Individualized time-interleaved stimulation parameter optimization method based on target probability distribution

Abstract

The application discloses an individualized time interference stimulation parameter optimization method based on a target point probability distribution and relates to the technical field of transcranial electrical stimulation. The application constructs a target point space probability density distribution atlas that matches the dynamic network characteristics of a patient, and uses the same to weight the maximum envelope amplitude in a candidate target point region, so as to maximize the weighted sum of the maximum envelope amplitudes in the candidate target point region and minimize the average value of the envelope amplitudes in the non-target cortical region as the target. Based on a three-dimensional finite element head model, electric field simulation and parameter optimization are performed on the time interference stimulation parameters to be selected, and the optimal scheme of the time interference stimulation parameters of the target individual is obtained. The application solves the spatial drift problem caused by the individualized target point positioning of transcranial electrical stimulation relying on a single image, and greatly improves the focusing of deep electric field distribution and individual adaptation.

Description

Technical Field

[0001] This invention relates to the field of transcranial electrical stimulation technology, specifically to a method for optimizing individualized temporal interference stimulation parameters based on target probability distribution. Background Technology

[0002] Combining multimodal neuroimaging for personalized stimulation target localization is an important direction for improving the therapeutic effects of non-invasive neuromodulation techniques. Personalized target selection requires not only consideration of anatomical structures but also integration of multimodal neuroimaging data such as functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) to comprehensively assess the functional structural connectivity of the individual brain, identify abnormal brain regions related to the disease, and thus achieve more precise intervention. In recent years, approaches combining brain imaging technology with personalized targeting and novel stimulation modalities (such as SAINT therapy) have achieved significant breakthroughs in clinical efficacy.

[0003] However, the stability of target localization methods based on neuroimaging-based individual functional connectivity analysis remains a significant challenge when expanding the application of such personalized image-guided therapies. Existing personalized paradigms heavily rely on functional connectivity data acquired from a single resting-state fMRI scan. Increasing research indicates that individual functional connectivity patterns are not entirely static trait indicators but exhibit a degree of intra- and inter-diurnal variability. This temporal variability can directly lead to spatial shifts in the coordinates of personalized transcranial stimulation targets determined from a single scan at different time points. For example, several studies on the core circuitry of depressive disorders have found that the calculated optimal individual target can shift spatially by 25-37 mm between different scanning sessions. Current methods for individualized target localization often treat the functional connectivity patterns obtained from a single scan as stable characteristics of the individual, failing to fully incorporate its dynamic temporal dimension. Therefore, functional connectivity analysis relying solely on cross-sectional scans is insufficient to fully characterize the robustness of individual functional networks, necessitating the development of target localization methods that integrate temporal information and are more robust to intra-individual variability.

[0004] On the other hand, transcranial temporal interferometry (TI) electrical stimulation, as an emerging non-invasive deep brain stimulation technique, induces an electric field distribution that is primarily influenced by individual tissue structure and dielectric properties, as well as temporal interferometry stimulation parameters (such as electrode location and excitation current). Due to the shunting effect of different layers of cranial tissue on current, traditional stimulation methods result in low and diffuse current density reaching the target brain region, limiting their focusing and targeting capabilities. Therefore, current technologies typically utilize individual transcranial electrical stimulation electric field simulation models (achieved by acquiring individual magnetic resonance imaging information, performing tissue segmentation, and constructing a three-dimensional finite element model) and parameter optimization methods to improve the intensity and focusing of the electric field generated by temporal interferometry stimulation in the target brain region.

[0005] However, traditional methods for optimizing temporal interference stimulation parameters typically only aim to maximize the electric field intensity of a single static target area. This optimization approach severs the connection between the dynamic changes in brain network connectivity and the physical electric field distribution, failing to adapt to the objective laws governing target location drift within individual functional networks. How to construct a target model that fits the dynamic network characteristics of patients, and use this model to drive the precise optimization of individualized temporal interference stimulation parameters, thereby significantly improving the focus and individual fit of deep electric field distribution, is a technical challenge that urgently needs to be addressed by those skilled in the art. Summary of the Invention

[0006] To address the spatial drift problem caused by the reliance on single images for individualized target localization in existing transcranial electrical stimulation (TCS) techniques, this invention aims to provide a method for optimizing individualized temporal interference stimulation parameters based on target probability distribution. This invention constructs a target probability density distribution model that fits the patient's dynamic network characteristics, and uses this model as spatial weights to drive a targeted optimization algorithm for individualized temporal interference stimulation parameters, significantly improving the focusing ability and individual fit of deep electric field distribution.

[0007] To achieve the above objectives, the specific technical solution adopted by the present invention is as follows:

[0008] In a first aspect, the present invention provides a method for optimizing individualized temporal interference stimulation parameters based on target probability distribution, comprising:

[0009] S1. Acquire and register three-dimensional structural images and resting-state functional magnetic resonance imaging data of the target individual in multiple trials, calculate the time-series functional connectivity strength between the cortical effect area and the deep effect area in each trial, and screen out the cortical effect area voxel with the strongest negative functional connectivity as the candidate target point for each individual in each trial.

[0010] S2. Within the candidate target area pre-defined by combining the prior information of the group, count the frequency of each voxel being identified as an individual candidate target in all trials, and then generate the target spatial probability density distribution map of the target individual through spatial interpolation.

[0011] S3. Perform tissue segmentation on the three-dimensional structural image of the target individual to construct a three-dimensional finite element head model;

[0012] S4. Using the target spatial probability density distribution map, the maximum envelope amplitude in the candidate target area is weighted. The goal is to maximize the weighted sum of the maximum envelope amplitude in the candidate target area while minimizing the average value of the envelope amplitude in the non-target cortical area. Based on the three-dimensional finite element head model, electric field simulation and parameter optimization are performed on the selected time interference stimulation parameters to obtain the optimal scheme of time interference stimulation parameters for the target individual.

[0013] As a preferred embodiment of the first aspect above, in S1, the method for extracting individual candidate targets in each trial is as follows: for the cortical effect area and deep effect area corresponding to the target intervention disease, the average time-domain signal sequence in the deep effect area is extracted from the resting-state functional magnetic resonance data of the current trial; then, the correlation coefficient between the time-domain signal sequence of each voxel in the cortical effect area and the average time-domain signal sequence is calculated through correlation analysis; then, the correlation coefficient corresponding to each voxel is converted by a standardization transformation method and used as the time-series functional connectivity strength corresponding to that voxel; the voxel with the smallest negative time-series functional connectivity strength in the cortical effect area that satisfies spatial constraints is selected as the individual candidate target in the current trial.

[0014] As a preferred embodiment of the first aspect above, the target intervention disease is depression, the corresponding cortical effect area is the dorsolateral prefrontal cortex, the corresponding deep effect area is the anterior cingulate cortex, and the spatial constraint is set such that the candidate target point is located in the gray matter of the brain and its spatial position is within 4 cm of the scalp surface.

[0015] As a preferred embodiment of the first aspect above, in S2, the method for delineating candidate target regions by combining prior information of the population is as follows: acquiring population image data including the patient group of the target intervention disease and the healthy control group, calculating the average functional connectivity network of the two groups based on the resting state functional image data, and delineating abnormal brain regions as candidate target regions in the standard three-dimensional brain space according to the network difference comparison results.

[0016] As a preferred embodiment of the first aspect above, in S2, the method for generating the target spatial probability density distribution map of the target individual is as follows: mapping all individual candidate target points identified in all trials to the candidate target area, counting the frequency of each voxel in the candidate target area being identified as an individual candidate target, and then using the frequency to represent the probability of being selected, and using density estimation or smoothing algorithms to generate a target spatial probability density distribution map covering the entire candidate target area.

[0017] As a preferred embodiment of the first aspect above, in S3, based on the T1-weighted and T2-weighted structural magnetic resonance imaging data of the target individual, the cranial tissue of the target individual is segmented into at least the scalp, skull, cerebrospinal fluid, gray matter, and white matter using a medical image segmentation algorithm. After smoothing and surface reconstruction of the segmented tissue mask, a three-dimensional finite element head model is generated through three-dimensional meshing. Finally, different tissue components are assigned corresponding conductivity and dielectric properties.

[0018] As a preferred embodiment of the first aspect above, in S4, each set of candidate time-interference stimulation parameters includes the combination position of the stimulation electrode array and the current distribution ratio between the electrodes.

[0019] As a preferred embodiment of the first aspect above, in S4, the method for obtaining the optimal solution of time-interference stimulation parameters through parameter optimization is as follows: For each set of candidate time-interference stimulation parameters, a stimulation electrode array that meets the position requirements of the current stimulation parameters is arranged on the basis of the three-dimensional finite element head model, and a sinusoidal current is injected according to the corresponding current distribution ratio and the preset difference frequency. The superimposed electric field distribution in the brain is obtained through electric field simulation, and then the maximum envelope amplitude of each voxel is calculated. Using the probability value of each voxel in the target spatial probability density distribution map as a weighting weight, the maximum envelope amplitude of all voxels in the candidate target area is weighted and summed. Then, the weighted sum is divided by the average value of the maximum envelope amplitude of all voxels in the non-target cortical area to obtain the weighted peak sum ratio. The time-interference stimulation parameter with the largest weighted peak sum ratio is selected as the optimal solution of time-interference stimulation parameters for the target individual.

[0020] In a second aspect, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, can implement the individualized time-interference stimulus parameter optimization method based on target probability distribution as described in any of the solutions of the first aspect above.

[0021] Thirdly, the present invention provides a computer electronic device, which includes a memory and a processor;

[0022] The memory is used to store computer programs;

[0023] The processor is configured to, when executing the computer program, implement the individualized time-interference stimulus parameter optimization method based on target probability distribution as described in any of the first aspects above.

[0024] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in:

[0025] (1) This invention abandons the traditional approach of relying on a single fMRI scan to determine the target point, and fully considers the fluctuation pattern of individual brain functional connectivity networks in the time dimension. By integrating spatial offset and recurrence rate data from multiple scans, this method has stronger robustness to intra-individual variations and solves the problem of large spatial displacement errors in traditional localization methods;

[0026] (2) The present invention adopts a dimensionality reduction strategy of prior delineation of regions by large queues and weighting of individual dynamic features, which not only ensures the correct pathological direction of target selection, but also realizes micro-calibration for individual brain specificity, and constructs a high-precision target probability density distribution model.

[0027] (3) This invention innovatively incorporates probability density distribution as a spatial weighting factor directly into the objective function of temporal interference electric field optimization, breaking through the limitation of previous methods that only pursued the maximum intensity of a single target point. This method enables the final stimulation electric field to not only be highly focused on the deep target area, but also to perfectly fit the abnormal weights of the patient's actual pathological network in terms of the distribution pattern of the electric field intensity, significantly reducing the bystander effect in non-target brain regions and greatly enhancing the therapeutic potential of individualized and precise intervention. Attached Figure Description

[0028] 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. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is a schematic diagram illustrating the steps of an individualized time-interference stimulus parameter optimization method based on target probability distribution.

[0030] Figure 2 This is a schematic diagram of the structure of a computer electronic device;

[0031] Figure 3 A schematic diagram of the overall technical route and process of the individualized time interferometry parameter optimization method based on target probability density distribution provided in the embodiments of the present invention;

[0032] Figure 4 This is a schematic diagram illustrating the probability density distribution of individual target spatial locations obtained through multiple fMRI scans and functional connectivity analysis and population prior information in an embodiment of the present invention.

[0033] Figure 5 This is a diagram showing the overall structure and the construction results of each tissue in the individual head anatomy simulation model in this embodiment of the invention;

[0034] Figure 6 This is a schematic diagram of individual time-interference stimulus parameter optimization based on target probability density distribution in an embodiment of the present invention;

[0035] Figure 7 This is an axial view of the spatial distribution and probability density map P(r) of the synthetic target points in an embodiment of the present invention, showing the positions of 10 synthetic target points, the geometric center of the dual-component mixed distribution, and the probability density distribution shape.

[0036] Figure 8 The single-target optimization method in Embodiment 2 of this invention and the method described herein are compared in V ROI Comparison of electric field envelope amplitude AM(r) distribution within the field (axial view);

[0037] Figure 9 This is a quantitative comparison chart of the electric field envelope amplitude at 10 candidate target points between the single-target optimization method in Embodiment 2 of the present invention and the present method, including (A) a bar chart of the electric field envelope amplitude at each target point and (B) a comparison of the AM mean value grouped by P weight. Detailed Implementation

[0038] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. Technical features in various embodiments of the present invention can be combined accordingly without mutual conflict.

[0039] In the description of this invention, it should be understood that when an element is considered to be "connected" to another element, it can be a direct connection to the other element or an indirect connection, i.e., there is an intermediate element. Conversely, when an element is said to be "directly" connected to another element, there is no intermediate element.

[0040] This invention provides a method for optimizing individualized temporal interference stimulation parameters based on target probability distribution, primarily applied in the planning and generation systems of non-invasive neuromodulation and individualized EEG stimulation protocols. For example... Figure 1 As shown, the individualized time-interference stimulus parameter optimization method includes the following steps S1 to S4. The specific implementation of each step is described in detail below.

[0041] S1. Acquire and register three-dimensional structural images and resting-state functional magnetic resonance imaging data of the target individual through multiple trials, calculate the time-series functional connectivity strength between the cortical effect area and the deep effect area in each trial, and screen out the cortical effect area voxel with the strongest negative functional connectivity as the candidate target point for each individual in each trial.

[0042] It should be noted that the target individual in this invention refers to a patient or subject who needs to receive transcranial temporal interference (TI) electrical stimulation. For each target individual, multimodal brain data needs to be collected from different samples at different sampling time points using a magnetic resonance imaging (MRI) scanner. The multimodal data includes at least two types: three-dimensional structural images and resting-state functional magnetic resonance imaging (fMRI) data. The three-dimensional structural images can be T1-weighted or T2-weighted structural MRI data. The specific number of stimulation trials is N≥2, but more trials should be collected whenever possible, to obtain a larger number of candidate target points for the individual and improve the accuracy of the target individual's spatial probability density distribution map.

[0043] To ensure that both 3D structural images and resting-state functional magnetic resonance imaging (fMRI) data can be used for subsequent data processing, necessary preprocessing and registration operations are required. The specific methods for preprocessing and registration can refer to general practices for magnetic resonance imaging. In embodiments of this invention, for 3D structural images, retargeting and tissue segmentation can be performed. For resting-state fMRI images, initial instability point removal, time-layer alignment, and head motion artifact correction can be performed sequentially, followed by spatial registration with the 3D structural images. Then, the images are normalized to a reference brain template (e.g., the MNI standard spatial brain template) and voxel resampling is performed. If necessary, interference components such as white matter, cerebrospinal fluid signals, and head motion parameters can be removed. Spatial smoothing is then performed to improve the signal-to-noise ratio. Finally, bandpass filtering can be performed to extract neural signals in the target frequency band.

[0044] In an embodiment of the present invention, the method for extracting individual candidate targets in each trial is as follows: for the cortical effect area and deep effect area corresponding to the target intervention disease, the average time-domain signal sequence in the deep effect area is extracted from the resting-state functional magnetic resonance data of the current trial; then, the correlation coefficient between the time-domain signal sequence of each voxel in the cortical effect area and the average time-domain signal sequence is calculated through correlation analysis; then, the correlation coefficient corresponding to each voxel is converted by a standardization transformation method and used as the time-series functional connectivity strength corresponding to that voxel; the voxel with the smallest negative time-series functional connectivity strength in the cortical effect area that satisfies the spatial constraints is selected as the individual candidate target for the current trial.

[0045] It should be noted that the cortical effect area and deep effect area are different for different diseases. Therefore, in this invention, the cortical effect area and deep effect area need to be determined according to the target disease to be intervened in. In the embodiments of this invention, if the target disease to be intervened in is depression, the corresponding cortical effect area can be selected from the dorsolateral prefrontal cortex (DLPFC), and the corresponding deep effect area can be selected from the anterior cingulate cortex (ACC).

[0046] It should be noted that the correlation analysis method used in this invention recommends the Pearson correlation analysis algorithm. The average time-domain signal sequence of all voxels in the deep effect region is used to obtain an average time-domain signal sequence. For any voxel v in the cortical effect region, the time-domain signal sequence of voxel v is subjected to Pearson correlation analysis with the average time-domain signal sequence of all voxels in the deep effect region to obtain an r value. Then, the Fisher's z-transformation algorithm is recommended as the normalization transformation method. The r value is converted into a z-score through the Fisher z-transformation, and the z-score can be regarded as the time-series functional connectivity strength of voxel v.

[0047] It should also be noted that during time-interference stimulation, the stimulation target should not be too deep into the scalp surface to ensure that the non-invasive electrical stimulation energy can effectively reach the target area. Therefore, when selecting individual candidate targets for each trial, both the time-series functional connectivity strength of the voxel and the corresponding spatial constraints should be considered. In other words, the voxel with the lowest time-series functional connectivity strength within the cortical effector area that satisfies the spatial constraints should be selected as the individual candidate target for the current trial. Since the z-score representing the time-series functional connectivity strength can be positive or negative, the minimum time-series functional connectivity strength is the largest absolute value among the negative intensity values.

[0048] S2. Within the candidate target area pre-defined by combining the prior information of the group, count the frequency of each voxel being identified as an individual candidate target in all trials, and then generate the target spatial probability density distribution map of the target individual through spatial interpolation.

[0049] It should be noted that the prior information of the population in this invention refers to prior information obtained through the analysis of large cohort group imaging data, rather than information obtained from the analysis of only a single individual or a subset of individuals. This ensures that the overall pathological direction of target selection is correct. The larger the sample size in the large cohort, the better, but there is no limitation on the specific number of samples.

[0050] In an embodiment of the present invention, the method for delineating candidate target regions by combining prior information of the population is as follows: acquiring population image data including a patient group with the target intervention disease and a healthy control group; calculating the average functional connectivity network of the two groups based on resting-state functional image data; performing a comparative analysis of the differences in the average functional connectivity networks of different brain regions of the two groups; and considering the differences exceeding a preset allowable upper limit as abnormal brain regions. Finally, based on the network difference comparison results, abnormal brain regions are delineated in a standard three-dimensional brain space. These abnormal brain regions can be used as candidate target regions to delineate the regional boundaries when generating a target spatial probability density distribution map.

[0051] It is important to note that this candidate target region does not need to be generated individually for different target individuals. It is obtained based on statistical analysis of a large amount of population data and can be directly reused in the target spatial probability density distribution map generation task for different target individuals after it is determined.

[0052] Once the candidate target region is determined, the method for generating the target spatial probability density distribution map of the target individual can be implemented as follows: map the individual candidate target points identified in all trials to the candidate target region, count the frequency of each voxel in the candidate target region being identified as an individual candidate target, and then use the frequency to represent the probability of being selected, and use density estimation or smoothing algorithms to generate a target spatial probability density distribution map covering the entire candidate target region.

[0053] It should be noted that since each individual candidate target is actually a voxel, after counting the frequency of each voxel within the candidate target region being identified as an individual candidate target, a large number of voxels are unvalued voxels for which frequency data was not collected. Only a small number of valuable voxels have frequency values ​​in the entire voxel space. The purpose of using density estimation or smoothing algorithms is to extend the interpolation of the values ​​in the valuable voxels to all unvalued voxels, thereby obtaining a continuous target spatial probability density distribution map in the voxel space. Because the structure and function of the cranial tissue have a certain degree of spatial continuity, the spatial Euclidean distance between each unvalued voxel and each valuable voxel needs to be considered during interpolation to ensure the smoothness of the final target spatial probability density distribution map. Based on this consideration, the choice of density estimation or smoothing algorithm can be optimized based on the final interpolation effect; Gaussian kernel density estimation or kernel smoothing algorithms are recommended.

[0054] S3. Perform tissue segmentation on the three-dimensional structural image of the target individual to construct a three-dimensional finite element head model.

[0055] It should be noted that tissue segmentation and the construction of a 3D finite element head model based on 3D structural image data obtained from magnetic resonance imaging (MRI) are existing technologies and can be implemented using existing segmentation algorithms and finite element simulation software. The 3D structural images used can be T1-weighted or T2-weighted structural MRI data acquired for the current target individual, and can be obtained from the data acquired in step S1. It is recommended to use both T1-weighted (T1w) and T2-weighted (T2w) structural MRI data for tissue segmentation. This combines the advantages of T1w's sensitivity to gray-white matter contrast and T2w's sensitivity to aqueous tissues (cerebrospinal fluid) and the boundaries of the inner and outer tables of the skull, thereby improving the accuracy of tissue segmentation.

[0056] In embodiments of this invention, based on T1-weighted and T2-weighted structural MRI data of the target individual, the cranial tissue of the target individual can be segmented into at least five tissue categories: scalp, skull, cerebrospinal fluid, gray matter, and white matter using a medical image segmentation algorithm. After smoothing and surface reconstruction of the segmented tissue masks, a three-dimensional finite element head model is generated through three-dimensional meshing. Finally, corresponding conductivity and dielectric properties are assigned to different tissue components. The medical image segmentation algorithm can employ 3D U-Net or SAM (Segment Anything Model) algorithms, etc. In addition to the five tissue categories of scalp, skull, cerebrospinal fluid, gray matter, and white matter, other tissues can be further subdivided to create a more refined head model.

[0057] S4. Using the target spatial probability density distribution map, the maximum envelope amplitude in the candidate target area is weighted. The goal is to maximize the weighted sum of the maximum envelope amplitude in the candidate target area while minimizing the average value of the envelope amplitude in the non-target cortical area. Based on the three-dimensional finite element head model, electric field simulation and parameter optimization are performed on the selected time interference stimulation parameters to obtain the optimal scheme of time interference stimulation parameters for the target individual.

[0058] It should be noted that during the parameter optimization process described above, for each set of candidate time-interference stimulation parameters, an electric field simulation based on a three-dimensional finite element head model is required to obtain the electric field distribution within the brain under that set of parameters. This allows for the comparison of the merits of different parameter schemes based on the optimization objective. The parameter types of the candidate time-interference stimulation parameters can be adjusted according to the actual adjustable stimulation parameters; for example, electrode position, number of electrodes, and current distribution between electrodes can all be included in the parameter optimization scope.

[0059] In this invention, the optimization objective of parameter optimization includes two sub-objectives. The first sub-objective is to maximize the weighted sum of the maximum envelope amplitudes within the candidate target region, thereby maximizing the weighted sum of the maximum envelope amplitudes of all voxels within the candidate target region. The second sub-objective is to minimize the average envelope amplitude in the non-target cortical region, thereby minimizing the average maximum envelope amplitude of all voxels in the non-target cortical region. This ensures that the stimulation electric field is focused within the candidate target region while suppressing stimulation intervention in other non-target regions as much as possible, reducing the exposure of non-target regions. Moreover, since this invention directly introduces the probability density distribution as a spatial weighting factor into the first sub-objective, the influence of all voxels on the objective function is no longer equivalent. Voxels that are frequently selected as individual candidate targets have higher weights. This ensures that the final stimulation electric field is not only highly focused on the deep target region, but also that the distribution pattern of the electric field intensity perfectly fits the abnormal weights of the patient's actual pathological network.

[0060] In embodiments of the present invention, each set of candidate time-interference stimulation parameters includes the combined positions of the stimulation electrode array and the current distribution ratio between the electrodes. During time-interference stimulation, the stimulation frequencies between the electrodes should be kept different, i.e., different frequency difference currents are input. The specific frequency difference can be determined by professional medical personnel according to the actual stimulation scheme. Therefore, when the combined positions of the stimulation electrode array and the current distribution ratio between the electrodes are used as time-interference stimulation parameters, the method for obtaining the optimal time-interference stimulation parameter scheme through parameter optimization is as follows:

[0061] For each set of candidate temporal interference stimulation parameters, an array of stimulation electrodes conforming to the positional requirements of the current stimulation parameters is arranged based on the aforementioned three-dimensional finite element head model. Sinusoidal current is injected according to the corresponding current distribution ratio and a preset difference frequency. The superimposed electric field distribution within the brain is obtained through electric field simulation, and then the maximum envelope amplitude AM of each voxel is calculated. Using the probability value of each voxel in the target spatial probability density distribution map as a weighting weight, the maximum envelope amplitude of all voxels within the candidate target region is weighted and summed. The weighted sum is then divided by the average maximum envelope amplitude of all voxels within the non-target cortical region to obtain the weighted peak-to-peak ratio (VPS). For each set of candidate time-interference stimulus parameters, after obtaining the corresponding weighted peak sum ratio, the time-interference stimulus parameter with the largest weighted peak sum ratio is selected as the optimal time-interference stimulus parameter scheme for the target individual.

[0062] The maximum envelope amplitude AM mentioned above can be calculated according to the corresponding principle formula. Assuming there are two electrodes, the time-interference stimulus will have different frequencies. and Two sinusoidal currents are injected through two electrode pairs, resulting in a superimposed electric field at any location r in the brain. It can be evaluated by the superposition of two electric fields, as shown in the following formula:

[0063]

[0064] in, and To optimize the injection current amplitude; and Each of the two electrode pairs is assigned a specific high-frequency carrier frequency. and This represents the electric field at position r. The superimposed electric field at position r. The calculation model for the maximum envelope amplitude AM is as follows:

[0065]

[0066] Where min() represents taking the smaller of the two electric field projection values ​​within the parentheses. Represents electric field and The angle between them; electric field and Together they determined a unique plane, in this Define a rotatable unit vector in the plane; represent The angle between this unit vector and the vector is used to... and The determined in-plane search for the optimal projection direction; Represents all Iterate through the values ​​to obtain the result. The largest Value, this value corresponding Multiply by 2 to get the corresponding maximum envelope amplitude AM.

[0067] Weighted peak and ratio ( The calculation formula for ) is as follows:

[0068]

[0069] in, It is a candidate target region pre-defined by combining prior information of the population; Corresponding to candidate target region A voxel within a space can be represented by spatial coordinates or a three-dimensional voxel index; It is a voxel in the spatial probability density distribution map of the target point The voxel value reflects the contribution of a specific coordinate to the network anomaly of that individual. In an embodiment of the invention, the target spatial probability density distribution map is normalized so that the sum of all voxel values ​​is 1, i.e. And in The integral within is 1. It is a voxel The maximum envelope amplitude at that point. Therefore, the molecular part This represents the weighted sum of the maximum envelope amplitudes of all voxels within the candidate target region. Additionally, the denominator... The maximum envelope amplitude is the average of all voxels within the non-target cortical region, calculated from the maximum envelope amplitude of all voxels within the remaining cortical region after removing the candidate target region. The average value is used to constrain the exposure amount in non-target areas.

[0070] In addition, when optimizing parameters, it is recommended to set safety constraints, such as setting the injection current. and The upper limit.

[0071] It should be noted that the method steps shown in S1 to S4 above can essentially be implemented in the form of computer programs or software functional modules.

[0072] Therefore, based on the same inventive concept, such as Figure 2 As shown, the present invention also provides a computer electronic device corresponding to the individualized time-interference stimulus parameter optimization method based on target probability distribution provided in the above embodiments, which includes a memory and a processor;

[0073] The memory is used to store computer programs;

[0074] The processor is configured to, when executing the computer program, implement the previously described method for optimizing individualized time-interference stimulation parameters based on target probability distribution.

[0075] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0076] Therefore, based on the same inventive concept, the present invention provides a computer-readable storage medium corresponding to the individualized time-interference stimulus parameter optimization method based on target probability distribution provided in the above embodiments. The storage medium stores a computer program, which, when executed by a processor, can realize the individualized time-interference stimulus parameter optimization method based on target probability distribution as described above.

[0077] Therefore, based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, can realize the individualized time-interference stimulus parameter optimization method based on target probability distribution as described above.

[0078] Specifically, in the computer-readable storage medium of the above three embodiments, the stored computer program is executed by a processor, which can perform the aforementioned steps S1 to S4.

[0079] It is understood that the aforementioned storage media may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Furthermore, the storage media may also be various media capable of storing program code, such as USB flash drives, external hard drives, magnetic disks, or optical discs.

[0080] It is understood that the processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0081] It should also be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. In the embodiments provided in this application, the division of steps or modules in the system and method is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple modules or steps may be combined or integrated together, and a module or step may also be split.

[0082] Furthermore, based on the same inventive concept, this invention provides a personalized temporal interference stimulation parameter optimization system based on target probability distribution. This system typically includes: a magnetic resonance imaging (MRI) scanner for acquiring multimodal brain data of an individual; a computing device (such as a high-performance computer, workstation, or cloud server) for running the personalized temporal interference stimulation parameter optimization method based on target probability distribution as described in S1-S4 above; and a TI (Transient Induced Stimulator) electrical stimulator that ultimately performs electrical stimulation according to the optimal temporal interference stimulation parameter scheme for the target individual. Through the coordinated operation of the above systems, this invention can overcome the problems of localization instability caused by traditional single scans, overcome errors caused by individual target spatial displacement, and achieve highly focused targeted intervention in deep brain regions.

[0083] The present invention will further demonstrate the detailed implementation process and technical effects of the individualized time-interference stimulus parameter optimization method based on target probability distribution shown in steps S1 to S4 above on specific data through a specific embodiment, so as to facilitate understanding of the essence of the present invention.

[0084] Example

[0085] In this embodiment, the specific process of the individualized time-interference stimulus parameter optimization method based on the target probability density distribution is as follows: Figure 3 As shown, it includes the following steps 1 to 4. Each step is described in detail below.

[0086] Step 1: Obtain individual candidate targets based on multi-trial functional connectivity analysis

[0087] This step involves acquiring and registering multiple trials of 3D structural images and resting-state functional magnetic resonance imaging (fMRI) data for the target individual. The temporal functional connectivity strength between the cortical effector and deep effector regions is calculated in each trial. Voxels of the cortical effector region with the strongest negative functional connectivity are selected as candidate target points for each individual trial. This step aims to eliminate random errors introduced by a single scan by acquiring brain network characteristics of the individual at different time points and calculating and selecting the optimal candidate target points for each scan. This step specifically includes the following sub-steps:

[0088] Step 101: Multiple Brain Imaging Data Acquisitions

[0089] Multimodal brain imaging data of the target individual were acquired at N different acquisition times. The acquired data included at least: high-resolution three-dimensional structural images (e.g., T1-weighted images acquired using MPRAGE sequences) and resting-state functional magnetic resonance imaging (fMRI) data. The time intervals between multiple acquisitions could be set according to clinical assessment needs (e.g., several days or weeks apart) to fully capture intra- or inter-diurnal fluctuations in the individual's functional connectivity patterns. In the subsequent actual test data of this embodiment, N=10.

[0090] Step 102: Multimodal image data preprocessing

[0091] To extract reliable oxygen-dependent (BOLD) signals, rigorous preprocessing was performed on the acquired structural and resting-state functional images using data processing software (such as SPM12 and the DPABI toolbox in MATLAB). The preprocessing steps specifically included: converting the raw DICOM images output from the MRI scanner to the NII standard format; for structural images: reorienting the anterior commissure to the origin, followed by tissue segmentation to generate probability maps containing gray matter, white matter, and cerebrospinal fluid, as well as corresponding inverse and forward transformation deformation fields; for resting-state functional images: firstly, deleting the first 10 time points to ensure the magnetic field signal reached a stable state, and allowing the subject to adapt to the scanning environment; performing time-layer correction, selecting intermediate slices as scanning reference layers to eliminate time delays between slices; and performing head movement correction, in which strict exclusion of... Data from any subject moving more than 3 mm in the x, y, or z directions and rotating more than 3 degrees on any axis were excluded to ensure signal quality. The corrected functional images were spatially registered with the corresponding 3D structural images and spatially normalized to the Montreal Neuroscience Institute (MNI) standard spatial brain template. The samples were then resampled to a 3 mm isotropic voxel size. White matter and cerebrospinal fluid signals, along with six motion parameters, were extracted to construct interference regressions, which were then removed. Spatial smoothing was performed using a Gaussian kernel with a full width at half maximum (FWHM) of 6 mm to improve the signal-to-noise ratio. Bandpass filtering was applied in the frequency range of 0.01 to 0.08 Hz to effectively separate low-frequency neural signals while reducing the effects of high-frequency physiological noise (such as heartbeat and respiration) and low-frequency instrument drift.

[0092] Step 103: Effect area delineation and time series functional connection calculation

[0093] In the standard MNI space, brain region masks responsible for different functions are defined. In this embodiment, taking the treatment of a specific neuropsychiatric disease (such as treatment-resistant depression) as an example, the dorsolateral prefrontal cortex (DLPFC) is selected as the cortical effect area, and the anterior cingulate cortex (ACC) is selected as the deep effect area. The left DLPFC mask is defined as follows: it contains gray matter voxels within three overlapping spheres, centered on the BA9 region (MNI coordinates: -36, 39, 43), the BA46 region (MNI coordinates: -44, 40, 29), and the standard 5 cm target area (MNI coordinates: -41, 16, 54), with each sphere having a radius of 20 mm. The left anterior cingulate cortex (ACC) region is precisely defined using an AAL (Automated Anatomical Labeling) brain atlas template. Based on the resting-state functional image data of each trial after preprocessing in step 102, the time-domain signal sequences of all voxels in the cortical effector and deep effector are extracted respectively. The time-domain signal sequences of all voxels in the deep effector are then averaged to obtain the average time-domain signal sequence of the deep effector. Next, Pearson correlation analysis is used to calculate the correlation coefficient r between the time-domain signal sequence of each voxel in the cortical effector and the average time-domain signal sequence of the deep effector. To satisfy the statistical assumption of normal distribution, Fisher's z-transformation is used to convert the r-value corresponding to each voxel in the cortical effector into a z-score, which represents the temporal functional connectivity strength between this voxel in the cortical effector and the deep effector. The temporal functional connectivity strengths corresponding to all voxels in the cortical effector constitute a connectivity matrix recording the connectivity strength between different voxels in the cortical effector and the deep effector. A connectivity matrix needs to be calculated for each trial.

[0094] Step 104: Individual candidate target screening under constraints

[0095] Based on the core circuit characteristics of diseases such as depression (e.g., the negative connectivity characteristics of the anterior cingulate cortex below the genu and the dorsolateral prefrontal cortex), in each trial, the voxel with the strongest negative functional connectivity is selected as the candidate stimulation target from the connectivity matrix calculated. This is achieved by finding the minimum value (the largest absolute value among all negative values) in the connectivity matrix, and the corresponding voxel is the individual candidate target for the current trial. However, in this embodiment, to ensure effective reach of subsequent non-invasive electrical stimulation of the scalp and enhance the feasibility of stimulation, spatial constraints are imposed on the selected individual candidate targets: the individual candidate targets must simultaneously satisfy the conditions of being located within the gray matter of the brain and having a spatial location within 4 cm of the scalp surface. Therefore, voxels that do not meet the spatial constraints should be excluded from the cortical effect area. Then, in each trial, the time-series functional connectivity strength values ​​of the remaining voxels need to be extracted from the connectivity matrix, and the minimum value within the negative domain is found. The voxel corresponding to this minimum value is then used as the final individual candidate target for the current trial.

[0096] The above steps are performed on the data collected in multiple trials to obtain the coordinates of multiple candidate target points for the individual in different trials.

[0097] Step 2: Construct an individual target spatial probability density distribution map by combining prior information from the population.

[0098] In this step, within the pre-defined candidate target region based on population prior information, the frequency with which each voxel is identified as an individual candidate target across all trials is counted. Then, spatial interpolation is used to generate a spatial probability density distribution map of the target individual. This step aims to overcome the limitations of traditional single-target localization by constructing a target spatial model that fits the dynamic characteristics of the individual patient through a computational strategy that transitions from population priors to individual features. This step specifically includes the following sub-steps:

[0099] Step 201: Delineation of prior candidate regions based on large queue populations

[0100] Acquire population imaging data from a large cohort sample; for example, extract resting-state functional imaging data from a major depressive disorder (MDD) patient group and a healthy control (HC) group, and calculate the average functional connectivity network. By comparing network differences between the two groups, abnormal brain regions with high prior probability are delineated in standard three-dimensional brain space and designated as target candidate regions (ROIs). This step utilizes the large-scale characteristics of the population to limit the macroscopic pathological spatial range for individual target optimization, avoiding computational redundancy and localization bias caused by blind whole-brain searches.

[0101] Step 202: Dynamic feature weighting based on individual multi-trial scans

[0102] Based on the aforementioned defined candidate target region for the target population, all individual candidate target points (i.e., the output of step 104) obtained from multiple independent resting-state functional magnetic resonance imaging (fMRI) scans of the target individual are mapped to the candidate target region according to their coordinates, such as... Figure 4 As shown in the figure. Then, the spatial distribution drift characteristics and recurrence rate of the individual's candidate target points in multiple scans are analyzed. Specifically, for each voxel in the target candidate region, the frequency of its identification as an individual candidate target is counted. This frequency reflects the probability that the corresponding voxel is a candidate target. Then, based on the voxels with the counted frequencies, combined with the spatial Euclidean distance between each non-valued voxel and each valued voxel, a 3D Gaussian kernel density estimation algorithm is used to assign a value to each voxel in the target candidate region.

[0103] Step 203: Generate spatial probability density distribution map of individual target points

[0104] Based on the above assignment results, the estimated values ​​of all voxels are normalized to assign each voxel a target weight between 0 and 1. The sum of the target weights of all voxels is 1, thus generating an individual target probability density distribution map for the subject in three-dimensional brain space. In this distribution model, regions with high weights represent the core brain regions where the individual's brain network functional connectivity abnormalities are most stable and the expected response to non-invasive electrical stimulation is most significant; while regions with low weights represent regions where abnormal features are sporadic or fluctuate greatly. This probability density distribution map will serve as the core high-precision three-dimensional spatial targeting constraint condition for subsequent electromagnetic field parameter optimization.

[0105] Step 3: Construct an individual head anatomical simulation model

[0106] In this step, tissue segmentation is performed on the three-dimensional structural image of the target individual to construct a three-dimensional finite element head model. To accurately assess and optimize the electric field distribution generated in the brain by temporal interference stimuli, a physical electromagnetic simulation model that conforms to the individual's actual anatomical structure must be established. This step specifically includes the following sub-steps:

[0107] Step 301: High-precision fine segmentation of cranial tissue

[0108] T1-weighted (T1w) and T2-weighted (T2w) structural MRI data of the subject's head were acquired. Leveraging the combined advantages of T1w's sensitivity to gray-white matter contrast and T2w's sensitivity to aqueous tissues (cerebrospinal fluid) and the boundaries of the inner and outer tables of the skull, a medical image segmentation algorithm was used to finely segment the subject's cranial tissue into five main layers: scalp, skull (which can be further subdivided into cancellous and compact bone), cerebrospinal fluid, gray matter, and white matter. Figure 5 As shown.

[0109] Step 302: Three-dimensional finite element modeling and dielectric property assignment

[0110] The segmented 3D tissue masks at each level are smoothed and their surfaces reconstructed, followed by 3D meshing to generate a high-fidelity 3D finite element head model containing millions of tetrahedral or hexahedral elements, such as... Figure 5 As shown. Subsequently, by consulting standard dielectric constant tables in the field of low-frequency electromagnetics, different tissue components (scalp, skull, cerebrospinal fluid, gray matter, white matter, etc.) were assigned isotropic or anisotropic dielectric properties such as conductivity. This individual head anatomy simulation model will realistically reflect the shunting effect of each layer of skull tissue on the injected current, thus providing accurate physical boundary conditions for subsequent calculations and simulations.

[0111] Step 4: Optimization of individual temporal interference stimulation parameters based on target probability density distribution

[0112] In this step, after completing the spatial modeling of the individual target probability density and the construction of the head finite element model, it is necessary to use the target spatial probability density distribution map to weight the maximum envelope amplitude within the candidate target region. The goal is to maximize the weighted sum of the maximum envelope amplitudes within the candidate target region while minimizing the average envelope amplitude in the non-target cortical region. Based on the three-dimensional finite element head model, electric field simulation and parameter optimization are performed on the selected time-interference stimulus parameters to obtain the optimal solution for the target individual's time-interference stimulus parameters. This step introduces the target probability density distribution as a spatial weighting factor into the optimization algorithm to maximize the electric field envelope intensity in the high-weighted target probability density region while minimizing the electric field intensity in the non-target region and the low-weighted region. Figure 6 As shown. This step specifically includes the following simulation calculations and parameter optimization processes:

[0113] Step 401: Calculate the superimposed electric field and envelope amplitude of the time interference.

[0114] The physical basis of temporal interference stimulation lies in using two or more sets of high-frequency alternating currents to generate low-frequency envelopes deep within the brain for neural modulation. In this embodiment, different frequencies are used during temporal interference stimulation. and Two sinusoidal currents are injected through two electrode pairs. and The frequencies are 2000Hz and 2040Hz respectively, generating a difference frequency envelope of 40Hz. This superimposed electric field is generated at any voxel r within the brain. It can be calculated using the aforementioned formula (1). Maximum envelope amplitude The calculation model is shown in the aforementioned formula (2).

[0115] Step 402: Construct the objective function weighted by probability density distribution

[0116] Traditional time-interference electric field optimization only aims to maximize the electric field intensity of a single target region. To ensure the stimulus electric field perfectly envelops the target region generated based on individual dynamic network characteristics, this embodiment establishes an envelope amplitude spatial integral objective function weighted by probability density distribution, namely, the weighted peak-to-sum ratio (P0). The calculation method is as shown in the aforementioned formula (3). In the calculation... At that time, since voxels are discrete within the actual region, the molecular part... Points can be earned through Weighted maximum envelope amplitude corresponding to all voxels Summation yields the result. Additionally, the denominator... The maximum envelope amplitude of all voxels in the remaining region after removing the candidate target region from the cortical region. The average is obtained. Dividing the numerator by the denominator gives the corresponding... .

[0117] Step 403: Iterative optimization of parameters under multiple constraints

[0118] Under safety constraints (in this embodiment, the injection current is set), and Each parameter is limited to a maximum current of 2mA. The system uses optimization algorithms (such as genetic algorithms, particle swarm optimization, or gradient descent) to iteratively calculate the combination positions and current distribution ratios of various head dial electrode arrays. Through continuous iterative solutions, the above objective function is optimized. The goal is to maximize the parameter solution and find the stimulation parameter scheme that best fits the probability density distribution of the target point of the individual (i.e., the optimal electrode placement position and the current injection intensity of each electrode).

[0119] The effectiveness of the method in this embodiment compared to the single-target optimization method will be verified below.

[0120] As mentioned earlier, traditional time-interference stimulation parameter optimization methods typically aim to maximize the electric field intensity at a single target point (hereinafter referred to as the "single-target optimization method"). To further verify the effectiveness of the method in this embodiment (an optimization method based on target probability density distribution) compared to the single-target optimization method, this embodiment constructs a set of candidate target points from multiple scans of a subject on an example head model. Parameter optimization is performed using both methods, and the electric field envelope amplitude (AM) generated by the optimal electrode schemes obtained by the two methods is compared at all candidate target points. This embodiment uses the MNI152 standard head model provided by the open-source transcranial electrical stimulation tool ROAST and its pre-generated lead field matrix. The matrix element dimension is [number of nodes × 3 × number of electrodes] = [444882 × 3 × 71], corresponding to the 71 electrode positions in the EEG 10-10 standard system. Tissue label information is extracted from the finite element tetrahedral mesh corresponding to this head model.

[0121] Considering the scarcity of multiple scan data collected from real-world subjects, this embodiment employs controlled synthesis to generate candidate targets from multiple scans. Based on the published clinical observation that individual sgACC target drift can reach 25–37 mm, and combined with the anatomical characteristic of the band-like structures of the prefrontal cortex spreading along the anteroposterior axis, this embodiment uses a mixture Gaussian distribution sampling: principal band components: centered at [-6, 16, -10] mm (the prior peak value of the left sgACC population), with a standard deviation of... =[14, 4, 4] mm, extract 7 candidate target points; distal cluster component: centered at [-6, 32, -10] mm, standard deviation =[4, 4, 4] mm, extract 3 candidate target points. After synthesis, a constraint is applied: all candidate target points must be located within the gray matter of the left hemisphere and at a depth not exceeding 40 mm from the scalp. The final spatial distribution of N=10 candidate target points is as follows: Figure 7 As shown.

[0122] As described in step 2 of this method, a target candidate region (VROI) is defined with [-6, 16, -10] as the center and a radius of 25 mm, and the intersection with the gray matter yields 4881 candidate nodes. Within the VROI, 3D Gaussian kernel density estimation (bandwidth h = 6 mm) is performed on 10 candidate target points, and the probability density distribution P(r) is obtained after normalization. This P(r) forms a main peak in the main strip component region and a secondary peak in the distal clique component region, exhibiting an overall asymmetric bimodal shape. Its geometric centroid position ([-9.4, 22.9, -12.0] mm) lies between the two component centers, deviating from the main peak position of P(r) by approximately 8 mm. Figure 7 As shown.

[0123] This embodiment compares the following two methods: 1. Single target optimization method (represented by the centroid method): using the geometric centroids of 10 candidate target points. As the optimization objective, maximize 2. This method: weighted integral with P(r) As the objective function. The average electric field value for non-target gray matter regions (calculated by randomly sampling 5000 nodes in the whole brain gray matter) is used to suppress the electric field intensity in non-target areas. The simulation parameters used in this embodiment include: both electrode pairs have a current of 2 mA. Electrode pair combinations are determined by traversing all non-overlapping 4-electrode pairs (approximately 2.9 million combinations) of the 71 electrodes in the EEG 10-10 system. For each combination, a score is calculated according to the aforementioned objective function, and the electrode configuration corresponding to the highest score is taken as the optimal solution.

[0124] electric field distribution comparison Figure 8 As shown, the optimal scheme obtained by the single-target method only forms a concentrated, moderate-intensity thermal zone (peak value approximately 0.27 V / m) around the centroid, failing to cover the main strip region elongated along the X-axis and the distal cluster region. The optimal scheme obtained by this method covers the entire high-density P(r) region, forming a distinct strong field band (peak value approximately 0.39 V / m) along the main strip direction, while also exhibiting considerable electric field intensity in the distal cluster region. Quantitative comparison... Figure 9 As shown, Figure 9 (A) shows the AM values ​​of the two methods at 10 candidate target sites. Our method achieved higher AM values ​​than the single-target method at 6 out of the 10 target sites, equal values ​​at 4 out of 10 target sites, and lower values ​​at none than the single-target method. Figure 9 In section (B), the 10 targets were divided into three groups based on their P-weights for mean comparison: the high P group (≥75% peak) showed an improvement of approximately 10%, the medium P group (60–75%) showed an improvement of approximately 30%, and the low P group (<60%) showed an improvement of approximately 41%. Overall, the mean AM of this method at the 10 candidate targets was 0.184 V / m, while the single-target method was 0.148 V / m, resulting in an overall improvement of approximately 24%.

[0125] The embodiments described above are merely some preferred implementations of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A method for optimizing individualized temporal interference stimulus parameters based on target probability distribution, characterized in that, include: S1. Acquire and register three-dimensional structural images and resting-state functional magnetic resonance imaging data of the target individual in multiple trials, calculate the time-series functional connectivity strength between the cortical effect area and the deep effect area in each trial, and screen out the cortical effect area voxel with the strongest negative functional connectivity as the candidate target point for each individual in each trial. S2. Within the candidate target area pre-defined by combining the prior information of the group, count the frequency of each voxel being identified as an individual candidate target in all trials, and then generate the target spatial probability density distribution map of the target individual through spatial interpolation. S3. Perform tissue segmentation on the three-dimensional structural image of the target individual to construct a three-dimensional finite element head model; S4. Using the target spatial probability density distribution map, the maximum envelope amplitude in the candidate target area is weighted. The goal is to maximize the weighted sum of the maximum envelope amplitude in the candidate target area while minimizing the average value of the envelope amplitude in the non-target cortical area. Based on the three-dimensional finite element head model, electric field simulation and parameter optimization are performed on the selected time interference stimulation parameters to obtain the optimal scheme of time interference stimulation parameters for the target individual.

2. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 1, characterized in that, In S1, the method for extracting individual candidate targets for each trial is as follows: For the cortical effect area and deep effect area corresponding to the target intervention disease, the average time-domain signal sequence in the deep effect area is extracted from the resting-state functional magnetic resonance data of the current trial. Then, the correlation coefficient between the time-domain signal sequence of each voxel in the cortical effect area and the average time-domain signal sequence is calculated through correlation analysis. The correlation coefficient corresponding to each voxel is then converted by a standardization transformation method and used as the time-series functional connectivity strength corresponding to that voxel. The voxel with the smallest negative time-series functional connectivity strength in the cortical effect area that satisfies the spatial constraints is selected as the individual candidate target for the current trial.

3. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 2, characterized in that, The target disease to be intervened is depression, the corresponding cortical effect area is the dorsolateral prefrontal cortex, the corresponding deep effect area is the anterior cingulate cortex, and the spatial constraint is set as follows: the candidate target point is located in the gray matter of the brain and its spatial position is within 4 cm of the scalp surface.

4. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 1, characterized in that, In S2, the method for delineating candidate target regions by combining prior information of the population is as follows: obtain population image data including the patient group of the target intervention disease and the healthy control group, calculate the average functional connectivity network of the two groups based on the resting state functional image data, and delineate abnormal brain regions as candidate target regions in the standard three-dimensional brain space according to the network difference comparison results.

5. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 1, characterized in that, In S2, the method for generating the target spatial probability density distribution map of the target individual is as follows: map all individual candidate target points identified in all trials to the candidate target area, count the frequency of each voxel in the candidate target area being identified as an individual candidate target, and then use the frequency to represent the probability of being selected, and use density estimation or smoothing algorithm to generate a target spatial probability density distribution map covering the entire candidate target area.

6. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 1, characterized in that, In step S3, based on the T1-weighted and T2-weighted structural magnetic resonance imaging data of the target individual, the cranial tissue of the target individual is segmented into at least the scalp, skull, cerebrospinal fluid, gray matter, and white matter using a medical image segmentation algorithm. After smoothing and surface reconstruction of the segmented tissue mask, a three-dimensional finite element head model is generated through three-dimensional meshing. Finally, corresponding conductivity attributes are assigned to different tissue components.

7. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 1, characterized in that, In S4, each set of candidate time-interference stimulation parameters includes the combination position of the stimulation electrode array and the current distribution ratio between the electrodes.

8. The method for optimizing individualized temporal interference stimulation parameters based on target probability distribution as described in claim 7, characterized in that, In S4, the method for obtaining the optimal scheme of time interference stimulation parameters through parameter optimization is as follows: for each set of candidate time interference stimulation parameters, a stimulation electrode array that meets the position requirements of the current stimulation parameters is arranged on the basis of the three-dimensional finite element head model, and sinusoidal current is injected according to the corresponding current distribution ratio and the preset difference frequency. The superimposed electric field distribution in the brain is obtained through electric field simulation, and then the maximum envelope amplitude of each voxel is calculated. Using the probability value of each voxel in the target spatial probability density distribution map as a weighting weight, the maximum envelope amplitude of all voxels in the candidate target region is weighted and summed. Then, the weighted sum is divided by the average maximum envelope amplitude of all voxels in the non-target cortical region to obtain the weighted peak sum ratio. The time-interference stimulation parameter with the largest weighted peak sum ratio is selected as the optimal time-interference stimulation parameter scheme for the target individual.

9. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it can implement the individualized time-interference stimulus parameter optimization method based on the target probability distribution as described in any one of claims 1 to 8.

10. A computer electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to, when executing the computer program, implement the individualized time-interference stimulus parameter optimization method based on target probability distribution as described in any one of claims 1 to 8.

Images