training-driven instructions as terminal control inputs to synchronously execute rehabilitation training operations and collect raw EEG feedback signals; perform frequency domain spatial mapping on the raw EEG feedback signals to dynamically extract neurophysiological topological nodes corresponding to phase synchronization extrema of specific brain regions; construct a whole-brain electrophysiological topological shell that conforms to the topological distribution of the cortex based on the spatial coordinate sequence of the neurophysiological topological nodes. Step 4: Analyze the contour projection trajectory and surface curvature field of the whole brain electrophysiological topological shell, calculate the morphological volume projection ratio and morphological compactness index, and integrate them into a neuromorphic dynamics tuple. Step 5: The neuromorphic dynamics tuple is used as the dynamic control input and the deviation is compared with the preset target benchmark to generate difficulty adjustment instructions and parameter correction instructions. The initial rehabilitation intervention parameter set is overwritten and updated through the difficulty adjustment instructions and parameter correction instructions to obtain the updated rehabilitation intervention parameter set. Step 6: Use the updated rehabilitation intervention parameter set as the input for periodic iteration, return to the terminal control input step of step 3 to form a closed-loop intervention process until the preset termination conditions are met.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to perform the method as described in claim 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in claim 8.

Images