Brain neuromodulation system for treating cognitive dysfunction and parameter optimization method thereof

Abstract

The application relates to a brain nerve regulation system for treating cognitive dysfunction and a parameter optimization method thereof, the method comprising the following steps: S1, providing a plurality of alternative stimulation schemes; S2, selecting one alternative stimulation scheme as an initial scheme for multiple treatments; S3, collecting resting-state electroencephalogram signals after each treatment, and calculating resting-state electroencephalogram power ratios after each treatment based on the resting-state electroencephalogram signals; S4, performing linear fitting on the resting-state electroencephalogram power ratios after the multiple treatments to obtain a slope; if the slope is less than or equal to 0, a new initial scheme is switched to perform multiple treatments on a target object; if the slope is greater than or equal to a first preset threshold, the current alternative stimulation scheme is maintained; if the slope is greater than 0 and less than the first preset threshold, an upward acceleration of the resting-state electroencephalogram power ratio changing with time is calculated, and a variance is calculated; S5, constructing a comprehensive evaluation index based on the slope, the acceleration and the variance; and selecting a next alternative stimulation scheme according to the comprehensive evaluation index; and S6, repeating steps S3-S5 for the same alternative stimulation scheme until a preset termination condition is met, and finally obtaining a treatment scheme.

Description

Technical Field

[0001] This application relates to the technical field of smart healthcare, and in particular to a brain neuromodulation system for treating cognitive impairment and a method for optimizing its parameters. Background Technology

[0002] Cognitive impairment (such as Alzheimer's disease, vascular dementia, and mild cognitive impairment) is a group of neurological diseases characterized by cognitive decline, severely impacting patients' quality of life and imposing a heavy care burden on society. In recent years, neuromodulation techniques based on electrical stimulation (such as transcranial direct current stimulation and transcranial alternating current stimulation) have been increasingly applied to clinical treatment research of cognitive impairment due to their non-invasiveness, safety, and minimal side effects. However, current electrical stimulation treatment protocols largely rely on standardized fixed parameters or physician experience-based selection, lacking sufficient consideration of individual differences in patients' brain function, leading to significant heterogeneity in clinical efficacy.

[0003] In the prior art, for example, CN119280719A discloses a transcranial focused ultrasound stimulation system for patients with cognitive impairment. It calculates the functional connectivity strength of brain regions through resting-state functional magnetic resonance imaging and identifies the weakest brain regions as personalized targets. It collects big data on the treatment of similar patients to train a deep prediction model, inputs the patient's target and cognitive impairment level to predict the treatment intensity, and determines the weakening coefficient through multi-model differences to obtain the final treatment intensity.

[0004] However, this method has significant drawbacks: First, the system is an open-loop structure, and once the target and intensity are determined, the treatment is fixed, lacking a real-time feedback and dynamic adjustment mechanism during the treatment process; second, efficacy evaluation is only performed by comparing before and after the completion of the full course of treatment, and if the plan is ineffective, the entire treatment time is wasted; third, the deep prediction model is highly dependent on large-scale, high-quality historical data, and the prediction accuracy is difficult to guarantee when data is scarce or there are large individual differences among patients; fourth, the classification of cognitive impairment levels has a certain degree of subjectivity and roughness, and intensity extrapolation based on level matching is difficult to achieve truly refined and individualized treatment. Summary of the Invention

[0005] The purpose of this invention is to provide a brain neuromodulation system for treating cognitive dysfunction and a method for optimizing its parameters, which partially solves or alleviates at least one of the above-mentioned shortcomings in the prior art, and can quickly screen out truly effective electrical stimulation programs for individuals, that is, screen out personalized neuromodulation parameters for individuals.

[0006] To solve the aforementioned technical problems, the present invention specifically adopts the following technical solution: A first aspect of the present invention is to provide a method for optimizing neuromodulation parameters in the treatment of cognitive dysfunction, comprising the following steps: S1: Provide multiple alternative stimulation schemes; the multiple alternative stimulation schemes are ranked according to the percentage of electric field space and lesion area coverage; S2: Select one alternative stimulation scheme from the plurality of alternative stimulation schemes as the initial scheme, and perform multiple treatments on the target subject; S3: Collect the resting-state EEG signal of the target object after each treatment, and calculate the resting-state EEG power ratio after each treatment based on the resting-state EEG signal; S4: Perform linear fitting on the resting-state EEG power ratio after multiple treatments to obtain the slope; if the slope is less than or equal to 0, switch to the first alternative stimulation scheme after the initial scheme as the new initial scheme to perform multiple treatments on the target subject, and execute step S3; if the slope is greater than or equal to a first preset threshold, maintain the current alternative stimulation scheme; if the slope is greater than 0 and less than the first preset threshold, calculate the acceleration of the resting-state EEG power ratio over time, and the variance after multiple consecutive treatments under the same alternative stimulation scheme; S5: Construct a comprehensive evaluation index based on the slope, acceleration, and variance; and select the next alternative stimulus based on the comprehensive evaluation index; S6: Repeat steps S3-S5 for the same alternative stimulation protocol until the preset termination condition is met and the final treatment plan is obtained.

[0007] Furthermore, step S5, which involves selecting the next alternative stimulus package based on the comprehensive evaluation index, specifically includes: Determine whether the comprehensive evaluation index is greater than or equal to a first preset index threshold. If the comprehensive evaluation index is greater than or equal to the first preset index threshold, the current attempt scheme is used as the plotting point, and at least one first-class candidate scheme with the same electrode position but different electrical stimulation parameters is matched for the current attempt scheme. The scheme search is performed in the at least one first-class candidate scheme along the slope growth direction with a preset search step size, and the next alternative stimulation scheme is selected from them using the Bayesian optimization algorithm. If the comprehensive evaluation index is between the first preset index threshold and the second preset index threshold, and the second preset index threshold is less than the first preset index threshold, then the Bayesian optimization algorithm is used to search for the next alternative stimulus scheme in the at least one first-class candidate scheme. When the comprehensive evaluation index is less than the second preset index threshold, the current attempt scheme is used as the plotting point to match at least one second-class candidate scheme with different electrode positions but the same electrical stimulation parameters, and the next alternative stimulation scheme is selected from them using a Bayesian optimization algorithm.

[0008] Furthermore, the initial scheme is the alternative stimulation scheme with the largest percentage of electric field space coverage area to lesion area among the multiple alternative stimulation schemes.

[0009] Furthermore, in step S5, the comprehensive evaluation index is... The calculation formula is: ; Where a is the slope under the current attempted scheme, σ 2 Let ξ be the variance of the current trial scheme, and ξ be the acceleration of the current trial scheme. The target slope is preset. For the maximum permissible variance, For the initial acceleration, , , The preset weighting coefficients are given, and they satisfy the following conditions: .

[0010] Further, in step S5, when the comprehensive evaluation index is between a first preset index threshold and a second preset index threshold, the step of using a Bayesian optimization algorithm to search for the next alternative stimulus scheme among the at least one first-class candidate schemes specifically includes: Based on historically tried solutions and their corresponding comprehensive evaluation indices, a probabilistic surrogate model is constructed to predict the comprehensive evaluation index and its confidence interval for all untried solutions in at least one first-class candidate solution. The score of each untried solution is automatically calculated using the acquisition function of the Bayesian optimization algorithm, and the solution with the highest score is selected as the next alternative stimulus solution.

[0011] Further, in step S5, when the comprehensive evaluation index is less than the second preset index threshold, the step of selecting the next alternative stimulus from the at least one second-type candidate scheme specifically includes: Start by trying the candidate test scheme with the largest percentage of coverage area from multiple second-category candidate schemes; Determine whether the number of attempts in at least one second-class candidate solution has reached a preset attempt threshold; If the preset attempt threshold is not reached, the next alternative stimulation scheme is selected from the at least one second-class candidate scheme according to the percentage of the coverage area of ​​the electric field space and the lesion area. If the preset attempt threshold has been reached, a Bayesian optimization algorithm is used to construct a probabilistic proxy model based on the attempted schemes and their corresponding comprehensive evaluation indices in at least one second-class candidate scheme. The comprehensive evaluation index and its confidence interval of all untried schemes in at least one second-class candidate scheme are predicted. The score of each untried scheme is automatically calculated through the acquisition function of the Bayesian optimization algorithm, and the scheme with the highest score is selected as the next alternative stimulus scheme.

[0012] Furthermore, the preset termination conditions include: the comprehensive evaluation index being greater than or equal to the preset target comprehensive evaluation index; or, there being no remaining options available to try among the multiple alternative stimulus options.

[0013] Further, in step S3, the resting-state EEG signal of the target subject after each treatment is collected, and the resting-state EEG power ratio after each treatment is calculated based on the resting-state EEG signal, including: Resting-state EEG signals were collected from the target subject after each treatment. The EEG power of the α-band, β-band, δ-band, and θ-band was extracted from the resting-state EEG signals to calculate the resting-state EEG power ratio after each treatment. The resting-state EEG power ratio was calculated as the sum of the α-band power and the β-band power divided by the sum of the δ-band power and the θ-band power.

[0014] A second aspect of the present invention is to provide a neuromodulation system for treating cognitive impairment, comprising: A storage module is configured to provide multiple alternative stimulation schemes, which are sorted by the percentage of electric field space to lesion area coverage. An electrical stimulation module is configured to select an initial protocol from the plurality of alternative stimulation protocols to perform multiple treatments on the target subject. The resting-state EEG power ratio calculation module is configured to collect the resting-state EEG signals of the target subject after each treatment, and calculate the resting-state EEG power ratio after each treatment based on the resting-state EEG signals; The treatment plan selection module is configured to perform linear fitting on the resting-state EEG power ratio after multiple treatments to obtain a slope. If the slope is less than or equal to 0, the current category of alternative stimulation plans is deemed invalid, and the first alternative stimulation plan after the initial plan is switched to as the new initial plan for multiple treatments on the target subject. If the slope is greater than or equal to a first preset threshold, the current alternative stimulation plan is maintained. If the slope is greater than 0 and less than the first preset threshold, the acceleration of the resting-state EEG power ratio over time and the variance after multiple consecutive treatments under the same alternative stimulation plan are calculated. A comprehensive evaluation index is constructed based on the slope, acceleration, and variance. The next alternative stimulation plan is selected based on the comprehensive evaluation index. The treatment plan determination module is configured to trigger the electrical stimulation module, the resting-state EEG power ratio calculation module, and the treatment plan selection module to perform corresponding functional operation steps each time the treatment plan selection module selects the next alternative stimulation plan, until the preset termination condition is met and the final treatment plan is obtained.

[0015] Further, the scheme selection module is specifically configured as follows: when it is determined that the comprehensive evaluation index is greater than the first preset index threshold, using the current attempt scheme as the plotting point, at least one first-class candidate scheme with the same electrode position but different electrical stimulation parameters is matched for the current attempt scheme, and scheme search is performed in the at least one first-class candidate scheme along the slope growth direction with a preset search step size, and the next alternative stimulation scheme is selected from them using a Bayesian optimization algorithm; when it is determined that the comprehensive evaluation index is between the first preset index threshold and the second preset index threshold, and the second preset index threshold is less than the first preset index threshold, the next alternative stimulation scheme is searched from the at least one first-class candidate scheme using a Bayesian optimization algorithm; when it is determined that the comprehensive evaluation index is less than the second preset index threshold, using the current attempt scheme as the plotting point, at least one second-class candidate scheme with different electrode position but the same electrical stimulation parameters is matched for the current attempt scheme, and the next alternative stimulation scheme is selected from them using a Bayesian optimization algorithm.

[0016] Beneficial technical effects: In the electrical stimulation therapy for cognitive impairment, rapidly identifying an effective treatment plan is a fundamental requirement of clinical practice; however, the treatment outcomes for cognitive impairment exhibit significant individual differences. Due to variations in each patient's brain anatomy and neural network connectivity patterns, standardized, fixed treatment plans often yield inconsistent results. Therefore, there is an urgent clinical need for "personalized" treatment plans that can be rapidly adjusted based on real-time individual feedback.

[0017] In view of this, this application proposes a method that relies solely on real-time post-treatment resting-state EEG data of the target subject for data analysis to obtain parameters characterizing the current treatment effect, such as slope and comprehensive evaluation index. Based on this slope and comprehensive evaluation index, a truly effective electrical stimulation protocol for the target subject is rapidly selected from a series of candidate protocols, including the corresponding electrical stimulation parameters and electrode positions (both of which can be referred to as neuromodulation parameters). This avoids the reduced response speed caused by introducing any other modality of data (task-based EEG, CT / MRA, scale test results). Furthermore, this application can quickly decide and execute the next round of alternative stimulation protocols based on the patient's immediate resting-state EEG data, achieving "real-time closed-loop" modulation, thereby improving the response speed.

[0018] Generally, relying on single resting-state EEG data for assessment carries a high probability of misjudgment. To mitigate this risk, measures are often taken. For example, a single increase in the R value might be accidental (e.g., due to adequate rest), or the R value might show an upward trend in the first few treatments but remain unchanged or show very little change in subsequent treatments, indicating a treatment plateau or bottleneck. Therefore, to reduce misjudgment, task-based EEG data or multimodal EEG data such as CT or MRA, or even scale assessment data, are typically introduced for comprehensive evaluation. However, all of these methods require a considerable amount of time to acquire the data, affecting the equipment's response efficiency. Furthermore, the comprehensive processing of different modalities of data increases algorithmic complexity. For instance, introducing a new data modality (such as task-based EEG, imaging data, or scales) not only means extended acquisition time and increased equipment requirements but also an exponential increase in algorithmic complexity. In addition, achieving temporal synchronization, spatial registration, feature fusion, and weight allocation for data from different modalities is itself a highly challenging technical problem, and conflicting information between different modalities can lead to decision-making biases.

[0019] Therefore, in order to improve response efficiency and simplify the algorithm architecture while ensuring the accuracy of the evaluation effect (i.e., reducing the probability of misjudgment to a certain extent), this application combines the slope, acceleration, and variance of the power ratio of resting-state EEG data after multiple treatments with linear fitting. These are evaluated from three dimensions: slope corresponds to the trend dimension, reflecting the average improvement rate of the brain's excitation / inhibition ratio in continuous treatment; acceleration corresponds to the curvature dimension, reflecting the rate of change of the slope itself, with positive acceleration indicating accelerated release of therapeutic effect (adaptation period) and negative acceleration indicating that neuronal response is close to saturation (bottleneck / plateau period); variance corresponds to the stability dimension, reflecting the degree of fluctuation of the R value, with low variance indicating stable and reliable therapeutic effect and high variance indicating that the data is dominated by noise such as emotion, sleep, and EEG artifacts. Based on the evaluation results, a decision can be made quickly to adjust the treatment plan, for example, by adjusting only the electrode position of the current treatment plan (i.e., spatial dimension adjustment) or by adjusting only the electrical stimulation parameters of the current treatment plan (i.e., cross-category adjustment).

[0020] In other words, without introducing any other modal data (task-based EEG, CT / MRA, scale test results), this application maps a simple one-dimensional numerical value into a three-dimensional state space (trend-slope + momentum-acceleration + confidence-variance) by mining the dynamic evolution characteristics of a single resting-state index R value in the time dimension. This achieves both high timeliness and high accuracy, and the algorithm structure is simple, making it easier to implement and promote.

[0021] Compared to the open-loop structure of the prior art, this application uses a closed-loop approach, adjusting the treatment in real time based on the patient's treatment effect (such as slope and comprehensive evaluation index). Furthermore, while the prior art relies on data from a large number of users to find a treatment plan universally applicable to a group, this application uses a series of existing alternative plans (applicable to a group or specific to a particular patient) as feedback from the target patient's previous real-time treatment effect to assist in the selection of alternative plans. This allows for the adjustment of neuromodulation parameters, such as electrode placement or electrical stimulation parameters, resulting in a personalized plan adapted to the target patient—that is, personalized neuromodulation parameters. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0023] Figure 1 This is a flowchart of a specific embodiment of a method for optimizing brain neuromodulation parameters for treating cognitive dysfunction according to this application.

[0024] Figure 2 This is a functional block diagram of a specific embodiment of a brain neuromodulation system for treating cognitive impairment according to this application.

[0025] Figure 3 This is a schematic diagram illustrating the distribution of two electrode pairs in an electrical stimulation protocol determined for patient D using the method of the present invention, as shown in the international 10-20 system electrode location diagram.

[0026] Figure 4 This is a schematic diagram illustrating some electrical stimulation parameters (electrical stimulation time, frequency, channel number, current intensity) in an electrical stimulation protocol determined for patient D using the method of the present invention.

[0027] Figure 5 This is a graph representing the statistical results of the ADASCog test (cognitive component) of the Alzheimer's Disease Assessment Scale for 17 subjects diagnosed with cognitive impairment, performed before, after, and one month after the corresponding electrical stimulation treatment regimen.

[0028] Figure 6This is a statistical chart reflecting the scores obtained from the Clinical Dementia Rating Scale test by 17 subjects diagnosed with cognitive impairment before, after, and one month after the corresponding electrical stimulation treatment.

[0029] Figure 7 This is a statistical chart reflecting the scores of 17 subjects diagnosed with cognitive impairment, obtained from the Monterey Cognitive Assessment Scale test before, after, and one month after the corresponding electrical stimulation treatment. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0031] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0032] In this document, the terms "upper," "lower," "inner," "outer," "front," "rear," "one end," and "the other end," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the present invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0033] In this document, unless otherwise explicitly specified and limited, the terms "installed," "equipped with," and "connected," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0034] In this document, "and / or" includes any and all combinations of one or more of the listed related items.

[0035] In this article, "multiple" means two or more, that is, it includes two, three, four, five, etc.

[0036] In one specific embodiment, this application includes the following steps: S1: Provide multiple alternative stimulation schemes, which are ranked according to the percentage of electric field space and lesion area coverage; S2: Select one alternative stimulation scheme from the plurality of alternative stimulation schemes as the initial scheme, and perform multiple treatments on the target subject; preferably, the alternative stimulation scheme ranked first is selected as the initial scheme; S3: Collect the resting-state EEG signal of the target object after each treatment, and calculate the resting-state EEG power ratio after each treatment based on the resting-state EEG signal; S4: Perform linear fitting on the resting-state EEG power ratio after multiple treatments to obtain the slope; if the slope is less than or equal to 0, it indicates that the currently attempted alternative stimulation scheme (such as the initial scheme) is ineffective or the treatment effect is not obvious, then switch to the next alternative stimulation scheme as the new initial scheme, and perform multiple treatments, that is, execute step S2 with the new initial scheme, and then execute step S3; if the slope is greater than or equal to the first preset threshold, it indicates that the current alternative stimulation scheme is an effective scheme, then maintain the current alternative stimulation scheme, and use it as the final scheme for the target object; if the slope is greater than 0 and less than the first preset threshold, then calculate the rising acceleration of the resting-state EEG power ratio over time, and the variance after multiple consecutive treatments under the same alternative stimulation scheme; S5: Construct a comprehensive evaluation index based on the slope, acceleration, and variance; and select the next alternative stimulation scheme according to the comprehensive evaluation index. Preferably, step S5 specifically includes: when the comprehensive evaluation index is greater than a first preset index threshold, using the current attempt scheme as the plotting point, matching at least one first-class candidate scheme with the same electrode position but different electrical stimulation parameters for the current attempt scheme, and performing a scheme search in the at least one first-class candidate scheme along the slope growth direction with a preset search step size to select the next alternative stimulation scheme (i.e., the next attempt scheme); preferably, performing a scheme search with a preset step size will yield multiple schemes, and a Bayesian optimization algorithm can be used to determine the one with the highest score as the scheme to be tried (i.e., the target scheme); specifically, see the following specific embodiments; since the comprehensive evaluation index is greater than the first preset index threshold, that is, the current attempt scheme has shown excellent efficacy, therefore, using the current attempt scheme as the anchor point, an efficient local search is performed along the slope growth direction to find possible local optimal solutions; that is, based on the existing good efficacy, by narrowing the search range, the optimization speed is accelerated and unnecessary parameter attempts are reduced; When the comprehensive evaluation index is between a first preset index threshold and a second preset index threshold, and the second preset index threshold is less than the first preset index threshold, a Bayesian optimization algorithm is used to search for the next alternative stimulus scheme in at least one first-class candidate scheme. Since the comprehensive evaluation index is between the two thresholds, that is, although the current attempt scheme shows some efficacy, it is not beneficial enough. Therefore, in the case where the efficacy is acceptable but there is uncertainty, a broader exploration is carried out within the current category to determine whether there is a better parameter combination, while avoiding prematurely falling into local optima. That is, compared with the search method when the comprehensive evaluation index is greater than the first preset index threshold, the search method without limiting the search direction is actually a global search in a larger range, which helps to discover other good parameter combinations that may exist at the current electrode position and increases the probability of finding the global optimum. When the comprehensive evaluation index is less than the second preset index threshold, the current attempt scheme is used as the plotting point to match at least one second-class candidate scheme with different electrode positions but the same electrical stimulation parameters. The next alternative stimulation scheme (i.e., the scheme to be tried) is searched among the at least one second-class candidate scheme using a Bayesian optimization algorithm. Since the comprehensive evaluation index is less than the second preset index threshold, it indicates that the current scheme is obviously not effective, that is, the current electrode position may not be the optimal target point. Therefore, it is necessary to relocate to a potentially more effective target point region and perform a cross-class global search. That is, by changing the electrode position, we can break away from the currently poor local region, explore a completely new parameter space, and avoid continuously wasting resources in the ineffective region.

[0037] S5: Repeat steps S3-S5 for the same alternative stimulation protocol until the preset termination conditions are met, and obtain the final treatment plan. The preset termination conditions include: the slope is greater than a first preset threshold, or the comprehensive evaluation index is greater than or equal to the preset target comprehensive evaluation index; or there is currently no treatment plan to try.

[0038] The solution of this application will be described in detail below with reference to specific embodiments. In one specific embodiment, a specific embodiment of the brain neuromodulation parameter optimization method for treating cognitive dysfunction that can be applied to this application will be disclosed, which includes the following steps: Step 101: Generate and sort multiple alternative stimulus schemes in the alternative stimulus scheme library.

[0039] This application first constructs a personalized library of alternative stimulation protocols for the target subject. The library contains multiple alternative stimulation protocols, each including at least electrode location information and electrical stimulation parameter information. The electrode location information is used to determine the attachment position of one or more stimulation electrodes on the scalp surface of the target subject (see [link to documentation]). Figure 3The diagram shows the electrode positions of the international 10-20 system. When a voltage is applied according to the corresponding electrical stimulation parameter information, an electric field space is formed between the multiple electrodes to target the target area (this is prior art and will not be described in detail here). Preferably, the electrical stimulation parameter information includes, but is not limited to, stimulation current intensity, stimulation frequency, pulse width, stimulation duration, and stimulation waveform type.

[0040] In some embodiments, the generation of the alternative stimulation protocol library is based on individualized brain imaging data of the target subject. For example, this is achieved by acquiring medical images of the target subject's head, including but not limited to structural magnetic resonance imaging and / or computed tomography (CT) scans. Based on the image data, a three-dimensional model of the target subject's head is reconstructed, and the three-dimensional spatial coordinates of the lesion region are annotated using medical image processing software.

[0041] Furthermore, an individualized head conductivity model of the target object was constructed using existing finite element methods. This model included different conductivity parameters for the scalp, skull, cerebrospinal fluid, gray matter, white matter, and lesion areas. Based on the head conductivity model, electric field simulation calculations were performed for each candidate electrode location combination to obtain the electric field distribution vector field generated by each candidate scheme in the target object's brain. ,in This represents the spatial location of any point within the brain. Specifically, the generation scheme for the alternative stimulus scheme library is existing technology, such as patent document CN119565030A, and therefore will not be elaborated upon here.

[0042] To quantify the spatial coverage effect of each alternative stimulus on the lesion region, this application uses the percentage of electric field space covered by the lesion region as the core ranking metric. Let the set of volumes or surface areas of the lesion region in three-dimensional space be denoted as . For ease of calculation, the projected area of ​​the lesion region on the cortical surface or standardized brain surface is taken as the target region area. For any alternative stimulation protocol... The brain electric field distribution obtained through simulation calculation is as follows: (Calculating the distribution of electric fields in the brain using simulation software is existing technology and will not be described in detail here.) Setting an effective threshold for electric field strength. When the amplitude of the electric field intensity at a certain point in the brain When it is determined that the point is covered by the effective electric field space, then the effective electric field coverage area is... Defined as: Then the percentage of coverage area The calculation formula is: in, This represents a function for calculating the two-dimensional surface area or projected area; the numerator represents the intersection area of ​​the effective electric field coverage area and the lesion area, and the denominator represents the total area of ​​the lesion area. Preferably, for different lesion depths and locations, the effective threshold of the electric field strength is... It can be adjusted according to the individualized neural excitability threshold of the target individual, and its value range is generally [value range missing]. to .

[0043] The percentage of coverage area corresponding to each of the alternative stimulus options was calculated. Subsequently, multiple alternative stimulus packages were arranged according to... The values ​​are sorted in descending order from highest to lowest, and the sorting results are stored in the storage module, forming an ordered library of candidate stimulus schemes. After sorting, the candidate stimulus scheme with the largest coverage percentage, i.e., satisfying... The plan, in which The total number of alternative stimulation protocols is marked as the initial protocol for subsequent first treatment attempts. The physical significance is that the higher the proportion of the lesion area covered by the electric field space, the higher the spatial precision of the neuromodulation's effect on the target point, and the more likely it is to produce an observable EEG response within fewer treatment attempts, thereby reducing the time cost and patient burden caused by ineffective attempts.

[0044] Step 102: Execute the initial protocol and collect EEG data.

[0045] After constructing and sorting the alternative stimulation protocol library, the top-ranked alternative stimulation protocol is read from the storage module as the initial protocol. The initial protocol is the percentage of electric field space and lesion area coverage among multiple alternative stimulation protocols. The most powerful alternative stimulus, i.e., the one that satisfies: The procedure involves attaching at least one stimulating electrode (e.g., a positive electrode) and a corresponding reference electrode (e.g., a negative electrode) to the target scalp according to the electrode position information of the initial plan. Electrical stimulation is then applied to the target scalp according to the electrical stimulation parameters specified in the initial plan. These parameters include stimulation current intensity (mA), stimulation frequency (Hz), pulse width (μs), stimulation duration (minutes), and stimulation waveform type (including but not limited to square wave, sine wave, or biphasic pulse wave). During each electrical stimulation application, the stimulation signal is continuously output until the preset stimulation duration is reached, completing a single treatment session.

[0046] To evaluate the therapeutic effect of the initial protocol, resting-state EEG signals were collected from the target subjects within predetermined time windows before and after each treatment. Specifically, multiple EEG recording electrodes were attached to the scalp surface of the target subjects according to the international EEG electrode placement standard 10-20 system to collect spontaneous activity signals from the whole brain or specific brain regions. Resting-state EEG signal acquisition was performed when the target subjects were awake, eyes closed, relaxed, and avoiding any specific cognitive tasks. Each acquisition session lasted at least 5 minutes to ensure sufficiently stable and statistically significant EEG data. The resting-state EEG signals collected before each treatment were recorded as the preprocessing baseline signal, and the resting-state EEG signals collected after each treatment were recorded as the postprocessing response signal. To ensure data quality, necessary preprocessing was performed on the acquired resting-state EEG signals before signal analysis. Preprocessing includes using a bandpass filter to remove low-frequency drift and high-frequency noise, with the passband range set from 0.5Hz to 70Hz; further, a notch filter is used to remove 50Hz power frequency interference; and independent component analysis (ICA) is used to remove artifacts such as electrooculogram (EOG), electromyogram (EMG), and electrocardiogram (ECG) signals to obtain a pure EEG signal. The preprocessed pure EEG signal is transmitted to a storage module and stored in association with the corresponding number of treatments, protocol index, and acquisition time point for subsequent calculation of the resting-state EEG power ratio. The above stimulation and acquisition process is repeated according to a preset rapid screening window period. Specifically, this rapid screening window period can be set based on the treatment cycle.

[0047] In a preferred embodiment of this application, the preset treatment cycle is 14 days, including 10 days of stimulation and 4 days of rest, i.e., two consecutive days of stimulation followed by one day of rest. Each stimulation session consists of 3-6 sessions in the morning and afternoon, each lasting 20 minutes. However, for subjects undergoing electrical stimulation for the first time, or for subjects who have not yet found an effective treatment plan, continuous stimulation for 3-6 days (i.e., a rapid screening window) is conducted to quickly identify an effective plan (though it may not be the optimal treatment). The EEG data before and after each stimulation within these 3-6 days are used as the basis for subsequent evaluation. The length of the rapid screening window can be adjusted based on the target subject's tolerance and clinical response, but should not be less than 3 days to ensure a sufficient sample size for subsequent linear fitting.

[0048] In each treatment session, resting-state EEG signals were first collected and stored before treatment, followed by electrical stimulation. After stimulation, resting-state EEG signals were collected and stored again after treatment. Thus, for each treatment session under the current alternative stimulation protocol... ( ,in (This represents the total number of treatments under the current protocol), and each treatment session recorded a pair of resting-state EEG signals, i.e., the pre-treatment signals. and post-treatment signals Multiple pairs of resting-state EEG signals form the raw data basis for evaluating the efficacy of the current alternative stimulation protocols, which are then used for frequency band power analysis and efficacy trend determination in subsequent steps. If the target patient experiences intolerance, adverse events, or equipment malfunctions during treatment, the current protocol is immediately terminated and the abnormal state is recorded. Simultaneously, the system automatically jumps to the next protocol in the alternative stimulation protocol library and restarts the above process to ensure the safety of the target patient and the robust operation of the system.

[0049] Step 103: Calculate the ratio of resting-state EEG power (R value) before and after each treatment.

[0050] Frequency band power analysis was performed on resting-state EEG signals collected before and after each treatment. The preprocessed, clean EEG signals were time-domain signals. These signals were first converted to the frequency domain using a Fast Fourier Transform to obtain the power spectral density distribution. Specifically, for the first... Pre-treatment signals for the next treatment and post-treatment signals Extract the middle segment of the acquisition period and the continuous segment with the most stable signal quality, respectively. Second data ( The values ​​range from 60 to 120 seconds. A Hanning window is used to window the truncated signal to reduce spectral leakage. Then, the power spectral density of each frequency band is calculated using a Fast Fourier Transform. Based on this, the power values ​​of four EEG frequency bands closely related to cognitive function and the balance between excitation and inhibition in the brain are extracted. These four frequency bands include: frequency band frequency band frequency band and Frequency bands. For each frequency band, the total power value of that band is obtained by integrating or summing the power spectral density of all frequency components within that band. The [number of]th [frequency band]... Before the second treatment Frequency band power is denoted as Before treatment Frequency band power is denoted as Before treatment Frequency band power is denoted as Before treatment Frequency band power is denoted as Similarly, the power of the corresponding frequency band after treatment is denoted as... , , and .

[0051] To eliminate the influence of individual differences and baseline fluctuations on absolute power values, this application uses the frequency band power ratio as the core evaluation indicator. Specifically, for each treatment, the resting-state EEG power ratio is calculated before and after treatment, and the resting-state EEG power ratio is defined as follows: Frequency band power and Sum of frequency band power divided by Frequency band power and The sum of power across frequency bands. The first... The resting-state EEG power ratio before the second treatment was recorded as follows: The calculation formula is as follows: ; The first The resting-state EEG power ratio after each treatment was recorded as follows: The calculation formula is as follows: ; The above ratio The physiological significance is as follows: frequency band and The power of a frequency band reflects the excitatory activity of the cerebral cortex, while frequency band and The power of a frequency band reflects inhibitory activity or impaired function. Patients with cognitive impairment typically exhibit... and Relatively enhanced frequency band power and The power in the frequency band is relatively reduced, therefore The lower the value, the more the brain's excitation / inhibition balance tends to be dominated by inhibition, indicating a pathological state. An elevated value indicates that the brain's excitation / inhibition balance is returning to normal, reflecting the effectiveness of the treatment.

[0052] Furthermore, to quantify the immediate improvement effect of a single treatment, the calculation of the first... Instantaneous changes after each treatment Defined as the difference between the resting-state EEG power ratio after treatment and the resting-state EEG power ratio before treatment: ; It reflects the immediate change in the brain's excitation / inhibition balance after a single electrical stimulation treatment. This indicates that the brain's excitability was relatively enhanced after the treatment. This indicates that the inhibitory effect in the brain has worsened or the excitability has decreased after treatment. This indicates that the treatment did not cause significant changes in neurophysiological state.

[0053] Once all of the current alternative stimulus options (i.e., the current trial option) have been completed... After each treatment, the ratio of resting-state EEG power was recorded. ( Arranged in chronological order, forming a length of Post-treatment ratio sequence This sequence reflects the cumulative change trend of post-treatment EEG status with increasing number of repeated treatments under the current alternative stimulation protocols. The post-treatment ratio sequence was stored and used as the basis for subsequent linear fitting and trend determination.

[0054] Step 104: Comprehensive evaluation and optimization of therapeutic effects.

[0055] Obtain all of the current alternative stimulus packages Resting-state EEG power ratio sequence after treatment Then, linear fitting was performed on the sequence to determine the overall trend of the resting-state EEG power ratio after treatment as the number of treatments increased. Specifically, the number of treatments was used as the basis for the analysis. ( () is used as the independent variable, and the ratio of resting-state EEG power after each treatment is used as the independent variable. Using the least squares method as the dependent variable, a univariate linear fit was performed to obtain the linear regression equation: ; in, The slope of the fitted line, The intercept. The slope. The calculation formula is: ; in, This represents the average number of treatments. This represents the mean of the resting-state EEG power ratio sequence after treatment. Slope This reflects the marginal increase in the resting-state EEG power ratio after treatment with increasing number of repeated treatments, under the current alternative stimulation protocols. This indicates that the brain's electroencephalogram (EEG) status showed an upward trend with the increase in the number of treatments, meaning that the brain's excitation / inhibition balance tended to normalize. This indicates that the brainwave state has stagnated. This indicates a downward trend in brain electrical activity, meaning that brain function is actually deteriorating.

[0056] Based on the slope The value is used to make a preliminary judgment on the effectiveness of the current alternative stimulus measures.

[0057] In some embodiments, if the slope If the current stimulus is ineffective, it indicates that the therapeutic electric field of the current protocol has failed to effectively target the brain region or that the stimulation parameters have failed to induce an effective neural response. In this case, the evaluation of the current protocol is terminated, and the second-ranked alternative stimulus protocol is switched to. Of course, if the slope of the second-ranked protocol is still less than or equal to 0 after multiple treatments, the third-ranked alternative stimulus protocol is tried. That is, each alternative stimulus protocol is traversed sequentially according to its ranking until an effective protocol (i.e., its corresponding slope > 0) is found, and this becomes the anchor point for searching and finding new protocols to try.

[0058] In some embodiments, if the slope ,in If the first preset threshold (i.e., the clinically effective threshold) is met, the current alternative stimulation regimen is determined to be effective, and there is no need to switch to other regimens. The current regimen is maintained as the final treatment plan for the target patient for periodic treatment. First preset threshold The range of values ​​is to Preferably, the value is taken as The specific value of this threshold is preset based on the target subject's baseline level and clinical response characteristics. In some embodiments, if the slope satisfy If the current alternative stimulation regimen has potential efficacy but the rate of improvement has not reached the clinically effective threshold, further refined evaluation is needed to determine the next optimization strategy. Specifically, when the determination result is... Then, the rising acceleration of the resting-state EEG power ratio sequence after treatment was further calculated. The rising acceleration was defined as the slope. The rate of change with increasing treatment frequency was specifically measured by analyzing the resting-state EEG power ratio sequence after treatment. Second-order polynomial fitting yielded the following: ; in, The coefficient of the quadratic term represents the acceleration due to the rising speed. That is, take as twice, that is acceleration upon ascent The physiological significance is as follows: This indicates that the rate at which the brainwave status improves after treatment is accelerating with the number of treatments, meaning that the current treatment plan is showing an accelerated improvement trend. This indicates that the rate of increase is slowing down, meaning that the effectiveness of the current plan is approaching saturation or has reached a bottleneck. This indicates that the EEG status rises steadily at an approximately constant rate after treatment. A negative acceleration is an important warning signal that the current treatment's effectiveness has reached a plateau or bottleneck.

[0059] Simultaneously, calculations are performed under the same alternative stimulus regimens for consecutive Resting-state EEG power ratio after treatment variance The formula for calculating variance is: ; In obtaining the slope acceleration upon ascent and variance Based on this, construct a comprehensive evaluation index. This is used to comprehensively assess the overall effectiveness of current alternative stimulus packages. Comprehensive Assessment Index The calculation formula is: ; in, For the target slope, The maximum permissible variance is preset. For the initial acceleration, , , The preset weighting coefficients are given, and they satisfy the following conditions: Target slope This refers to the maximum linear slope value achieved by the target subject during historical treatment cycles, or a preset clinically effective slope threshold. Maximum permissible variance This is used to evaluate whether the dispersion of the resting-state EEG power ratio after treatment is within an acceptable range. Its value is set based on the baseline noise level of the target subject. The baseline noise level is calibrated by repeatedly acquiring resting-state EEG signals before the first treatment and calculating the standard deviation of their power ratios. Initial acceleration For the first time, a linear slope was detected. satisfy The corresponding acceleration value at that moment, that is, when the system first determines that the current plan has potential efficacy, the acceleration value at that moment is recorded as the benchmark for subsequent acceleration assessment.

[0060] The calculated comprehensive evaluation index P is compared with the preset index threshold to guide the next step of scheme selection.

[0061] Specifically, when the comprehensive evaluation index P is greater than the first preset index threshold τ1, the comprehensive efficacy of the current alternative is judged to be excellent, indicating that the therapeutic effect, stability and development trend of the current alternative are good and there is still potential for further improvement. At this time, the electrode position should be kept unchanged, but the electrical stimulation parameters should be changed to perform a local search: take the current trial or the initial alternative as the plotting point, match it with at least one first-class candidate alternative with the same electrode position but different electrical stimulation parameters, and search for alternatives in at least one first-class candidate alternative along the slope growth direction with a preset search step size to obtain a subset of candidate alternatives (including at least one first-class candidate alternative). Then, the Bayesian optimization algorithm is used to find the alternative stimulation alternative with the highest score from the subset of candidate alternatives as the alternative to be tried.

[0062] For example, if the slope of the current attempt is greater than 0 but less than the first preset threshold, and when it is determined that the comprehensive evaluation index P is greater than the first preset index threshold, before using the Bayesian optimization algorithm to search for the next attempt, if increasing the current intensity of the current attempt brings better therapeutic effect compared to the previous attempt (e.g., the slope is greater than 0 and increases), then "increasing the current intensity" is the direction of slope increase. Searching with a preset search step size means that the difference between the current intensity amplitude of the current attempt and the current intensity amplitude of the previous attempt is used to continue searching for a solution among multiple first-class candidate solutions, resulting in a subset of candidate solutions. This means constraining the search space of the Bayesian optimization algorithm to the neighborhood of the current electrode position to enter a local fine-tuning mode. Based on the constrained search space, the score of each candidate stimulus solution in the subset of candidate solutions is recalculated through the acquisition function, and the solution with the highest score is selected as the next candidate stimulus solution to be tried (i.e., the local fine-tuning solution).

[0063] In some embodiments, the preset search step size is a "multiple" or "fraction" of the interval between each alternative stimulus: when searching along a single parameter direction, the search step size is usually an integer multiple or fraction of the preset parameter interval; for example, if the parameter interval between each alternative stimulus is 1mA, then the step size can be 1mA, 2mA or 0.1mA.

[0064] Of course, in other embodiments, compared to the previous attempt, the increased electrical stimulation parameters in the current attempt can be any other electrical stimulation parameter (such as frequency) in addition to current intensity, or any number of all electrical stimulation parameters.

[0065] Of course, if none of the local fine-tuning schemes attempted within the neighborhood of the current electrode location achieve a linear slope for the resting-state EEG power ratio... Reaching the first preset threshold If the local fine-tuning mode is exited, the scheme with the largest comprehensive evaluation index is selected as the new anchor point, and a cross-category search is performed (see subsequent embodiments).

[0066] For example, if the comprehensive evaluation index of the current attempt scheme (not the initial scheme) is greater than or equal to the first preset index threshold, then along the current parameter adjustment direction (such as the direction of increased current intensity or frequency compared to the previous attempt scheme), multiple candidate schemes with different electrical stimulation parameters but the same electrode position are found with a preset parameter step size (such as the current step size increasing from 0.1mA to 0.3mA, and / or the frequency step size increasing from 0.5Hz to 1.5Hz). Taking the adjustment of current intensity as an example, a subset of candidate schemes is obtained with a preset search step size of 0.1mA: candidate A1-5.1mA; candidate B1-5.2mA, etc. Through the above-mentioned preset parameter step size as the search step size strategy, the system can accelerate the exploration of better parameter combinations along the verified effective direction, under the premise that the current scheme performs well, and avoid making too many redundant minor attempts near the optimal solution, thereby improving search efficiency and accelerating convergence speed.

[0067] However, for the initial scheme, since there is no historical adjustment direction to refer to, both positive and negative directions are generated simultaneously with a preset step size to search for multiple schemes. Then, a Bayesian optimization algorithm is used to find the candidate stimulus scheme with the highest score as the scheme to be tried. For example, taking the adjustment of current as an example, a subset of candidate schemes in both positive and negative directions is generated simultaneously with a preset search step size: increasing direction - candidate A2 - 5.1mA; candidate A3 - 5.2mA...; decreasing direction - candidate B2 - 4.9mA; candidate B2 - 4.8mA...

[0068] Of course, even when searching with a preset step size, it is necessary to set an upper limit on the number of searches (that is, not endlessly trying multiple first-type candidate solutions, but trying multiple times with a preset threshold). For example, 10-20 times. Furthermore, provided that P is greater than the first preset exponential threshold, a larger P allows for a more appropriate increase in the number of local search attempts, because the current direction has good prospects and is worth exploring in depth; correspondingly, a smaller P allows for a more appropriate decrease in the number of searches.

[0069] When the comprehensive evaluation index Given the first preset index threshold Second preset index threshold Between (i.e.) The study determined that while the current alternative stimulus was effective, the therapeutic effect could be further improved. A Bayesian optimization algorithm was then used to perform a global search within at least one first-class candidate stimulus, i.e., within the current class, to obtain the next alternative stimulus for treatment.

[0070] Furthermore, if at least one first-class candidate program has been tried and no alternative stimulation program with a comprehensive evaluation index greater than or equal to the preset target comprehensive evaluation index is found, then the alternative stimulation program with the largest comprehensive evaluation index among all tried alternative stimulation programs is used as the anchor point. At least one second-class candidate program with different electrode positions but the same electrical stimulation parameters is matched to the current tried program, and one alternative stimulation program is selected from these for treatment, that is, a cross-class search (or a cross-class global search) is performed. In this paper, "global search" refers to searching for a program among all first-class candidate programs or all second-class candidate programs corresponding to the anchor point. Correspondingly, "local search" refers to searching for a program among all first-class candidate programs or all second-class candidate programs with corresponding constraints (e.g., slope growth direction, preset search step size; or, for example, preset number of attempts).

[0071] When the comprehensive evaluation index Less than the second preset index threshold If the overall efficacy of the current alternative stimulation protocols is deemed severely insufficient, the current or initial protocol is immediately used as an anchor point to match at least one second-class candidate protocol with different electrode positions but the same electrical stimulation parameters. A Bayesian optimization algorithm is then used to select one of these alternative stimulation protocols as the next protocol to be tried for the treatment of the subject.

[0072] Specifically, when At the same time, and before the step of searching for the next alternative stimulus among multiple second-type candidate solutions, the procedure also includes the step of: Multiple first-class candidate solutions are tried sequentially according to the percentage of the electric field space to the lesion area coverage. When it is determined that the number of solutions tried in the first-class candidate solutions has reached a preset threshold, the solution with the highest comprehensive evaluation index is used as the anchor point to match multiple second-class candidate solutions, and a Bayesian optimization algorithm is used to select the next alternative stimulus solution. Of course, if the preset threshold is not reached, the next alternative stimulus solution is tried in descending order according to the percentage of the electric field space to the lesion area coverage.

[0073] Furthermore, when or If, when searching for the next alternative stimulus option among multiple first-class candidate options, it is found that there are no untried options in the current category, then a cross-class search is performed using the alternative stimulus option with the highest comprehensive evaluation index among all tried alternative stimulus options as the anchor point.

[0074] For example, since the number of first-class candidate schemes is relatively small, when multiple (less than the preset attempt threshold) candidate stimulation schemes have the same electrode position but different electrical stimulation parameters, and there are no first-class candidate schemes to continue trying, the candidate stimulation scheme with the largest comprehensive evaluation index is used as the anchor point to find multiple second-class candidate stimulation schemes with the same electrical stimulation parameters (of course, "same" here includes the same in a strict sense, as well as the case where the deviation of any parameter or multiple parameters is within the preset deviation range), but different electrode positions. Then, a candidate stimulation scheme is selected from them as the initial scheme using a Bayesian optimization algorithm.

[0075] In this embodiment, the specific steps of selecting a candidate stimulation scheme using the Bayesian optimization algorithm include: constructing a surrogate model based on the electrical stimulation parameters corresponding to all historically tried schemes and their corresponding comprehensive evaluation indices (for example, constructing a Gaussian surrogate model based on the electrical stimulation parameters such as electrode position and current magnitude of each tried scheme, and the comprehensive evaluation index, slope, and variance calculated based on the resting-state EEG power ratio collected under the electrical stimulation parameters, so as to construct the mapping relationship between electrical stimulation parameters and comprehensive evaluation index / slope / variance), then using the surrogate model to predict all untried schemes searched in the search space, obtaining their corresponding comprehensive evaluation index and confidence interval (i.e., variance), and then automatically calculating the score of each untried scheme among multiple first-class candidate schemes through the acquisition function of the Bayesian optimization algorithm, and taking the scheme with the highest score as the next candidate stimulation scheme to be tried.

[0076] Although all three scenarios utilize Bayesian optimization algorithms to determine the next alternative stimulus, their prediction ranges differ. Specifically: As mentioned earlier, when it is determined that P is greater than the first preset exponential threshold, since the search range is constrained to the local neighborhood of the current electrode position, only the fine-tuning of the electrical stimulation parameters is considered, and the search is conducted along the slope growth direction. Therefore, when using the Bayesian optimization algorithm to find the next attempt scheme, it only predicts some schemes searched in the local range near the current scheme (that is, the schemes near the current attempt scheme in the same class of alternative stimulation schemes).

[0077] As mentioned earlier, when it is determined that P is between the first preset exponential threshold and the second preset exponential threshold, the next attempt is found directly in the search space, that is, in all the first category of alternative stimulus schemes corresponding to the anchor point, by using the Bayesian optimization algorithm.

[0078] As mentioned earlier, when it is determined that P is less than the second preset exponential threshold, the next solution to be tried is found by using the Bayesian optimization algorithm among all the second-class candidate solutions for the anchor point.

[0079] In some embodiments, the first preset exponential threshold The range of values ​​is to The second preset index threshold The range of values ​​is to And satisfy The specific values ​​are preset based on the individual differences and clinical response characteristics of the target subjects.

[0080] In other embodiments, when it is determined that P is less than a second preset exponential threshold, the step of matching multiple second-type candidate solutions with the current solution as the anchor point specifically includes the following steps: Using the current scheme as the anchor point, multiple first-class candidate schemes are matched. Among these multiple first-class candidate schemes, a preset number of first-class candidate schemes are tried in descending order of coverage percentage. If no candidate stimulus scheme with a slope greater than the second preset index threshold is found, the candidate stimulus scheme with the largest comprehensive evaluation index is used as the anchor point, and multiple second-class candidate schemes are matched for it. Then, based on the same principle, the Bayesian optimization algorithm is used to find the next scheme to be tried from the multiple second-class candidate schemes.

[0081] For example, when multiple first-class candidate solutions are matched for the current attempt, the multiple first-class candidate solutions are sorted by coverage percentage, and multiple solutions are tried in descending order starting from the candidate test solution with the largest coverage percentage. If the number of attempts reaches a preset attempt threshold and no solution with a slope greater than or equal to a second preset threshold is found, the search strategy is changed: cross-class search is performed, that is, the candidate stimulus solution with the largest comprehensive evaluation index among the current attempt and the multiple first-class candidate solutions is used as the anchor point, and multiple second-class candidate solutions are matched for it. Then, a Bayesian optimization algorithm is used to construct a probabilistic surrogate model based on all historical attempts and their corresponding slopes, variances, and comprehensive evaluation indices, to predict the comprehensive flat index and its confidence interval (such as variance) of all unattended second-class candidate solutions, and the score of each unattended second-class candidate solution is automatically calculated through the collection function of the Bayesian optimization algorithm. The solution with the highest score is used as the next candidate stimulus solution to be tried.

[0082] Furthermore, if the Bayesian optimization algorithm still fails to find a solution with a slope greater than or equal to the first preset exponential threshold, it can prompt that the parameters need to be manually adjusted.

[0083] When evaluating treatment efficacy, relying solely on resting-state EEG data carries a high risk of misjudgment. To mitigate this risk, measures are taken. For example, a single increase in the R value might be accidental (e.g., due to adequate rest), or the R value might show an upward trend in the first few treatments but remain unchanged or show very little change in subsequent treatments, indicating a treatment plateau. Generally, to reduce misjudgment, task-based EEG data or multimodal EEG data such as CT or MRA, or even scale-based assessments, are typically introduced for comprehensive evaluation. However, all of these methods require a considerable amount of time to acquire the data, impacting equipment response efficiency; furthermore, the integrated processing of different modalities increases algorithmic complexity. Therefore, in order to improve the efficiency of the effect while ensuring the accuracy of the assessment (i.e., reducing the probability of misjudgment), this application combines the slope, acceleration, and variance of the power ratio of resting-state EEG data after multiple treatments with linear fitting. These are evaluated from three dimensions: slope corresponds to the trend dimension, reflecting the average rate of improvement of the brain's excitation / inhibition ratio in continuous treatment; acceleration corresponds to the curvature dimension, reflecting the rate of change of the slope itself, with positive acceleration indicating accelerated release of therapeutic effect (adaptation period) and negative acceleration indicating that neuronal response is close to saturation (bottleneck / plateau period); variance corresponds to the stability dimension, reflecting the degree of fluctuation of the R value, with low variance indicating stable and reliable therapeutic effect and high variance indicating that the data is dominated by noise such as emotion, sleep, and EEG artifacts. Based on the evaluation results, a decision can be made quickly to adjust the treatment plan, for example, by adjusting only the electrode position of the current treatment plan (i.e., spatial dimension adjustment) or by adjusting only the electrical stimulation parameters of the current treatment plan (i.e., cross-category adjustment).

[0084] In other words, without introducing any other modal data (task-based EEG, CT / MRA, scale test results), this application maps a simple one-dimensional numerical value into a three-dimensional state space (trend-slope + momentum-acceleration + confidence-variance) by mining the dynamic evolution characteristics of a single resting-state index R value in the time dimension, thereby achieving both high timeliness and high accuracy.

[0085] Step 105: Iterate and optimize until the termination condition is met, and output the final solution.

[0086] Slope under the current alternative stimulus package and comprehensive evaluation index Afterward, maintain the current alternative stimulus or select the next alternative from the alternative stimulus pool, and repeat the stimulus, data collection, calculation and evaluation process until the preset termination conditions are met.

[0087] In some embodiments, the iterative process is judged according to preset termination conditions. When any termination condition is met, the iteration stops and the current or historically best alternative stimulus regimen is taken as the final treatment plan for the target object. Preset termination conditions include, but are not limited to: First termination condition: There are no remaining regimens available to try in the alternative stimulus regimen library, i.e., all All alternative stimulus options have been tried, and the search space has been fully explored. Second termination condition: The overall evaluation index of the current alternative stimulus options reaches the preset target index threshold. Third termination condition: The slope is greater than the first preset threshold.

[0088] In some embodiments, when the second or third termination condition is met, the current alternative stimulation protocol is determined as the final treatment plan for the target patient, and the electrode position information and electrical stimulation parameter information corresponding to the protocol are output. When the first termination condition is met, i.e., there are no remaining protocols available for trying in the alternative stimulation protocol library, it indicates that the system has fully traversed all pre-generated alternative stimulation protocols and still has not found a protocol that can meet the second or third termination condition. At this time, the comprehensive evaluation index in the historical record is used. The highest-ranking alternative stimulation regimen is selected as the final treatment plan for the target patient. Specifically, the comprehensive evaluation index of all tried regimens is retrieved from the storage module. Find satisfaction Scheme Index ,in The total number of solutions tried. For the first The comprehensive evaluation index of each tried treatment plan is used to output the electrode location information and electrical stimulation parameter information corresponding to that plan as the final treatment plan. Value rather than slope The rationale for backtracking selection is as follows: The value comprehensively reflects three dimensions: efficacy level, stability, and development trend. It allows for a more holistic evaluation of the overall efficacy of the regimen, avoiding the selection of a regimen with rapid but highly unstable efficacy solely based on the slope. If multiple regimens exist... If the values ​​are the same or very close, choose the slope from them. The most comprehensive approach will be the final treatment plan.

[0089] In some preferred embodiments, during the search for a solution based on the percentage of coverage area, if the slope of multiple consecutive alternative stimulus solutions (e.g., three consecutive solutions, but less than a preset attempt threshold) If all values ​​are less than or equal to 0, meaning multiple consecutive schemes are deemed invalid, the fast-track mechanism is triggered. The fast-track mechanism skips schemes based on coverage percentage. The process involves sequentially trying intermediate solutions, and then directly selecting the coverage percentage from the remaining untried solutions. The median solution between the highest and lowest values, or the solution furthest from the currently tried solution in parameter space, is attempted first to quickly escape the invalid region and avoid wasting too many treatment attempts on consecutive invalid solutions. Furthermore, if the number of attempts reaches a preset threshold without meeting the termination condition, a cross-category search is initiated. Of course, the count of consecutive invalid solutions is updated when any valid solution (i.e., ...) is found. When the comprehensive evaluation index is greater than or equal to the preset target comprehensive evaluation index, it will be reset to zero.

[0090] In some preferred embodiments, during the iterative optimization process, if the variance of the current alternative stimulus is... Exceeding the maximum permissible variance And slope Less than the first preset threshold This indicates that the efficacy indicators of the current regimen fluctuate greatly and the efficacy level has not reached the target, making the treatment effect unreliable. In this case, we should not wait for the complete evaluation cycle of the current regimen (i.e., not complete all evaluations). (Second treatment), early termination of the current protocol evaluation and immediate switching to the next alternative stimulation protocol to save treatment time and costs. Maximum permissible variance The settings take into account the individualized baseline noise level of the target subjects: before the first treatment, resting-state EEG signals were collected multiple times consecutively, and the standard deviation of the resting-state EEG power ratio was calculated. The maximum permissible variance is set based on the baseline noise level. ,in This is a preset proportionality coefficient, with a value ranging from 2 to 5. If the current variance... If the current data is determined to be dominated by noise, the weighting coefficient in the comprehensive evaluation index will be reduced. The decision weights, and the acceleration of the rise. Set to zero.

[0091] In some preferred embodiments, the total number of treatments in the iterative optimization process does not exceed a preset safety limit, which is pre-set based on the target subject's tolerance and ethical requirements, and generally does not exceed 30 treatments. If the second or third termination condition is not met even after reaching the safety limit, the iteration is forcibly terminated, and the comprehensive evaluation index from the historical records is used. The highest-ranking alternative stimulation regimen is selected as the final treatment plan, while a warning notification is sent to the clinician, indicating that the target subject does not respond well to the regimens in the existing alternative stimulation regimen library, and suggesting that the lesion be re-labeled or a new alternative stimulation regimen library be generated.

[0092] By repeating the above calculation and selection steps until the first, second, or third termination condition is met, or the safety upper limit is reached, the corresponding alternative stimulation scheme is output as the final treatment plan, completing the entire parameter optimization process of the neuromodulation method for treating cognitive dysfunction. The final treatment plan contains complete electrode location information and electrical stimulation parameter information, which can be used to guide subsequent long-term clinical electrical stimulation therapy.

[0093] See Figure 3 and Figure 4 To utilize the scheme ultimately determined by the method described in this embodiment, the method includes: applying an electrical stimulation signal with a frequency of 2040 Hz and a current intensity of 1.88 mA to the electrical stimulation channel between the first electrode (i.e., the positive electrode) with electrode number CH13 and the second electrode (i.e., the negative electrode) with electrode number CH32 in the first electrode pair, wherein the electrode position of the first electrode CH13 is at AF8 and the electrode position of the second electrode CH32 is at TP7; applying an electrical stimulation signal with a frequency of 2000 Hz (i.e., the frequency difference between the two electrical stimulation channels is 40 Hz) and a current intensity of 1.88 mA to the electrical stimulation channel between the third electrode (i.e., the positive electrode) with electrode number CH14 and the fourth electrode (i.e., the negative electrode) with electrode number CH33 in the second electrode pair.

[0094] For subjects trying electrical stimulation therapy for the first time, or those whose physiological state has changed significantly recently, the above approach can quickly identify effective electrical stimulation protocols within a short period. While this protocol may not be the most optimal, it is one that can have a positive effect. Therefore, to further identify even more effective protocols quickly, in some embodiments, the actual therapeutic effect of the final treatment protocol can be monitored, and then further optimized based on the actual results to find even better protocols, i.e., to further optimize the neuromodulation parameters. For example, if it is necessary to continue treatment using the final protocol for at least one course (e.g., one month per course, three groups per week, each group including 3-6 electrical stimulations in the morning and afternoon), and calculate its comprehensive evaluation index using the same principles as in steps S4-S5 (the difference being that here each group is used as the evaluation unit, i.e., the comprehensive evaluation index is calculated based on the EEG data before and after each group's treatment). If its comprehensive evaluation index reaches the preset target comprehensive evaluation index, it is adopted as the final protocol. If it does not reach the preset target comprehensive evaluation index, the protocol is used as an anchor point, and the search strategy for the next alternative test protocol is selected based on the region where the comprehensive evaluation index is located (e.g., a local or global search is performed in the first category of candidate protocols corresponding to the anchor point, or a cross-category search is performed, i.e., a global search is performed in the second category of candidate protocols corresponding to the anchor point. The specific principles of local search, global search, and cross-category search are described above).

[0095] Example 2: Based on the method of Example 1 above, this application also provides a neuromodulation system for treating cognitive dysfunction, comprising: A storage module is configured to provide multiple alternative stimulation schemes, which are sorted by the percentage of electric field space to lesion area coverage. An electrical stimulation module is configured to select an initial protocol from a plurality of alternative stimulation protocols and perform multiple treatments on the target object; specifically, the electrical stimulation module is externally connected to an existing electrical stimulation device and sends the electrical stimulation parameters of the initial protocol to the electrical stimulation device to trigger the electrical stimulation device to perform treatment; The resting-state EEG power ratio calculation module is configured to collect resting-state EEG signals of the target subject after each treatment and calculate the resting-state EEG power ratio after each treatment based on the resting-state EEG signals. Specifically, the resting-state EEG power ratio calculation module communicates with an external electrical stimulation device and obtains the resting-state EEG signals from it. The treatment plan selection module is configured to perform linear fitting on the resting-state EEG power ratio after multiple treatments to obtain a slope. If the slope is less than or equal to a preset amplitude, such as 0, the current category of alternative stimulation plans is determined to be invalid, and the first alternative stimulation plan after the initial plan is switched to as the new initial plan for multiple treatments on the target subject. If the slope is greater than or equal to a first preset threshold, the current alternative stimulation plan is maintained. If the slope is greater than 0 and less than the first preset threshold, the acceleration of the resting-state EEG power ratio over time and the variance after multiple consecutive treatments under the same alternative stimulation plan are calculated. A comprehensive evaluation index is constructed based on the slope, acceleration, and variance. The next alternative stimulation plan is selected based on the comprehensive evaluation index. The treatment plan determination module is configured to repeatedly execute the functional steps of the treatment plan selection module until the preset termination conditions are met and the final treatment plan is obtained.

[0096] Further, the scheme selection module is specifically configured as follows: when it is determined that the comprehensive evaluation index is greater than the first preset index threshold, using the current attempt scheme as the plotting point, at least one first-class candidate scheme with the same electrode position but different electrical stimulation parameters is matched for the current attempt scheme, and scheme search is performed in the at least one first-class candidate scheme along the slope growth direction with a preset search step size, and the next alternative stimulation scheme is selected from them using a Bayesian optimization algorithm; when it is determined that the comprehensive evaluation index is between the first preset index threshold and the second preset index threshold, and the second preset index threshold is less than the first preset index threshold, the next alternative stimulation scheme is searched from the at least one first-class candidate scheme using a Bayesian optimization algorithm; when it is determined that the comprehensive evaluation index is less than the second preset index threshold, using the current attempt scheme as the plotting point, at least one second-class candidate scheme with different electrode position but the same electrical stimulation parameters is matched for the current attempt scheme, and the next alternative stimulation scheme is selected from them using a Bayesian optimization algorithm.

[0097] Example 3: This application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the following... Figure 1 The method shown.

[0098] Example 4: To verify the effectiveness of the scheme determined by the method of the present invention, the above-described method of the present invention was used to find individual electrical stimulation schemes for 17 patients diagnosed with dementia and cognitive impairment. Each patient underwent treatment under their respective electrical stimulation scheme for one month, with three treatment sessions per week. Each treatment session included five electrical stimulation sessions in the morning and five in the afternoon. Furthermore, before treatment (i.e., before the first treatment), after treatment (i.e., after the last treatment in the one-month treatment period), and for one month after treatment (during which no electrical stimulation treatment was performed), the subjects underwent corresponding tests, including: Clinical Dementia Rating Scale tests, Alzheimer's Disease Assessment Scale - Cognitive Section tests, and Monterey Cognitive Assessment Scale tests. The results were statistically analyzed to obtain the corresponding data, such as... Figures 5-7 Therefore, based on statistical analyses conducted before treatment, after treatment, and at one month post-treatment follow-up, the scores of the 17 subjects on the Clinical Dementia Rating Scale, the ADASCog (Alzheimer's Disease Assessment Scale - Cognitive Subscale) test, and the Monterey Cognitive Assessment Scale test all showed significant improvement. like Figure 5 As shown, before treatment, the mean ADASCog test score of the 17 subjects was close to 20, while after treatment, the mean ADASCog test score of the 17 subjects dropped significantly to below 20; and one month after the completion of treatment, the mean ADASCog test score of the 17 subjects at the follow-up was stable at around the mean ADASCog test score after treatment.

[0099] like Figure 6 As shown, before treatment, the mean score of the Clinical Dementia Rating Scale obtained by the 17 subjects was over 4, while after treatment the mean score of the Clinical Dementia Rating Scale obtained by the subjects dropped significantly to below 4; and one month after the completion of treatment, the mean score of the Clinical Dementia Rating Scale obtained by the 17 subjects during follow-up was stable at around 4, the mean score after treatment.

[0100] like Figure 7 As shown, before treatment, the mean score of the Monterey Cognitive Assessment Scale (MCA) obtained by the 17 subjects was close to 10, while after treatment, the mean score of the Monterey Cognitive Assessment Scale obtained by the 17 subjects increased significantly and approached 15; and one month after the completion of treatment, the mean score of the Monterey Cognitive Assessment Scale obtained by the 17 subjects during follow-up was stable at around 15, the mean score after treatment.

[0101] Among them, the ADASCog is one of the most commonly used scales for assessing cognitive function in Alzheimer's disease in clinical practice. It is mainly used to assess the cognitive function of patients, and the lower the score, the better the cognition. The Clinical Dementia Rating Scale is a scale used in clinical practice to assess the severity of dementia. The lower the score, the clearer the degree. The Montreal Cognitive Assessment Scale is a screening scale used in clinical practice to assess cognitive function. All of these scales indicate that the higher the score, the better the cognition.

[0102] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0103] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0104] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims. All of these forms are within the protection scope of this application.

Claims

1. A method for optimizing brain neural modulation parameters for treating cognitive impairment, characterized in that, Includes the following steps: S1: Provide multiple alternative stimulation schemes; the multiple alternative stimulation schemes are ranked according to the percentage of electric field space and lesion area coverage; S2: Select one alternative stimulation scheme from the plurality of alternative stimulation schemes as the initial scheme, and perform multiple treatments on the target subject; S3: Collect the resting-state EEG signal of the target object after each treatment, and calculate the resting-state EEG power ratio after each treatment based on the resting-state EEG signal; S4: Linear fitting was performed on the ratio of resting-state EEG power after multiple treatments to obtain the slope; If the slope is less than or equal to 0, then switch to the first alternative stimulation scheme after the initial scheme as the new initial scheme to treat the target object multiple times, and execute step S3; If the slope is greater than or equal to the first preset threshold, the current alternative stimulation scheme is maintained; if the slope is greater than 0 and less than the first preset threshold, the acceleration of the rise of the resting-state EEG power ratio over time and the variance after multiple consecutive treatments under the same alternative stimulation scheme are calculated. S5: Construct a comprehensive evaluation index based on the slope, acceleration, and variance; and select the next alternative stimulus based on the comprehensive evaluation index; S6: Repeat steps S3-S5 for the same alternative stimulation protocol until the preset termination condition is met and the final treatment plan is obtained.

2. The method for optimizing brain neural modulation parameters for treating cognitive dysfunction according to claim 1, characterized in that: Step S5, which involves selecting the next alternative stimulus package based on the comprehensive assessment index, specifically includes: Determine whether the comprehensive evaluation index is greater than or equal to a first preset index threshold. If the comprehensive evaluation index is greater than or equal to the first preset index threshold, the current attempt scheme is used as the plotting point to match at least one first-class candidate scheme with the same electrode position but different electrical stimulation parameters. The scheme is searched in the at least one first-class candidate scheme along the slope growth direction with a preset search step size to obtain a subset of candidate schemes. Then, the next alternative stimulation scheme is selected from the subset of candidate schemes using the Bayesian optimization algorithm. If the comprehensive evaluation index is between the first preset index threshold and the second preset index threshold, and the second preset index threshold is less than the first preset index threshold, then the Bayesian optimization algorithm is used to search for the next alternative stimulus scheme in the at least one first-class candidate scheme. When the comprehensive evaluation index is less than the second preset index threshold, the current attempt scheme is used as the plotting point to match at least one second-class candidate scheme with different electrode positions but the same electrical stimulation parameters, and the next alternative stimulation scheme is selected from them using a Bayesian optimization algorithm.

3. The method for optimizing brain neural regulation parameters for treating cognitive dysfunction according to claim 2, characterized in that: The initial scheme is the alternative stimulation scheme with the largest percentage of electric field space coverage area to lesion area among the multiple alternative stimulation schemes.

4. The method for optimizing brain neural modulation parameters for treating cognitive dysfunction according to claim 1, characterized in that: In step S5, the comprehensive evaluation index is calculated. The calculation formula is: ； Where a is the slope under the current attempt, σ 2 Let ξ be the variance of the current trial scheme, and ξ be the acceleration of the current trial scheme. The target slope is preset. For the maximum permissible variance, For the initial acceleration, , , The preset weighting coefficients satisfy the following conditions: .

5. The method for optimizing brain neural regulation parameters for treating cognitive dysfunction according to claim 2, characterized in that: In step S5, when the comprehensive evaluation index is between a first preset index threshold and a second preset index threshold, the step of using a Bayesian optimization algorithm to search for the next alternative stimulus scheme among the at least one first-class candidate schemes specifically includes: Based on historically tried solutions and their corresponding comprehensive evaluation indices, a probabilistic surrogate model is constructed to predict the comprehensive evaluation index and its confidence interval for all untried solutions in at least one first-class candidate solution. The score of each untried solution is automatically calculated using the acquisition function of the Bayesian optimization algorithm, and the solution with the highest score is selected as the next alternative stimulus solution.

6. The method for optimizing brain neural modulation parameters for treating cognitive dysfunction according to claim 2, characterized in that: In step S5, when the comprehensive evaluation index is less than the second preset index threshold, the step of selecting the next alternative stimulus from the at least one second-type candidate scheme specifically includes: Start by trying the candidate test scheme with the largest percentage of coverage area from multiple second-category candidate schemes; Determine whether the number of attempts in at least one second-class candidate solution has reached a preset attempt threshold; If the preset attempt threshold is not reached, the next alternative stimulation scheme is selected from the at least one second-class candidate scheme according to the percentage of the coverage area of ​​the electric field space and the lesion area. If the preset attempt threshold has been reached, a Bayesian optimization algorithm is used to construct a probabilistic proxy model based on the attempted schemes and their corresponding comprehensive evaluation indices in at least one second-class candidate scheme. The comprehensive evaluation index and its confidence interval of all untried schemes in at least one second-class candidate scheme are predicted. The score of each untried scheme is automatically calculated through the acquisition function of the Bayesian optimization algorithm, and the scheme with the highest score is selected as the next alternative stimulus scheme.

7. The method for optimizing brain neural modulation parameters for treating cognitive dysfunction according to claim 1, characterized in that: The preset termination conditions include: the comprehensive evaluation index is greater than or equal to the preset target comprehensive evaluation index; or, there are no remaining options available to try among the multiple alternative stimulus options.

8. A method for optimizing brain neural modulation parameters for treating cognitive dysfunction according to any one of claims 1-7, characterized in that: In step S3, resting-state EEG signals are collected from the target subject after each treatment, and the resting-state EEG power ratio after each treatment is calculated based on the resting-state EEG signals, including: Resting-state EEG signals were collected from the target subject after each treatment. The EEG power of the α-band, β-band, δ-band, and θ-band was extracted from the resting-state EEG signals to calculate the resting-state EEG power ratio after each treatment. The resting-state EEG power ratio was calculated as the sum of the α-band power and the β-band power divided by the sum of the δ-band power and the θ-band power.

9. A neuromodulation system for treating cognitive impairment, characterized in that, include: A storage module is configured to provide multiple alternative stimulation schemes, which are sorted by the percentage of electric field space to lesion area coverage. An electrical stimulation module is configured to select an initial protocol from the plurality of alternative stimulation protocols to perform multiple treatments on the target subject. The resting-state EEG power ratio calculation module is configured to collect the resting-state EEG signals of the target subject after each treatment, and calculate the resting-state EEG power ratio after each treatment based on the resting-state EEG signals; The treatment plan selection module is configured to perform linear fitting on the resting-state EEG power ratio after multiple treatments to obtain a slope; if the slope is less than or equal to 0, the current category of alternative stimulation plans is determined to be invalid, and the first alternative stimulation plan after the initial plan is switched to be used as the new initial plan for multiple treatments on the target subject; if the slope is greater than or equal to a first preset threshold, the current alternative stimulation plan is maintained; if the slope is greater than 0 and less than the first preset threshold, the acceleration of the resting-state EEG power ratio over time and the variance after multiple consecutive treatments under the same alternative stimulation plan are calculated; and a comprehensive evaluation index is constructed based on the slope, acceleration, and variance. And select the next alternative stimulus package based on the comprehensive assessment index; The treatment plan determination module is configured to trigger the electrical stimulation module, the resting-state EEG power ratio calculation module, and the treatment plan selection module to perform corresponding functional operation steps each time the treatment plan selection module selects the next alternative stimulation plan, until the preset termination condition is met and the final treatment plan is obtained.

10. A neuromodulation system for treating cognitive impairment according to claim 9, characterized in that, The scheme selection module is specifically configured as follows: When it is determined that the comprehensive evaluation index is greater than the first preset index threshold, the current attempt scheme is used as the plotting point to match at least one first-class candidate scheme with the same electrode position but different electrical stimulation parameters. The scheme is searched in the at least one first-class candidate scheme along the slope growth direction with a preset search step size, and the next alternative stimulation scheme is selected from them using the Bayesian optimization algorithm. When it is determined that the comprehensive evaluation index is between the first preset index threshold and the second preset index threshold, and the second preset index threshold is less than the first preset index threshold, the next alternative stimulus scheme is obtained by searching among the at least one first-class candidate scheme using the Bayesian optimization algorithm. When it is determined that the comprehensive evaluation index is less than the second preset index threshold, the current attempt scheme is used as the plotting point to match at least one second-class candidate scheme with different electrode positions but the same electrical stimulation parameters, and the next alternative stimulation scheme is selected from them using the Bayesian optimization algorithm.

Images