Transcranial direct current stimulation intervention system

Abstract

This application relates to the field of transcranial direct current stimulation technology. Specifically, it discloses a transcranial direct current stimulation intervention system, which uses data processing and image analysis algorithms based on machine learning and artificial intelligence technologies to analyze the pre- and post-treatment brain structural networks and brain functional networks marked as effective, so as to capture the pre-treatment brain structure-functional network characteristics. When actually identifying the possibility of positive brain responses and detecting potential effectiveness for new individuals, the brain network state characteristics of new individuals are reflected by extracting and splicing the characteristics of the brain structural network and brain functional network of new individuals, and the brain network state characteristics of the new individuals are used as an index to query the feature information with similar responses from the brain network state database with effective interventions, so as to identify the possibility of positive brain responses of new individuals, thereby predicting whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated.

Description

Technical Field

[0001] This application relates to the technical field of transcranial direct current stimulation, and more specifically, to a transcranial direct current stimulation intervention system. Background Art

[0002] Transcranial Direct Current Stimulation (tDCS) is a non-invasive brain stimulation technique that applies a weak direct current through electrodes on the scalp to regulate the activity level of cortical neurons. tDCS has been used for various clinical and research purposes, including improving cognitive function, treating depression, alleviating chronic pain, etc. Although tDCS has potential, its effects vary from person to person, and it may not have a significant effect on some individuals.

[0003] Chinese Patent CN111870808B provides an intervention method for transcranial direct current stimulation. By combining non-invasive brain network detection (magnetic resonance scanning and brain network analysis), it pre-evaluates the treatment response rate of tDCS for an individual, so as to achieve the purpose of applying tDCS to intervene in emotional problems in a targeted manner. Simply put, it is to specifically exclude individuals who may not respond to tDCS intervention, greatly avoiding the waste of medical resources.

[0004] However, in the above-mentioned intervention method of transcranial direct current stimulation, it specifically analyzes the multi-person big data of brain images based on two aspects of brain structural network and brain functional network image data, so as to capture the morphological characteristics of the brain network distribution that has a positive response to the classical tDCS intervention method, and is used to identify the possibility of positive response of the brain of a new individual. This method of capturing the morphological characteristics of the brain network distribution with positive response and comparing them with the morphological characteristics of the brain network distribution of a new individual to identify the possibility of positive response, but because the brain structure and functional connection patterns of each individual are different, the characteristics captured based on group data may not fully reflect the uniqueness of the individual. That is to say, this method of comparing the morphological characteristics of the group brain network distribution may be difficult to fully consider the subtle differences between individuals, thus affecting the reliability of the assessment of the response possibility of new individuals.

[0005] Therefore, an optimized transcranial direct current stimulation intervention system is desired. Summary of the Invention

[0006] The present application provides a transcranial direct current stimulation intervention system, which can match the similar brain network states of individual patients from the database of known effective tDCS intervention on brain network states through a more intelligent query response method, which helps to assist in predicting the potential effectiveness of tDCS intervention for different individual patients, especially for those who may not respond to the standard treatment plan.

[0007] According to an aspect of the present application, there is provided a transcranial direct current stimulation intervention system, including:

[0008] An effective intervention brain network data acquisition module, configured to obtain a data set of {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} labeled as effective;

[0009] An effective intervention brain network analysis module, configured to perform feature extraction and aggregation analysis on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the data set of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} labeled as effective to obtain a set of pre-treatment brain structural-functional network features;

[0010] A patient object brain network data acquisition module, configured to obtain the brain structural network and brain functional network of a patient object to be evaluated;

[0011] A patient object brain network analysis module, configured to perform feature extraction and feature splicing on the brain structural network and brain functional network of the patient object to be evaluated to obtain the encoded features of the brain structural-functional network of the patient object to be evaluated;

[0012] A patient object brain state query response module, configured to perform fast query dynamic response processing on the encoded features of the brain structural-functional network of the patient object to be evaluated and the set of pre-treatment brain structural-functional network features to obtain the encoded features of the brain state query response of the patient object to be evaluated;

[0013] An intervention effectiveness detection module, configured to determine whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated based on the encoded features of the brain state query response of the patient object to be evaluated.

[0014] A transcranial direct current stimulation intervention system provided by the present application uses data processing and image analysis algorithms based on machine learning and artificial intelligence technologies to analyze the pre-treatment and post-treatment brain structural networks and brain functional networks marked as effective, so as to capture the pre-treatment brain structure-functional network features, which are used to reflect the anatomical connectivity and functional interaction between different brain regions where these tDCS interventions are effective, thereby forming a database of effective intervention brain network states. When actually identifying the possibility of positive brain responses and detecting potential effectiveness for a new individual, the brain network state features of the new individual are reflected by extracting and splicing the features of the brain structural network and brain functional network of the new individual, and using the brain network state features of the new individual as an index to query similar feature information from the database of effective intervention brain network states, so as to identify the possibility of positive brain responses of the new individual, thereby predicting whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated. In this way, it is possible to match the similar brain network states of personalized patients from the database of known effective tDCS intervention brain network states through a more intelligent query response method, which helps to assist in predicting the potential effectiveness of tDCS intervention for different individual patients, especially for those individuals who may not respond to the standard treatment plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application and do not limit the present application.

[0016] Figure 1 It is a schematic block diagram of the transcranial direct current stimulation intervention system of the embodiment of the present application.

[0017] Figure 2 It is a schematic diagram of the data flow of the transcranial direct current stimulation intervention system of the embodiment of the present application.

[0018] Figure 3 It is a schematic block diagram of the effective intervention brain network analysis module in the transcranial direct current stimulation intervention system of the embodiment of the present application.

[0019] Figure 4 It is a schematic block diagram of the patient object brain state query response module in the transcranial direct current stimulation intervention system of the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts also belong to the scope of protection of the present application.

[0021] In response to the above technical problems, in the technical solution of the present application, a transcranial direct current stimulation intervention system is proposed, which can evaluate the potential effectiveness of classical tDCS intervention paradigms for different patients and the possibility of positive brain responses by using machine learning and artificial intelligence technologies. Specifically, the transcranial direct current stimulation intervention system uses data processing and image analysis algorithms based on machine learning and artificial intelligence technologies to analyze the pre- and post-treatment brain structural networks and brain functional networks labeled as effective, so as to capture the pre-treatment brain structure-functional network characteristics, which are used to reflect the anatomical connectivity and functional interaction between different brain regions where these tDCS interventions are effective, thereby forming a database of effective brain network states for intervention. When actually identifying the possibility of positive brain responses and detecting potential effectiveness for a new individual, the brain network state characteristics of the new individual are reflected by extracting and splicing the characteristics of the new individual's brain structural network and brain functional network, and the brain network state characteristics of the new individual are used as an index to query similar feature information from the database of effective brain network states for intervention, so as to identify the possibility of positive brain responses of the new individual, thereby predicting whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated. In this way, it is possible to match the similar brain network states of personalized patients from the known effective tDCS intervention brain network state database through a more intelligent query response method, which helps to assist in predicting the potential effectiveness of tDCS intervention for different individual patients, especially for those individuals who may not respond to the standard treatment plan.

[0022] As Figure 1 and Figure 2As shown, the transcranial direct current stimulation intervention system includes: an effective intervention brain network data acquisition module 10, configured to obtain datasets of {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} marked as effective; an effective intervention brain network analysis module 20, configured to perform feature extraction and aggregation analysis on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the datasets marked as effective to obtain a set of pre-treatment brain structure-functional network features; a patient object brain network data acquisition module 30, configured to obtain the brain structural network and brain functional network of a patient object to be evaluated; a patient object brain network analysis module 40, configured to perform feature extraction and feature splicing on the brain structural network and brain functional network of the patient object to be evaluated to obtain the brain structure-functional network coding features of the patient object to be evaluated; a patient object brain state query response module 50, configured to perform fast query dynamic response processing on the brain structure-functional network coding features of the patient object to be evaluated and the set of pre-treatment brain structure-functional network features to obtain the brain state query response coding features of the patient object to be evaluated; and an intervention effectiveness detection module 60, configured to determine whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated based on the brain state query response coding features of the patient object to be evaluated.

[0023] Exemplarily, in the effective intervention brain network data acquisition module 10, datasets of {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} marked as effective are obtained. It should be understood that the efficacy of transcranial direct current stimulation (tDCS) varies due to individual differences. Not all individuals who receive tDCS intervention will show significant therapeutic effects. By collecting and analyzing the pre-treatment brain structural network and pre-treatment brain functional network, brain network-related feature information effective for tDCS intervention can be extracted, which helps to predict the potential efficacy of new patients. Obtaining these data can help construct a "dataset of brain network states effective for intervention". This dataset contains the brain structural and functional network features of individuals known to have a good response to tDCS. When evaluating new patients, it is possible to predict whether they will have a positive response to tDCS by comparing their brain network states with the features in the database. Pre-screening individuals who may not respond to tDCS can avoid unnecessary treatment attempts, save medical resources and personal time, and at the same time improve the success rate of clinical treatment. For patients who may not benefit from the standard tDCS regimen, other more suitable treatment methods can be explored, thus achieving more personalized medical care.

[0024] Exemplarily, in the effective intervention brain network analysis module 20, feature extraction and aggregation analysis are performed on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the dataset marked as effective to obtain a set of pre-treatment brain structure-functional network features. It should be understood that by performing feature extraction and aggregation analysis on the pre-treatment brain structure and functional network datasets marked as effective, the anatomical connectivity and functional interaction features between those brain regions that have a positive response to tDCS intervention can be captured to form a set containing effective pre-treatment brain structure-functional network features, which helps to identify whether a new individual has a similar positive response potential before receiving tDCS treatment..

[0025] In one embodiment, as Figure 3 shown, the effective intervention brain network analysis module 20 includes: a pre-treatment brain feature extraction unit 21, configured to perform feature extraction on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the dataset marked as effective to obtain a set of pre-treatment brain structural network feature vectors and a set of pre-treatment brain functional network feature vectors; a pre-treatment brain feature splicing unit 22, configured to splice the corresponding pre-treatment brain structural network feature vectors and pre-treatment brain functional network feature vectors in each group of the set of pre-treatment brain structural network feature vectors and the set of pre-treatment brain functional network feature vectors to obtain a set of pre-treatment brain structure-functional network feature vectors as the set of pre-treatment brain structure-functional network features.

[0026] Exemplarily, in the pre-treatment brain feature extraction unit 21, feature extraction is performed on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the dataset labeled as valid to obtain a set of pre-treatment brain structural network feature vectors and a set of pre-treatment brain functional network feature vectors. It should be understood that since each data in the dataset labeled as valid for the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} is image data, among which the brain structural network (such as the nerve fiber network) and the brain functional network (such as the functional connection between different regions of the brain) are from different individuals who have received tDCS intervention and are effective. In order to be able to extract brain network-related feature information about the effectiveness of tDCS intervention from each image data in these datasets, so as to provide a basis for the subsequent evaluation of the effectiveness of tDCS intervention for new individual patient objects, in the technical solution of this application, feature extraction is performed on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the dataset labeled as valid to obtain a set of pre-treatment brain structural network feature vectors and a set of pre-treatment brain functional network feature vectors. In particular, in a specific example of this application, a brain network feature extractor based on a dilated convolutional neural network model can be used to perform feature extraction on different brain network image data respectively, so as to capture the pre-treatment brain structural network features and brain functional network features labeled as valid, which are used to reflect the brain network information of different individuals with effective tDCS intervention, such as the connection tightness and activity level of specific regions of the brain. These features help to construct a database for personalized query matching and response of the brain network state in the subsequent stage. For example, a higher connection strength between certain regions, or a higher activation level shown in certain regions may be a sign of an effective tDCS response.

[0027] In one embodiment, the pre-treatment brain feature extraction unit is configured to: extract each pre-treatment brain structural network and each pre-treatment brain functional network from a dataset, and respectively perform feature extraction on the each pre-treatment brain structural network and the each pre-treatment brain functional network through a brain network feature extractor based on a dilated convolutional neural network model to obtain a set of pre-treatment brain structural network feature vectors and a set of pre-treatment brain functional network feature vectors. It should be understood that the pre-treatment brain structural network and the pre-treatment brain functional network include high-quality diffusion tensor imaging (DTI) and functional magnetic resonance imaging (fMRI) data. These imaging data reflect the strength of nerve fiber connections between different regions of the brain (structural network) and the change of blood oxygenation level-dependent signal over time (functional network). The pre-treatment brain structural network and the pre-treatment brain functional network of each individual are respectively input into the brain network feature extractor based on the dilated convolutional neural network model. This model uses the dilated convolution technique to expand the receptive field without reducing the resolution, so as to be able to capture spatial information in a larger range. After being processed by the brain network feature extractor based on the dilated convolutional neural network model, what is output is a set of pre-treatment brain structural network feature vectors and a set of brain functional network feature vectors for each individual. These feature vectors not only retain the key information in the original images, but also are abstracted and concentrated through deep learning algorithms, making the subsequent aggregation analysis and comparison more efficient and accurate.

[0028] Exemplarily, in the pre-treatment brain feature splicing unit 22, for each pair of corresponding pre-treatment brain structural network feature vectors and pre-treatment brain functional network feature vectors in the set of pre-treatment brain structural network feature vectors and the set of pre-treatment brain functional network feature vectors, feature splicing is performed to obtain a set of pre-treatment brain structural-functional network feature vectors as the set of pre-treatment brain structural-functional network features. It should be understood that since the brain structural network (such as white matter fiber connections) and the functional network (such as the temporal correlation of neuronal activities) provide different but complementary information about the brain. The structural network reflects the physical connection mode of the brain, while the functional network shows the dynamic interaction of these connections in actual operation. In order to effectively fuse these two types of network features to obtain a more comprehensive description and representation of the brain state, in the technical solution of this application, further, for each pair of corresponding pre-treatment brain structural network feature vectors and pre-treatment brain functional network feature vectors in the set of pre-treatment brain structural network feature vectors and the set of pre-treatment brain functional network feature vectors, feature splicing is performed to obtain a set of pre-treatment brain structural-functional network feature vectors. Through the feature fusion method of feature splicing, the pre-treatment brain structural network features and functional network features can be combined, so as to capture more complex neural activity patterns, avoid the problem that considering only the structural or functional features of the brain network is not sufficient to fully characterize the individual's response to tDCS intervention, and provide a basis for predicting the effectiveness of individual tDCS intervention for subsequent patients.

[0029] In one embodiment, first, each feature vector in the set of pre-treatment brain structural network feature vectors and the set of pre-treatment brain functional network feature vectors is aligned according to the same individual and brain region. Next, for each pair of corresponding pre-treatment brain structural network feature vectors and pre-treatment brain functional network feature vectors, they are combined into a new and more rich pre-treatment brain structural-functional network feature vector through a splicing operation. This new pre-treatment brain structural-functional network feature vector contains both structural information (such as nerve fiber connection strength) and functional information (such as synchronous activities between different brain regions), so as to be able to more comprehensively describe the state of the brain.

[0030] Exemplarily, in the patient object brain network data acquisition module 30, the brain structure network and the brain function network of the patient object to be evaluated are acquired. It should be understood that by acquiring the brain structure network and the brain function network of the patient to be evaluated, the current state of the brain of the patient object to be evaluated can be understood in detail. These information provide important basic data for subsequent analysis, enabling doctors to formulate more precise treatment plans based on individual differences. At the same time, by acquiring the brain structure network and the brain function network of the patient object to be evaluated, it can be compared with known effective tDCS intervention cases. In this application, machine learning algorithms are used to match the similarity between the brain network characteristics of the new patient and those of individuals who have a good response to tDCS. If a high degree of similarity is found between the two, it indicates that this new patient may also benefit from the same tDCS intervention; conversely, if the matching degree is low, other treatment methods may need to be considered or the tDCS parameters may need to be adjusted to meet individual needs.

[0031] Exemplarily, in the patient object brain network analysis module 40, feature extraction and feature splicing are performed on the brain structure network and the brain function network of the patient object to be evaluated to obtain the brain structure-functional network coding features of the patient object to be evaluated. It should be understood that when evaluating and predicting the classical tDcs intervention paradigm for the patient object to be evaluated, by acquiring the brain structure network and the brain function network of the patient object to be evaluated, and performing feature extraction and feature splicing on the brain structure network and the brain function network of the patient object to be evaluated to obtain the brain structure-functional network coding vector of the patient object to be evaluated. It should be understood that since the brain structure and function characteristics of each person are unique, it is difficult to fully represent the uniqueness of an individual relying only on a single type of feature (such as only considering the brain structure network or only considering the brain function network). By analyzing the structure and function networks simultaneously, the subtle differences between individuals can be captured more comprehensively, thus realizing a more personalized prediction of the effectiveness of tDcs intervention. In addition, the same processing method as that for the dataset labeled as effective above is adopted here, which can provide a more accurate data basis in the subsequent brain state query matching process, facilitating the query response of thickness and the prediction and evaluation of the effectiveness of tDcs intervention.

[0032] In one embodiment, the patient object brain network analysis module is configured to: after using the brain network feature extractor based on the dilated convolutional neural network model to perform feature extraction on the brain structure network and the brain function network of the patient object to be evaluated respectively, then splice the obtained brain structure network feature vector of the patient object to be evaluated and the brain function network feature vector of the patient object to be evaluated to obtain the brain structure-functional network coding vector of the patient object to be evaluated as the brain structure-functional network coding features of the patient object to be evaluated.

[0033] Exemplarily, in the patient object brain state query response module 50, a fast query dynamic response process is performed on the set of the brain structure-functional network coding features of the patient object to be evaluated and the pre-treatment brain structure-functional network features to obtain the brain state query response coding features of the patient object to be evaluated. It should be understood that since the set of the brain structure-functional network coding vectors of the patient object to be evaluated and the pre-treatment brain structure-functional network feature vectors respectively contain the brain structure-functional network state features of the patient object to be evaluated and the brain network state database marked as effective for tDcs intervention, when predicting and evaluating whether the classical tDcs intervention for this patient object to be evaluated is effective, it is necessary to query and match the brain network state features of this patient object with each known intervention-effective brain network state in the database to match the similar brain network state of the personalized patient object for subsequent classification and effectiveness evaluation tasks of the classical tDcs intervention paradigm. However, since the traditional feature matching query method is based on the overall database information for screening, the query response speed and accuracy are relatively low. Therefore, in the technical solution of this application, a fast query dynamic response process is further performed on the set of the brain structure-functional network coding features of the patient object to be evaluated and the pre-treatment brain structure-functional network features to obtain the brain state query response coding features of the patient object to be evaluated. In particular, the process of the fast query dynamic response process realizes the process from rough positioning to fine-grained matching and then to deep semantic coding by integrating technologies such as cross-entropy calculation, local context awareness, linear transformation, and heterogeneous transformer structure, aiming to optimize the speed and accuracy of the brain state query response of the patient object to be evaluated.

[0034] In one embodiment, as Figure 4 shown, the patient object brain state query response module 50 includes: a patient object fast query positioning factor calculation unit 51, configured to calculate the patient object fast query positioning factors of the brain structure-functional network coding vector of the patient object to be evaluated relative to each pre-treatment brain structure-functional network feature vector in the set of the pre-treatment brain structure-functional network feature vectors to obtain a set of patient object fast query positioning factors; a brain state fine-grained matching search window determination unit 52, configured to determine a brain state fine-grained matching search window of the patient object based on the set of the patient object fast query positioning factors; and a fine-grained query response coding unit 53, configured to perform a fine-grained query response coding on the set of the brain structure-functional network coding vector of the patient object to be evaluated and the pre-treatment brain structure-functional network feature vectors based on the brain state fine-grained matching search window of the patient object to obtain the brain state query response coding features of the patient object to be evaluated.

[0035] In one embodiment, the patient object fast query and positioning factor calculation unit 51 is configured to: calculate the cross entropy of the brain structure-functional network coding vector of the patient object to be evaluated with respect to each of the pre-treatment brain structure-functional network feature vectors in the set of pre-treatment brain structure-functional network feature vectors to obtain the set of patient object fast query and positioning factors. Specifically, this process can be expressed by the formula:

[0036]

[0037]

[0038]

[0039] wherein, is the set of pre-treatment brain structure-functional network feature vectors, are respectively the 1st, 2nd, th, and th pre-treatment brain structure-functional network feature vectors in the set of pre-treatment brain structure-functional network feature vectors, are the eigenvalues at each position in the brain structure-functional network coding vector of the patient object to be evaluated, is the number of eigenvalues, is the th eigenvalue at each position in the pre-treatment brain structure-functional network feature vector, is the cross entropy between the brain structure-functional network coding vector of the patient object to be evaluated and the th pre-treatment brain structure-functional network feature vector, that is, the patient object fast query and positioning factor, is the set of patient object fast query and positioning factors, are respectively the 1st, 2nd, th, and th patient object fast query and positioning factors.

[0040] It should be understood that by calculating the cross-entropy, the semantic differences between the brain structure-functional network coding vectors of the patients to be evaluated and each known effective case in the database can be quantified. As a metric for measuring the difference between two probability distributions, the cross-entropy is used here to evaluate the similarity between two brain network states. A smaller cross-entropy value means the two are closer, while a larger value indicates a greater difference. Therefore, this method can quickly narrow down the search scope, screen out the most likely relevant features, and greatly improve the speed and efficiency of the query. For each patient to be evaluated, finding the effective treatment case with the most similar brain network state is the key to achieving personalized medicine. The cross-entropy calculation helps the system identify those historical cases that are most similar to the new patient in terms of structure and function, thus providing a scientific basis for formulating personalized tDCS intervention plans. In addition, as a metric, the cross-entropy can exclude those obviously mismatched features at an early stage and reduce the risk of misjudgment. This means that only when the patient to be evaluated shows a high degree of consistency with the historical case in multiple dimensions will it be considered a potential effective treatment target. This not only improves the reliability of the prediction results but also saves time and resources for subsequent detailed analysis. Based on the set of fast query location factors calculated from the cross-entropy, the system can dynamically adjust its query strategy and prioritize those features with higher similarity for further analysis.

[0041] In one embodiment, the brain state fine-grained matching search window determination unit 52 is configured to: use the pre-treatment brain structure-functional network feature vector corresponding to the minimum value in the set of fast query location factors of the patient object as the location matching pre-treatment brain structure-functional network feature vector; based on the location matching pre-treatment brain structure-functional network feature vector, determine the brain state fine-grained matching search window of the patient object, where the vector at the center position of the brain state fine-grained matching search window of the patient object is the location matching pre-treatment brain structure-functional network feature vector, and each pre-treatment brain structure-functional network feature vector in the brain state fine-grained matching search window of the patient object is defined as a fine-grained query pre-treatment brain structure-functional network feature vector to obtain a set of fine-grained query pre-treatment brain structure-functional network feature vectors. Specifically, this process can be expressed by the formula:

[0042]

[0043]

[0044]

[0045] Among them, is to return the corresponding one of the minimum values , To locate and match the eigenvectors of the pre-treatment brain structure-functional network, represents calculating the square of the one-norm of a vector, represents calculating the value of the logarithmic function with base 2, represents floor processing, is the matching search window factor, ( ) is the starting point of the fine-grained matching search window for the brain state of the patient object, ( ) is the ending point of the fine-grained matching search window for the brain state of the patient object, is the set of fine-grained query pre-treatment brain structure-functional network eigenvectors, are respectively each fine-grained query pre-treatment brain structure-functional network eigenvector in the set of fine-grained query pre-treatment brain structure-functional network eigenvectors.

[0046] It should be understood that by selecting the pre-treatment brain structure-functional network eigenvector corresponding to the minimum value in the set of fast query location factors as the location matching eigenvector, the system can accurately locate the historical case most similar to the patient to be evaluated. This precise matching is the basis for subsequent more detailed analysis, ensuring that the results of the preliminary screening have a high degree of credibility. This step effectively narrows the search scope, reduces unnecessary consumption of computing resources, and at the same time improves the speed and accuracy of matching. Next, the system constructs a fine-grained matching search window based on the location-matched pre-treatment brain structure-functional network eigenvector. This window is centered around the location-matched eigenvector and defines a set of fine-grained query pre-treatment brain structure-functional network eigenvectors around it. These eigenvectors represent a series of historical cases that are closest to the patient to be evaluated, forming a local and refined feature space. In this way, the system can perform in-depth analysis in a smaller and more relevant dataset, thereby capturing more subtle brain network differences and providing more accurate matching results. The introduction of the fine-grained matching search window enables the system to more carefully compare the similarities and differences between the patient to be evaluated and historical cases. By further analyzing within this local feature space, the system can identify key factors that may affect the tDCS intervention effect, such as changes in the connection strength or activity patterns of specific brain regions. This refined analysis helps improve the ability of personalized prediction, enabling doctors to formulate more precise treatment plans according to the unique situation of each patient. In addition, each eigenvector in the fine-grained matching search window is defined as a set of fine-grained query pre-treatment brain structure-functional network eigenvectors, which not only enhances the intelligence level of the system but also supports more efficient decision-making assistance.

[0047] In one embodiment, the fine-grained query response encoding unit 53 is configured to: perform a linear transformation on the brain structure-functional network encoding vector of the patient object to be evaluated to obtain a brain structure-functional network query vector and a brain structure-functional network value vector of the patient object to be evaluated; use each fine-grained pre-treatment brain structure-functional network feature vector in the set of fine-grained pre-treatment brain structure-functional network feature vectors as a key vector, and input the brain structure-functional network query vector, the brain structure-functional network value vector, and the key vector of the patient object to be evaluated into the fine-grained query encoding module based on the heterogeneous transformer structure to obtain a brain state query response encoding vector of the patient object to be evaluated as the brain state query response encoding feature of the patient object to be evaluated. Specifically, this process can be represented by the formula:

[0048]

[0049]

[0050]

[0051] Wherein, is the brain structure-functional network encoding vector of the patient object to be evaluated, and are the query weight matrix and the query bias vector respectively, and are the value weight matrix and the value bias vector respectively, and are the brain structure-functional network query vector and the brain structure-functional network value vector of the patient object to be evaluated respectively, is the -th fine-grained pre-treatment brain structure-functional network feature vector in the set of fine-grained pre-treatment brain structure-functional network feature vectors, is the length of the -th fine-grained pre-treatment brain structure-functional network feature vector, is matrix multiplication, is the normalized exponential function, is the brain state query response encoding vector of the patient object to be evaluated.

[0052] It should be understood that, first, by performing a linear transformation on the brain structure-functional network encoding vector of the patient object to be evaluated, the system generates a query vector and a value vector. This transformation not only preserves the spatial distribution characteristics of the original features but also enhances the feature representation ability through a mapping operation. The query vector is used to find the most similar historical cases in the database, while the value vector carries more detailed information, which helps with subsequent in-depth analysis. Next, the system inputs these query vectors, value vectors, and each fine-grained query pre-treatment brain structure-functional network feature vector (as key vectors) extracted from the fine-grained matching search window into the fine-grained query encoding module based on a heterogeneous transformer structure. The heterogeneous transformer structure allows the system to process different types of data within the same framework, thus more effectively capturing complex relationships. In this way, the system can efficiently complete the matching task for large-scale data in a short time while ensuring the accuracy of the results. The fine-grained query encoding module uses the attention mechanism to focus on the most relevant historical cases, thereby achieving a deep semantic understanding of the patient's brain state. During this process, the interaction between the query vector and the key vector helps the system identify which historical cases are most similar to the current patient; while the value vector provides additional information, enabling the system to further refine the matching results. This method not only improves the quality of the matching but also reveals the potential patterns hidden behind the data, providing a more solid foundation for personalized prediction. The finally output brain state query response encoding vector of the patient object to be evaluated, as the brain state query response encoding feature, represents the comprehensive description after multiple layers of analysis. This feature integrates information from different dimensions, including structural connectivity and functional interaction, and can more comprehensively reflect the unique brain state of the patient. Doctors can evaluate the possibility of tDCS intervention and its expected effects based on these detailed features, thereby formulating a more precise treatment plan. For example, if the brain state of a new patient is highly consistent with known effective cases, it indicates that the patient may also benefit from a similar tDCS intervention; conversely, if significant differences are found, it may be necessary to adjust the intervention strategy or explore other treatment methods.

[0053] In summary, in the process of fast query dynamic response processing, by introducing a fast positioning mechanism, that is, using cross-entropy calculation to evaluate the semantic distance between the brain structure-functional network coding vector of the patient object to be evaluated and each pre-treatment brain structure-functional network feature vector, the semantic difference between the two can be quantified, so as to quickly narrow down the search scope to the most likely relevant area and improve the response speed of query matching. Furthermore, a best-matched pre-treatment brain structure-functional network feature vector is found through preliminary screening. This method reduces unnecessary consumption of computing resources and accelerates the preliminary screening process. Secondly, in order to further improve the matching query accuracy, the system delimits a fine-grained matching search window around the best-matched pre-treatment brain structure-functional network feature vector. This window focuses on the local area around the pre-treatment brain structure-functional network features with known effective interventions determined by the preliminary screening. This strategy enhances the model's deep correlation semantic understanding ability between the brain structure-functional network state of the patient object to be evaluated and each known effective pre-treatment brain structure-functional network state, enabling the system to capture more subtle differences in pre-treatment brain states, so as to provide a more accurate query response result and feature representation of the patient object's brain state, and provide a basis for the classification judgment of whether the subsequent classical tDcs intervention paradigm is effective for the patient object to be evaluated.

[0054] Exemplarily, in the intervention effectiveness detection module 60, based on the brain state query response coding features of the patient object to be evaluated, it is determined whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated. In one embodiment, the intervention effectiveness detection module is configured to: input the brain state query response coding vector of the patient object to be evaluated into an evaluation module based on a classifier to obtain an evaluation result, and the evaluation result is used to represent whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated. That is to say, the characterization information of the brain state query response of the patient object to be evaluated is used for classification processing to identify the possibility of positive brain responses of new individuals, so as to predict whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated. In this way, through a more intelligent query response method, a similar brain network state of an individual patient can be matched from a database of known effective tDCS intervention brain network states, which helps to screen out the brain network features after treatment using similar pre-treatment brain network states, so as to assist in predicting the potential effectiveness of tDCS intervention for different individual patients, especially for those who may not respond to the standard treatment plan.

[0055] In a specific embodiment of the present application, inputting the encoding vector of the brain state query response of the patient object to be evaluated into the classifier-based evaluation module to obtain an evaluation result includes: performing a fully connected encoding on the encoding vector of the brain state query response of the patient object to be evaluated using the fully connected layer of the classifier-based evaluation module to obtain a fully connected encoded encoding vector of the brain state query response of the patient object to be evaluated. The role of the fully connected layer is to convert the input high-dimensional feature vector into a low-dimensional representation while capturing the complex relationships between features. This process enhances the learning ability of the model, enabling it to better understand the internal patterns of the input data. Then, passing the fully connected encoded encoding vector of the brain state query response of the patient object to be evaluated through the Softmax classification function of the classifier-based evaluation module to obtain a probability distribution belonging to different categories. The Softmax function can convert the raw scores on multiple categories into probability values, and the sum of these probability values is equal to 1, facilitating the interpretation of the relative likelihood of each category. In the present application, two probability values will be obtained: the first probability belonging to the presence of a positive treatment response and the second probability belonging to the absence of a positive treatment response. These two probability values reflect the probabilities of different outcomes that the patient to be evaluated may have after receiving tDCS intervention. Finally, based on the comparison between the first probability and the second probability, the final classification result is determined. For example, if the first probability is higher than the second probability, it is determined that the classical tDcs intervention paradigm is effective for the patient object to be evaluated; conversely, if the second probability is higher, it indicates that the classical tDcs intervention paradigm is ineffective for the patient object to be evaluated, and other treatment options may need to be considered.

[0056] Preferably, obtaining an evaluation result by passing the encoding vector of the brain state query response of the patient object to be evaluated through the classifier-based evaluation module includes:

[0057] Calculating the distance between each pair of eigenvalues of the encoding vector of the brain state query response of the patient object to be evaluated, such as the L2 distance, and taking the square root of the distance to obtain a matrix representing the encoding distance of the brain state query response of the patient object to be evaluated, that is

[0058]

[0059] where represents the encoding vector of the brain state query response of the patient object to be evaluated, and represent the and eigenvalues of the encoding vector of the brain state query response of the patient object to be evaluated, represents the and distance between the Denote the eigenvalue at the position of in the brain state query response coding distance representation matrix of the patient object to be evaluated;

[0060] Obtain the self - correlation matrix of the brain state query response coding of the patient object to be evaluated for the brain state query response coding vector of the patient object to be evaluated, that is , where represents the brain state query response coding vector of the patient object to be evaluated, represents the transpose of the vector, represents matrix multiplication, represents the self - correlation matrix of the brain state query response coding of the patient object to be evaluated;

[0061] Multiply the brain state query response coding vector of the patient object to be evaluated with the brain state query response coding distance representation matrix to obtain the first - level mapping vector of the brain state query response coding of the patient object to be evaluated, that is , where represents the brain state query response coding vector of the patient object to be evaluated, represents the brain state query response coding distance representation matrix of the patient object to be evaluated, represents matrix multiplication, represents the first - level mapping vector of the brain state query response coding of the patient object to be evaluated;

[0062] Multiply the first - level mapping vector of the brain state query response coding of the patient object to be evaluated with the matrix product of the brain state query response coding distance representation matrix and the self - correlation matrix of the brain state query response coding of the patient object to be evaluated to obtain the multi - level mapping vector of the brain state query response coding of the patient object to be evaluated , where represents the first - level mapping vector of the brain state query response coding of the patient object to be evaluated, represents the brain state query response coding distance representation matrix of the patient object to be evaluated, represents matrix multiplication, represents the self - correlation matrix of the brain state query response coding of the patient object to be evaluated, represents the multi - level mapping vector of the brain state query response coding of the patient object to be evaluated;

[0063] Dot - multiply the multi - level mapping vector of the brain state query response coding of the patient object to be evaluated with the associated eigenvector of the brain state query response coding of the patient object to be evaluated composed of the eigenvalues of the self - correlation matrix of the brain state query response coding of the patient object to be evaluated (interpolate or fill with zeros if the number of eigenvalues is insufficient) to obtain the optimized brain state query response coding vector of the patient object to be evaluated;

[0064] The optimized brain state query response encoding vector of the patient subject to be evaluated is passed through a classifier-based evaluation module to obtain an evaluation result.

[0065] Considering that the brain structure-function network coding vector of the patient to be evaluated represents the semantic coding splicing features of the brain structure network and the brain function network of the patient to be evaluated, and the set of the pre-treatment brain structure-function network feature vectors represents the semantic coding splicing features of the pre-treatment brain structure network and the pre-treatment brain function network of multiple patient objects marked as valid, when performing dynamic semantic query encoding based on a fast positioning mechanism, differences in source data formats and additional interference introduced during the data acquisition process will cause positioning accuracy to be imbalanced, resulting in the lack of decision-making of the query coding feature instance of the brain state query response coding vector of the patient to be evaluated obtained by the dynamic semantic query encoding, thereby affecting the accuracy of the evaluation result obtained by the classifier-based evaluation module.

[0066] Therefore, the self-correlation of the complete similarity instantiation of the brain state query response encoding vector of the patient to be evaluated is instantiated by a linear target mapping representation based on the similarity distance representation matrix of the brain state query response encoding vector of the patient to be evaluated, and the negative impact factor of the association mismatch is compensated by the association fusion kernel bias to improve the degree of instance determination of the eigenvalue of the brain state query response encoding vector of the patient to be evaluated under the similarity restriction, that is, the degree of significance of the eigenvalue as an instance for classification regression determination, and improve the accuracy of the evaluation result obtained by the classifier-based evaluation module of the brain state query response encoding vector of the patient to be evaluated. In this way, similar brain network states of personalized patients can be matched from the known effective tDCS intervention brain network state database through a more intelligent query response method, which helps to assist in predicting the potential effectiveness of tDCS intervention for different individual patients, especially for those who may not respond to standard treatment plans.

[0067] In summary, the transcranial direct current stimulation intervention system according to the embodiments of the present application is elucidated. It uses data processing and image analysis algorithms based on machine learning and artificial intelligence technologies to analyze the pre- and post-treatment brain structure networks and brain function networks labeled as effective, so as to capture the pre-treatment brain structure-functional network characteristics, which are used to reflect the anatomical connectivity and functional interaction between different brain regions where these tDCS interventions are effective, thereby forming a database of effective brain network states. When actually identifying the possibility of positive brain responses and detecting potential effectiveness for a new individual, the characteristics of the new individual's brain structure network and brain function network are extracted and spliced to reflect the characteristics of the new individual's brain network state, and the characteristics of the new individual's brain network state are used as an index to query for similar characteristic information from the database of effective brain network states of the intervention, so as to identify the possibility of positive brain responses of the new individual, and thus predict whether the classical tDcs intervention paradigm is effective for the patient to be evaluated. In this way, it is possible to match the similar brain network states of personalized patients from the known effective tDCS intervention brain network state database through a more intelligent query response method, which helps to assist in predicting the potential effectiveness of tDCS interventions for different individual patients, especially for those individuals who may not respond to the standard treatment plan.

[0068] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, article or device. Without more limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including that element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present invention is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, the features described with reference to certain examples may be combined in other examples.

[0069] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a computer software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0070] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the present invention, and all of them belong to the protection scope of the present invention.

Claims

1. A transcranial direct current stimulation intervention system, characterized in that, Including: An effective intervention brain network data acquisition module for acquiring datasets of {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} labeled as effective; An effective intervention brain network analysis module for performing feature extraction and aggregation analysis on each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the datasets of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} labeled as effective to obtain a set of pre-treatment brain structural-functional network features; A patient object brain network data acquisition module for acquiring the brain structural network and brain functional network of a patient object to be evaluated; A patient object brain network analysis module for performing feature extraction and feature splicing on the brain structural network and brain functional network of the patient object to be evaluated to obtain the brain structural-functional network coding features of the patient object to be evaluated; A patient object brain state query response module for performing fast query dynamic response processing on the brain structural-functional network coding features of the patient object to be evaluated and the set of pre-treatment brain structural-functional network features to obtain the brain state query response coding features of the patient object to be evaluated; An intervention effectiveness detection module for determining whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated based on the brain state query response coding features of the patient object to be evaluated; Wherein, the patient object brain network analysis module is used to: after using a brain network feature extractor based on a dilated convolutional neural network model to perform feature extraction on the brain structural network and brain functional network of the patient object to be evaluated respectively, then splice the obtained brain structural network feature vector of the patient object to be evaluated and the brain functional network feature vector of the patient object to be evaluated to obtain a brain structural-functional network coding vector of the patient object to be evaluated as the brain structural-functional network coding features of the patient object to be evaluated; Wherein, the patient object brain state query response module includes: A patient object fast query positioning factor calculation unit for calculating the patient object fast query positioning factors of the brain structural-functional network coding vector of the patient object to be evaluated relative to each of the pre-treatment brain structural-functional network feature vectors in the set of pre-treatment brain structural-functional network features to obtain a set of patient object fast query positioning factors; A brain state fine-grained matching search window determination unit for determining a brain state fine-grained matching search window of the patient object based on the set of patient object fast query positioning factors; A fine-grained query response coding unit for performing fine-grained query response coding on the brain structural-functional network coding vector of the patient object to be evaluated and the set of pre-treatment brain structural-functional network features based on the brain state fine-grained matching search window of the patient object to obtain the brain state query response coding features of the patient object to be evaluated.

2. The transcranial direct current stimulation intervention system according to claim 1, wherein The effective intervention brain network analysis module includes: The pre-treatment brain feature extraction unit is used to extract features from each of the {pre-treatment brain structural network, pre-treatment brain functional network, post-treatment brain structural network, post-treatment brain functional network} in the dataset marked as valid to obtain a set of pre-treatment brain structural network feature vectors and a set of pre-treatment brain functional network feature vectors; The pre-treatment brain feature splicing unit is used to splice the corresponding pre-treatment brain structural network feature vectors and pre-treatment brain functional network feature vectors in the set of pre-treatment brain structural network feature vectors and the set of pre-treatment brain functional network feature vectors to obtain a set of pre-treatment brain structural-functional network feature vectors as the set of pre-treatment brain structural-functional network features.

3. The transcranial direct current stimulation intervention system according to claim 2, characterized in that, The pre-treatment brain feature extraction unit is used to: extract each pre-treatment brain structural network and each pre-treatment brain functional network from the dataset, and respectively perform feature extraction on the each pre-treatment brain structural network and the each pre-treatment brain functional network through a brain network feature extractor based on a dilated convolutional neural network model to obtain the set of pre-treatment brain structural network feature vectors and the set of pre-treatment brain functional network feature vectors.

4. The transcranial direct current stimulation intervention system according to claim 3, wherein, The patient object rapid query positioning factor calculation unit is used to: calculate the cross entropy of the brain structural-functional network coding vector of the patient object to be evaluated with respect to each pre-treatment brain structural-functional network feature vector in the set of pre-treatment brain structural-functional network feature vectors to obtain the set of patient object rapid query positioning factors.

5. The transcranial direct current stimulation intervention system according to claim 4, characterized in that, The brain state fine-grained matching search window determination unit is used to: Use the pre-treatment brain structural-functional network feature vector corresponding to the minimum value in the set of patient object rapid query positioning factors as the positioning matching pre-treatment brain structural-functional network feature vector; Based on the positioning matching pre-treatment brain structural-functional network feature vector, determine the patient object brain state fine-grained matching search window, where the vector at the central position of the patient object brain state fine-grained matching search window is the positioning matching pre-treatment brain structural-functional network feature vector, and each pre-treatment brain structural-functional network feature vector in the patient object brain state fine-grained matching search window is defined as a fine-grained query pre-treatment brain structural-functional network feature vector to obtain a set of fine-grained query pre-treatment brain structural-functional network feature vectors.

6. The transcranial direct current stimulation intervention system according to claim 5, wherein The fine-grained query response encoding unit is used to: Perform a linear transformation on the brain structural-functional network coding vector of the patient object to be evaluated to obtain a brain structural-functional network query vector and a brain structural-functional network value vector of the patient object to be evaluated; Using each of the fine-grained query pre-treatment brain structure-functional network feature vectors in the set of fine-grained query pre-treatment brain structure-functional network feature vectors as key vectors, input the brain structure-functional network query vector of the patient object to be evaluated, the brain structure-functional network value vector of the patient object to be evaluated, and the key vectors into the fine-grained query encoding module based on the heterogeneous transformer structure to obtain the brain state query response encoding vector of the patient object to be evaluated as the brain state query response encoding feature of the patient object to be evaluated.

7. The transcranial direct current stimulation intervention system according to claim 6, characterized in that, The intervention effectiveness detection module is configured to: input the brain state query response encoding vector of the patient object to be evaluated into the evaluation module based on a classifier to obtain an evaluation result, and the evaluation result is used to indicate whether the classical tDcs intervention paradigm is effective for the patient object to be evaluated.

Images