A method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets
By using a brain network virtual modulation method, the area under the improvement rate curve of candidate targets is calculated and normalized, which solves the problem of the lack of unified evaluation of candidate targets in the existing technology, realizes the quantitative evaluation and ranking of candidate targets, and improves the objectivity and repeatability of the evaluation.
Patent Information
- Application Number
- CN202611203624.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-10
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies lack a unified evaluation standard after identifying transcranial electrical stimulation candidate targets, making it difficult to conduct quantitative comparisons and rankings, which leads to difficulties in determining the relative priority of multiple candidate targets.
A brain network-based virtual modulation method was adopted. By calculating the changes in the total network deviation of candidate target-related connections before and after virtual modulation, an improvement rate curve was constructed using linear interpolation, and the candidates were ranked by normalized AUC scores to form a unified priority result for candidate targets.
It enables quantitative evaluation and ranking of multiple candidate targets without the need for real stimulation, improves the objectivity and reproducibility of candidate target evaluation, provides a unified evaluation standard, and is applicable to different disease types and brain function data modalities.
Smart Images

Figure CN123120195A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of brain functional data processing, brain functional network analysis, and non-invasive neuromodulation-assisted decision-making technology, specifically to a method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets based on virtual modulation of brain networks. Background Technology
[0002] Non-invasive neuromodulation techniques such as transcranial electrical stimulation (tES) intervene in neural activity and brain function by applying weak currents to specific locations on the scalp. They have significant application value in areas such as cognitive function improvement, intervention for mental disorders, and brain function rehabilitation. In the design of a tES protocol, the choice of stimulation location directly affects the spatial directionality of the stimulation effect, the interpretability of the protocol, and the design of subsequent parameters, making it a crucial preliminary step in the electrical stimulation intervention process.
[0003] Currently, transcranial electrical stimulation candidate targets are usually determined based on empirical brain region localization, previous literature reports, hypothesis of symptom-related brain regions, or abnormal results obtained from EEG and brain imaging analysis.
[0004] CN118718251B discloses an intervention optimization method for transcranial electrical stimulation (TCS). This method acquires multimodal magnetic resonance imaging (MRI) data of the target subject, fuses functional and structural connectivity information between brain regions to determine the target brain regions, and further combines individual electrical conduction models and brain electric field distribution calculations to determine the stimulation matrix. This approach is based on multimodal brain imaging and individual electric field modeling, focusing on the joint optimization of the target brain regions and the stimulation matrix.
[0005] CN120285454A discloses a transcranial stimulation method for the human brain. This method selects core brain regions by calculating the degree centrality and betweenness centrality of brain network nodes, locates abnormal core brain regions and related brain regions using a pre-set anomaly detection model, and achieves closed-loop control through real-time state feedback during subsequent stimulation. This scheme is based on brain network topological features and anomaly detection results, focusing on abnormal brain region identification and stimulation strategy generation.
[0006] The methods described above can all provide a basis for identifying relevant brain regions or candidate targets. However, due to differences in analytical dimensions and evaluation indicators among different methods, the results obtained usually lack uniform comparability. When multiple candidate targets with literature support, abnormal characteristics, or neurophysiological significance exist for the same research subject, existing methods still struggle to quantitatively compare and rank the candidate targets based on unified evaluation criteria after the candidate target set is formed.
[0007] Therefore, there is an urgent need for a method for quantitative assessment and ranking of candidate targets before actual transcranial electrical stimulation (TCS) procedures. This method should be able to uniformly model, quantitatively evaluate, and rank multiple candidate targets based on existing brain function data, generating repeatable priority results for candidate targets before stimulation, thus providing an objective computational reference for subsequent TCS target selection. Summary of the Invention
[0008] To address the problem in existing technologies that, after initial identification of transcranial electrical stimulation (TCS) candidate targets, lack a pre-stimulation quantitative comparison mechanism based on a unified evaluation standard, making it difficult to establish a relative priority ranking of multiple candidate targets, this invention provides a method for quantitative evaluation and ranking of TCS candidate targets based on virtual modulation of brain networks. This method uses a unified virtual modulation framework to evaluate the changes in the overall network deviation during the adjustment of candidate target-related connections to a reference state, thereby generating comparable quantitative results between different candidate targets under the same evaluation standard.
[0009] A method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets includes the following steps: Step 1: Obtain the brain network to be evaluated and the reference brain network, perform node and matrix consistency verification on both, and perform standardization processing according to the connection properties of the brain network; wherein, the brain network to be evaluated and the reference brain network are both N×N connection matrices, the rows and columns of the matrix correspond to network nodes, and the matrix elements represent the functional connection strength between the corresponding nodes. Step 2: Obtain pre-determined candidate target information, map each candidate target to a candidate node in the brain network A to be evaluated and the reference brain network R after processing in Step 1, and form a candidate target set; wherein, each candidate target corresponds to a candidate node, and the candidate target set includes all or part of the network nodes to be quantified and ranked. Step 3: Based on the brain network A to be evaluated and the reference brain network R processed in Step 1, calculate the initial full network deviation of the brain network to be evaluated relative to the reference brain network, and calculate the initial abnormality of the relevant connections of each candidate node in the candidate target set relative to the reference brain network R before virtual adjustment. Step 4: Set multiple virtual adjustment intensity values within a preset virtual adjustment intensity range to form a virtual adjustment intensity sequence; Step 5: Traverse each candidate node in the candidate target set and each virtual modulation intensity in the virtual modulation intensity sequence, perform linear interpolation virtual modulation on the relevant connections between the candidate node and other nodes, and construct the corresponding virtual modulation backbrain network; Step 6: Calculate the deviation of each virtual-modulated brain network relative to the reference brain network, and calculate the improvement rate of each candidate node under the corresponding virtual modulation intensity based on the initial full network deviation and the virtual-modulated full network deviation; form the improvement rate curve corresponding to the candidate node based on the improvement rate of the same candidate node under different virtual modulation intensities. Step 7: Calculate the area under the improvement rate curve corresponding to each candidate node, and normalize the area under the curve using the initial outlier of the candidate node to obtain the normalized AUC score of each candidate node. Step 8: Sort all candidate nodes in descending order according to the normalized AUC score, and output the quantitative evaluation and relative priority ranking results of transcranial electrical stimulation candidate targets.
[0010] Furthermore, the brain network is an undirected weighted brain functional connectivity network constructed based on EEG, magnetoencephalography, functional magnetic resonance imaging, or other brain functional data. Depending on the data source and construction scale of the brain network, each network node corresponds to a signal acquisition location, a preset brain region, a source spatial location, or other preset brain functional unit. The functional connectivity strength is calculated using correlation, coherence, phase synchronization, or other functional connectivity indicators capable of forming an undirected weighted connectivity matrix. This invention does not limit the specific brain network construction method or functional connectivity calculation method.
[0011] Furthermore, step one specifically includes: The consistency check between the brain network A to be evaluated and the reference brain network R includes: Verify whether the matrix dimensions, number of nodes, node labels, node arrangement order, and connection strength expression methods of the brain network to be evaluated and the reference brain network are consistent; when the node sets are consistent but the node arrangement order is inconsistent, the rows and columns of the connection matrix are rearranged synchronously according to the node labels; when the node sets are inconsistent, the two are transformed into a unified node space according to the preset node mapping relationship; if node mapping cannot be completed, or if a connection matrix with consistent node correspondence cannot be formed after node mapping, the evaluation process is terminated. The connection matrix that has completed the consistency check is standardized, including setting the main diagonal elements of the brain network to be evaluated A and the reference brain network R to zero to complete the symmetry process.
[0012] Furthermore, the candidate target set in step two is predetermined by one of the following methods: All nodes in the brain network were used as candidate targets; Candidate targets are selected based on prior research, prior knowledge, specific diseases, cognitive tasks, or preliminary brain function analysis results; or Based on the initial difference analysis results between the brain network to be evaluated and the reference brain network, abnormal nodes with functional connectivity strength deviations exceeding a preset threshold were selected as candidate targets.
[0013] Furthermore, in step three, the initial full-network deviation of the brain network A to be evaluated relative to the reference brain network R... And the virtual modulation of the hindbrain network in step six Total deviation of the reference brain network R All values were calculated using the Frobenius norm.
[0014] Furthermore, the initial anomaly quantity of candidate node i in step three... , is the sum of the absolute differences in the connections between the candidate node and other nodes in the brain network A to be evaluated and the reference brain network R, where the calculation process excludes the candidate node's own connections: In this process, j iterates through all network nodes except for candidate node i. This represents the connection strength between candidate node i and candidate node j in the brain network A to be evaluated. This represents the connection strength between candidate node i and candidate node j in the reference brain network R.
[0015] Furthermore, in step four, the virtual adjustment intensity parameter is a dimensionless calculation parameter, and its theoretical value range is [0,1]; the virtual adjustment intensity sequence is constructed in the following way: Based on actual assessment needs, the actual virtual adjustment intensity range is selected within the range of [0,1]. Multiple discrete value points are set at equal or non-equal intervals within the interval to form a virtual adjustment intensity sequence; When the virtual adjustment strength parameter is 0, the connection strength of the candidate node remains unchanged from the initial connection strength in the brain network A to be evaluated. When the virtual adjustment strength parameter is set to 1, the relevant connection strength of the candidate node is set to the strength of the corresponding connection in the reference brain network R.
[0016] Furthermore, step five specifically includes: Local adjustment: at each virtual adjustment intensity In this case, only the relevant connections between candidate node i and other nodes are updated, while the connections between other nodes that are not involved in candidate node i remain at the original connection values in the brain network A to be evaluated. For any node The virtual adjusted connection strength is calculated according to the following formula: [ in,[ Indicates the virtual adjustment intensity of candidate node i. The virtual adjusted connection strength between node j and node j; This indicates the original connection strength of the corresponding connection in the brain network to be evaluated; This indicates the connection strength of the corresponding connection in the reference brain network; To maintain the symmetry of the connected matrix after virtual adjustment, the matrix elements at symmetrical positions are updated synchronously, and the elements on the main diagonal are set to zero.
[0017] Wherein, the virtual adjustment intensity parameter It only represents the calculated proportion by which the brain network connection strength approaches the reference state. It is a mathematical calculation variable and does not correspond to the current intensity, stimulation frequency, stimulation duration or other physical stimulation parameters in actual transcranial electrical stimulation.
[0018] Furthermore, the improvement rate in step six uses the initial full network deviation. Deviation of the entire network after adjustment of each virtual adjustment intensity calculate: in This is a preset, extremely small positive number, used to avoid zero denominators and improve the stability of numerical calculations.
[0019] Furthermore, in step seven, the area under the improvement rate curve is calculated using a numerical integration method; The normalized AUC score is obtained by dividing the area under the curve by the sum of the initial outlier and the smallest positive number of the corresponding candidate node, and is used to characterize the overall improvement of the entire network corresponding to a unit initial outlier.
[0020] Furthermore, the output results in step eight include one or more of the following: candidate target labels, normalized AUC scores, and relative priority rankings, as well as the optional output of improvement rate curve data corresponding to each candidate target.
[0021] The beneficial effects of this invention are as follows: The method described in this invention addresses the need for candidate target evaluation before actual transcranial electrical stimulation (TCS) is performed. Without the need for actual stimulation, it uses existing brain functional connectivity networks to perform unified virtual adjustment, quantitative evaluation, and ranking output on multiple pre-determined candidate targets, forming a relative priority result for the candidate targets. This can provide a quantitative calculation reference for subsequent TCS target selection. The method uses a unified virtual adjustment framework to evaluate the changes in the deviation of the entire network during the adjustment of candidate target-related connections to the reference state, thereby forming comparable quantitative results between different candidate targets under the same evaluation criteria, which helps to improve the objectivity and repeatability of the candidate target evaluation and ranking process before transcranial electrical stimulation. The method employs linear interpolation to virtually adjust the connections related to candidate targets, characterizing the computational process of gradually adjusting these connections from the state to the reference state. This method has clear calculation rules, well-defined parameter meanings, and low implementation complexity, making it suitable for rapid quantitative assessment before transcranial electrical stimulation. The method uses the area under the improvement rate curve to characterize the comprehensive evaluation results of candidate targets within a preset virtual adjustment intensity range, and further normalizes the area under the curve by combining the initial anomaly of the candidate target's related connections, thereby reducing the impact of differences in the initial anomaly scale of different candidate targets on the ranking results and improving the comparability of the quantitative evaluation results of candidate targets. The method is not limited to specific disease types, experimental tasks, brain functional data modalities, or brain network construction algorithms. When the brain network to be evaluated and the reference brain network use the same brain functional data modalities and functional connectivity indicators, and can be represented as an undirected weighted brain functional connectivity matrix with consistent node correspondence, the candidate target quantitative evaluation and ranking process can be executed, thus having good versatility and scalability. Attached Figure Description
[0022] Figure 1 This is a flowchart of the transcranial electrical stimulation candidate target quantification evaluation and ranking method described in this invention; Figure 2 This is a schematic diagram of the process of the transcranial electrical stimulation candidate target quantification evaluation and ranking method. Detailed Implementation
[0023] To make the objectives, technical solutions, beneficial effects, and significant advancements of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings provided in the examples of the present invention. Obviously, all the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0024] In the description of this application, unless otherwise expressly specified and limited, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance; the term "multiple" refers to two or more; unless otherwise specified or explained, the terms "connected," "fixed," etc., should be interpreted broadly. For example, "connected" can be a fixed connection, a detachable connection, an integral connection, or an electrical connection; "connected" can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.
[0025] like Figure 1 , Figure 2 As shown, a method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets specifically includes the following steps: Step 1: Obtain the brain network to be evaluated and the reference brain network, and perform consistency verification and standardization on the two.
[0026] S11: Obtain the brain network to be evaluated and the reference brain network.
[0027] This step is performed after the brain functional connectivity network is constructed to obtain the brain network to be evaluated, A, and the reference brain network, R.
[0028] Both the brain network A to be evaluated and the reference brain network R can be constructed based on electroencephalography (EEG), magnetoencephalography (MEG), functional magnetic resonance imaging (fMRI), or other data capable of characterizing brain functional activity. Depending on the source of the brain functional data and the scale of brain network construction, the network nodes of the brain network A to be evaluated and the reference brain network R can correspond to signal acquisition locations, preset brain regions, source spatial locations, or other preset brain functional units. This invention does not limit the specific acquisition method or preprocessing method of the brain functional data.
[0029] Both the brain network A to be evaluated and the reference brain network R are undirected weighted brain functional connectivity networks, and are represented as N×N connectivity matrices. The rows and columns of the connectivity matrix correspond to network nodes, and the matrix elements represent the functional connectivity strength between the corresponding network nodes.
[0030] The brain functional connectivity strength can be calculated using functional connectivity indices that form an undirected weighted connectivity matrix. This invention does not limit the specific method for calculating functional connectivity; however, in the same candidate target evaluation process, the brain network A to be evaluated and the reference brain network R should use the same functional connectivity indices and consistent connectivity strength processing methods to ensure their comparability. In this embodiment, the Pearson correlation coefficient is used to calculate the functional connectivity strength between different network nodes to construct the corresponding undirected weighted brain functional connectivity matrix.
[0031] Transcranial electrical stimulation candidate targets are represented as candidate nodes in the brain network. In this embodiment, each candidate target corresponds to a candidate node in the brain network.
[0032] The brain network A to be evaluated is used to characterize the brain functional connectivity status of the subject or group to be evaluated. It can be the individual brain network of a single subject or the average brain network of a group of subjects with preset common characteristics.
[0033] The reference brain network R is used to characterize the reference brain functional state compared with the brain network to be evaluated. It can be obtained by averaging the brain networks of a healthy control group, or by using the average brain network of the group corresponding to the preset target state, a pre-constructed standardized brain network template, or other brain networks with clear reference significance.
[0034] S12: Perform consistency verification and standardization on the brain network to be evaluated and the reference brain network.
[0035] The brain network A to be evaluated and the reference brain network R should have the same or mutually mappable node definitions, and use the same functional connectivity index and consistent connection strength processing method to ensure that the two are comparable.
[0036] Consistency checks were performed on the matrix dimensions, number of nodes, node labels, and node order of the brain network A to be evaluated and the reference brain network R.
[0037] When two nodes have the same set of nodes but different node order, the rows and columns of the corresponding connection matrix are rearranged synchronously according to the node labels so that the matrix elements at the same position in the two connection matrices correspond to the same node connection.
[0038] When the node sets of the two are inconsistent, they are converted to a unified node space according to the preset node mapping relationship; when a reliable node mapping relationship cannot be established, or a connection matrix with consistent node correspondence still cannot be formed after node mapping, the current candidate target evaluation process is terminated.
[0039] After completing the consistency check, matrix standardization was performed on the brain network A to be evaluated and the reference brain network R.
[0040] The connection matrices of the brain network A to be evaluated and the reference brain network R are checked for symmetry. When the matrix contains asymmetric terms due to numerical calculation errors, it is symmetricized according to the following formula: Set the main diagonal elements of the brain network A to be evaluated and the reference brain network R to zero to eliminate the influence of the nodes' own connections on subsequent calculations.
[0041] Through the above processing, the brain network A to be evaluated and the reference brain network R have consistent node space, node arrangement order, matrix dimensions and connection strength expression, providing a unified brain network input for subsequent virtual modulation, full network deviation calculation and candidate target scoring.
[0042] Step 2: Within the brain network A to be evaluated and the reference brain network R processed in Step 1, determine the set of candidate transcranial electrical stimulation targets.
[0043] Within the brain network space after consistency verification and standardization, pre-determined candidate target information is obtained. Each candidate target is mapped to the brain network space of the unified node processed in step one. Each candidate target corresponds to a unique candidate node. Based on the mapped candidate nodes, a set of transcranial electrical stimulation candidate targets is constructed, denoted as: in, M represents the total number of candidate targets, and N represents the total number of brain network nodes. This represents the candidate node in the brain network corresponding to the m-th candidate target.
[0044] In this embodiment, each candidate target point corresponds to a candidate node in the brain network. Based on the data source and node definition of the brain network, the candidate node may correspond to a signal acquisition location, a preset brain region, a source spatial location, or other preset brain functional units.
[0045] The candidate target set can be all nodes in the brain network, or it can be a subset of nodes determined based on research objectives, literature review, prior knowledge, or preliminary screening results. This invention does not limit the specific number of candidate targets or the method of pre-determining them.
[0046] The candidate target set is used to limit the range of candidate nodes for subsequent initial anomaly calculation, virtual adjustment, network deviation calculation, quantitative scoring and sorting.
[0047] Step 3: Calculate the initial total network deviation of the brain network to be evaluated relative to the reference brain network, and calculate the initial abnormality of the relevant connections of each candidate node in the candidate target set relative to the reference brain network R before virtual adjustment.
[0048] S31: Calculate the initial full network deviation.
[0049] Based on the brain network A to be evaluated and the reference brain network R obtained after step one, the initial full-network deviation of brain network A relative to the reference brain network R is calculated. .
[0050] In this embodiment, the Frobenius norm is used to calculate the initial full network deviation. As shown in the following formula: in, The Frobenius norm of a matrix can be further expressed as: in, This represents the connection strength between node p and node q in the brain network A to be evaluated. This represents the connection strength between corresponding nodes in the reference brain network R.
[0051] The initial full network deviation This is used to characterize the overall degree of difference between the brain network A to be evaluated and the reference brain network R before virtual modulation. Under the condition that the brain network A to be evaluated and the reference brain network R have the same node definitions and connection strength expressions, The larger the value, the greater the deviation between the overall connection states of the two.
[0052] The initial full network deviation As a benchmark for evaluating the effects of subsequent virtual adjustment, it is used in subsequent steps to calculate the relative improvement of the entire network deviation under different virtual adjustment intensities.
[0053] S32: Calculate the initial anomaly quantity of candidate nodes.
[0054] For the candidate target set For each candidate node i in the network, calculate the initial abnormality of the associated connections of that candidate node relative to the reference brain network R before virtual conditioning. In this embodiment, the initial anomaly quantity of candidate node i The calculation is based on the sum of the absolute values of the differences between the candidate node and its related connections with other nodes in the brain network A to be evaluated and the reference brain network R, as shown in the following formula: In this process, j iterates through all network nodes except for candidate node i. This represents the connection strength between candidate node i and candidate node j in the brain network A to be evaluated. This represents the connection strength between candidate node i and candidate node j in the reference brain network R.
[0055] Initial outlier of candidate nodes This is used to characterize the degree to which the overall connectivity of candidate node i deviates from the reference brain network before virtual regulation. The larger the value, the greater the initial difference between the candidate node's associated connections and the corresponding connections in the reference brain network.
[0056] The initial anomaly of the candidate node This is used to normalize the results of the virtual adjustment comprehensive evaluation in order to reduce the impact of the differences in the initial anomaly scale of the related connections of different candidate nodes on the candidate target scoring and ranking results.
[0057] Step 4: Set multiple virtual adjustment intensity values within the preset virtual adjustment intensity range to form a virtual adjustment intensity sequence.
[0058] S41: Set the virtual adjustment intensity parameter and its value range.
[0059] Define virtual regulation intensity parameters The virtual adjustment intensity parameter is a dimensionless computational parameter used to characterize the proportion by which the relevant connections of candidate nodes move closer to the corresponding connection states in the reference brain network R from the initial connection states in the brain network to be evaluated A.
[0060] The virtual adjustment intensity parameter The theoretical allowable range is [0,1]. Based on specific evaluation requirements, the actual virtual adjustment intensity range is preset from the theoretical allowable range. ,in .
[0061] Therefore, the actual virtual adjustment intensity range can cover the entire [0,1] range, or it can be any preset sub-range within the [0,1] range, such as [0,0.5], [0,0.8] or [0.2,0.7].
[0062] when At this time, corresponding to the zero virtual adjustment state, the relevant connection strength of the candidate nodes remains unchanged from the initial connection strength in the brain network A to be evaluated; when When the state is fully virtual, the strength of the candidate node's associated connection is set to the strength of the corresponding connection in the reference brain network R.
[0063] When the actual virtual adjustment intensity range does not contain 0 or 1, such as when the range is set to [0.2, 0.7], the changes in the relevant connections of the candidate nodes are only calculated and evaluated within the set range. It is not required that the actual value sequence must contain zero virtual adjustment state and full virtual adjustment state.
[0064] The virtual adjustment intensity parameter It only represents the calculated proportion by which the brain network connection strength approaches the reference state. It is a mathematical calculation variable and does not correspond to the current intensity, stimulation frequency, stimulation duration or other physical stimulation parameters in actual transcranial electrical stimulation.
[0065] S42: Construct a virtual modulation intensity sequence.
[0066] In this example, the virtual adjustment intensity sequence is set at equal intervals. Within the preset virtual adjustment intensity range... Within, according to the preset step size Several virtual adjustment intensity values are set to form a virtual adjustment intensity sequence: Where K is the number of virtual adjustment intensity values, and each virtual adjustment intensity value satisfies: = The step size The sampling density of the sequence is determined based on preset computational accuracy and computational requirements. A smaller step size can increase the sampling density of virtual adjustment intensity values, improving the numerical resolution of subsequent improvement rate curves and their area under the curve calculations, but correspondingly increasing the computational load; a larger step size can reduce the computational load, but may reduce the accuracy of characterizing local changes in the improvement rate curve.
[0067] In this embodiment, the actual virtual adjustment intensity range is set to [0,1], and the step size is set to 0.05, thereby forming a virtual adjustment intensity sequence. .
[0068] During the same candidate target evaluation process, each candidate node uses the same virtual modulated intensity sequence to ensure the comparability of evaluation results between different candidate nodes.
[0069] Step 5: Traverse each candidate node in the candidate target set and each virtual modulation intensity in the virtual modulation intensity sequence, perform linear interpolation virtual modulation on the relevant connections between the candidate node and other nodes, and construct the corresponding virtual modulation postbrain network.
[0070] S51: Traverse candidate nodes and virtually adjust intensity.
[0071] Traverse the set of candidate targets determined in step two For each candidate node i, iterate through the virtual adjustment intensity sequence constructed in step four. Each virtual adjustment intensity For each group (i, Each candidate node was constructed with a corresponding virtual modulated hindbrain network to ensure that the network state of each candidate node under different modulation intensities was recorded and evaluated separately.
[0072] S52: Perform linear interpolation virtual adjustment on the relevant connections of candidate nodes.
[0073] For the currently processed candidate node i and the virtual adjustment intensity Based on the standardized brain network A to be evaluated in step one, only the connection strength between candidate node i and other nodes is virtually adjusted by linear interpolation. For any node i that satisfies the following conditions... For network node j, the virtual adjusted connection strength is calculated according to the following formula: [ in,[ Indicates the virtual adjustment intensity applied to candidate node i. The virtual adjusted connection strength between node j and node j; This represents the original connection strength of the connection between candidate node i and candidate node j in the brain network A to be evaluated. This represents the connection strength between candidate node i and candidate node j in the reference brain network R.
[0074] S53: Formation of a virtual regulatory hindbrain network.
[0075] Since this embodiment uses an undirected weighted brain functional connectivity network, the connection strength between candidate node i and node j is updated. Simultaneously update the corresponding symmetric positions in the connection matrix to satisfy the symmetry constraints of the undirected network: For connections between nodes other than candidate node i, their original connection strengths in the brain network A to be evaluated are kept unchanged, i.e.: At the same time, the virtual regulation of the hindbrain network will be implemented. Set the main diagonal elements to zero, that is: This allows us to obtain the virtual adjustment intensity of candidate node i. The corresponding virtual regulatory hindbrain network To simulate the virtual adjustment of the strength of the relevant connections of candidate node i from the state to be evaluated to the reference state. The calculation process of gradually approaching the target.
[0076] In this process, the virtual regulation process of different candidate nodes is independent of each other. Each candidate node uses the brain network A to be evaluated after step one as the starting network for virtual regulation and does not perform cumulative regulation on the virtual regulation results of other candidate nodes.
[0077] Step 6: Calculate the deviation and improvement rate of the entire network after virtual adjustment, and generate an improvement rate curve.
[0078] S61: Calculate the total network deviation after virtual adjustment.
[0079] For each virtual modulated hindbrain network obtained in step five Calculate its total deviation relative to the reference brain network R. As shown in the following formula: in, Denotes the Frobenius norm; Used to characterize the virtual adjustment strength of candidate node i Below, the degree of overall deviation of the virtual-modulated brain network relative to the reference brain network is observed. The smaller the value, the more it indicates virtual regulation of the hindbrain network. The smaller the overall deviation from the reference brain network R.
[0080] S62: Calculate the improvement rate under different virtual adjustment intensities.
[0081] Based on the initial full network deviation obtained in step S31 of step three The virtual adjusted total network deviation obtained in step S61 Calculate the virtual adjustment intensity of candidate node i. Improvement rate As shown in the following formula: Here, ε is a preset minimum positive number used to avoid zero denominators and improve the stability of numerical calculations. Used to represent the given virtual adjustment strength Under the condition that the virtual adjustment corresponding to candidate node i causes the change ratio of the total network deviation relative to the initial state.
[0082] when When, it indicates that the deviation of the virtual-modulated brain network relative to the reference brain network is lower than the initial deviation of the entire network; when When the time is specified, it indicates that the total network deviation remains unchanged before and after the virtual adjustment.
[0083] S63: Form the improvement rate curve of candidate nodes.
[0084] Traversing the virtual modulation intensity sequence For all values in the virtual modulation intensity sequence, calculate the candidate node i in the virtual modulation intensity sequence in turn. various virtual adjustment intensities The improvement rate was obtained by calculating the improvement rate sequence. .
[0085] According to the order of virtual adjustment intensity from small to large, the virtual adjustment intensities are... and the corresponding improvement rate Composition of coordinate points: Based on this, an improvement rate curve for candidate node i is generated, which is used to characterize the change in the degree of improvement of the entire network deviation corresponding to the candidate node as the virtual adjustment intensity changes.
[0086] Step 7: Calculate the area under the improvement rate curve and obtain the normalized AUC score of the candidate nodes.
[0087] S71: Calculate the area under the improvement rate curve.
[0088] Integrating the improvement rate curve of candidate node i obtained in step six, we obtain the area under the improvement rate curve corresponding to candidate node i. As shown in the following formula: in and These represent the lower and upper limits of the actual virtual adjustment intensity range, respectively. This is used to comprehensively characterize the overall improvement of candidate node i on the deviation of the entire network within a preset virtual adjustment intensity range.
[0089] Since the virtual adjustment intensity sequence Λ constructed in step four consists of multiple discrete values, the trapezoidal integral method is used in this embodiment to calculate the area under the improvement rate curve. As shown in the following formula: ( Where K is the number of virtual regulation intensity values in the virtual regulation intensity sequence; and These represent the values of two adjacent virtual adjustment intensities; This represents the interval between adjacent virtual adjustment intensity values. When the virtual adjustment intensity sequence is set using an equal-interval method, the interval between each adjacent value is a preset step size. .
[0090] In other implementations, other numerical integration methods capable of calculating the area under the curve based on discrete improvement rate data may also be used.
[0091] Under the same virtual adjustment intensity range and value sequence, The larger the value, the greater the overall improvement of the deviation of the entire network by candidate node i under multiple virtual adjustment intensities.
[0092] S72: Calculate the normalized AUC score.
[0093] Using the initial anomaly values of candidate nodes obtained in step three The area under the improvement rate curve for candidate node i Normalization is performed to obtain the normalized AUC score of candidate node i. As shown in the following formula: Here, ε is a preset minimum positive number used to avoid zero denominators and improve the stability of numerical calculations.
[0094] The normalized AUC score This is used to characterize the overall improvement of the entire network corresponding to a unit initial anomaly of candidate node i, in order to reduce the impact of differences in the initial anomaly scale of related connections among different candidate nodes on the evaluation results. Under the same computational conditions, The larger the value, the higher the relative evaluation priority of candidate node i. Normalized AUC score It serves as an evaluation metric for quantitative comparison and relative priority ranking among different candidate targets.
[0095] Step 8: Sort the candidate targets and output the quantitative evaluation results.
[0096] Normalized AUC score for each candidate node i in the candidate target set The candidate targets are sorted in descending order to obtain their relative priority ranking. Under the same computational conditions, the higher the normalized AUC score, the greater the overall improvement of the entire network corresponding to the unit initial outlier of the candidate node, and the higher its relative priority.
[0097] Output the quantitative evaluation and ranking results of candidate targets. The output includes candidate target labels, normalized AUC scores and rankings, and may also include improvement rate curve data for each candidate target and improvement rates under different virtual modulation intensities.
[0098] The ranking results are used to compare different candidate targets under a unified evaluation framework and to provide a calculation reference for determining candidate targets before actual transcranial electrical stimulation is performed.
[0099] The ranking results represent the relative priority of each candidate target within the brain network virtual regulation evaluation framework described in this invention, and are not directly equivalent to the clinical effect or actual stimulation effect of real transcranial electrical stimulation.
[0100] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is merely for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in the embodiments can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.
Claims
1. A method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets, characterized in that, Includes the following steps: Step 1: Obtain the brain network to be evaluated and the reference brain network, perform node and matrix consistency verification on both, and perform standardization processing according to the connection properties of the brain network; wherein, the brain network to be evaluated and the reference brain network are both N×N connection matrices, the rows and columns of the matrix correspond to network nodes, and the matrix elements represent the functional connection strength between the corresponding nodes. Step 2: Obtain the pre-determined candidate target information, map each candidate target to a candidate node in the brain network to be evaluated and the reference brain network processed in Step 1, and form a candidate target set; wherein, each candidate target corresponds to a candidate node, and the candidate target set includes all or part of the network nodes to be quantitatively evaluated and ranked. Step 3: Based on the brain network to be evaluated and the reference brain network processed in Step 1, calculate the initial full network deviation of the brain network to be evaluated relative to the reference brain network, and calculate the initial abnormality of the relevant connections of each candidate node in the candidate target set relative to the reference brain network before virtual adjustment. Step 4: Set multiple virtual adjustment intensity values within a preset virtual adjustment intensity range to form a virtual adjustment intensity sequence; Step 5: Traverse each candidate node in the candidate target set and each virtual modulation intensity in the virtual modulation intensity sequence, perform linear interpolation virtual modulation on the relevant connections between the candidate node and other nodes, and construct the corresponding virtual modulation backbrain network; Step 6: Calculate the deviation of each virtual-modulated brain network relative to the reference brain network, and calculate the improvement rate of each candidate node under the corresponding virtual modulation intensity based on the initial full network deviation and the virtual-modulated full network deviation; form the improvement rate curve corresponding to the candidate node based on the improvement rate of the same candidate node under different virtual modulation intensities. Step 7: Calculate the area under the improvement rate curve corresponding to each candidate node, and normalize the area under the curve using the initial outlier of the candidate node to obtain the normalized AUC score of each candidate node. Step 8: Sort all candidate nodes in descending order according to the normalized AUC score, and output the quantitative evaluation and relative priority ranking results of transcranial electrical stimulation candidate targets.
2. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, Step one specifically includes: The consistency of the brain network to be evaluated and the reference brain network is checked, including: Verify whether the matrix dimensions, number of nodes, node labels, node arrangement order, and connection strength expression methods of the brain network to be evaluated and the reference brain network are consistent; when the node sets are consistent but the node arrangement order is inconsistent, the rows and columns of the connection matrix are rearranged synchronously according to the node labels; when the node sets are inconsistent, the two are transformed into a unified node space according to the preset node mapping relationship; if node mapping cannot be completed, or if a connection matrix with consistent node correspondence cannot be formed after node mapping, the evaluation process is terminated. The connection matrix that has completed the consistency check is standardized, including setting the main diagonal elements of the brain network to be evaluated and the reference brain network to zero and performing symmetry processing.
3. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, The candidate target set in step two is predetermined by one of the following methods: All nodes in the brain network were used as candidate targets; Candidate targets are selected based on prior research, prior knowledge, specific diseases, cognitive tasks, or preliminary brain function analysis results; or Based on the initial difference analysis results between the brain network to be evaluated and the reference brain network, abnormal nodes with functional connectivity strength deviations exceeding a preset threshold were selected as candidate targets.
4. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, In step three, the initial whole-network deviation of the brain network to be evaluated relative to the reference brain network is... Quantity, and the virtual regulation of the hindbrain network in step six. Total deviation relative to the reference brain network All values were calculated using the Frobenius norm.
5. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, The initial anomaly of candidate node i in step three , is the sum of the absolute differences in the connections between the candidate node and other nodes in the brain network to be evaluated and the reference brain network, where the calculation process excludes the candidate node's own connections: In this process, j iterates through all network nodes except for candidate node i. This represents the connection strength between node i and node j in the brain network A to be evaluated. This represents the connection strength between node i and node j in the reference brain network R.
6. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, In step four, the virtual adjustment intensity parameter is a dimensionless calculation parameter with a theoretical value range of [0,1]; the virtual adjustment intensity sequence is constructed in the following way: Based on actual assessment needs, the actual virtual adjustment intensity range is selected within the range of [0,1]. Multiple discrete value points are set at equal or non-equal intervals within the interval to form a virtual adjustment intensity sequence; When the virtual adjustment strength is 0, the connection strength of the candidate node remains unchanged from the initial connection strength in the brain network A to be evaluated. When the virtual adjustment strength parameter is set to 1, the relevant connection strength of the candidate node is set to the strength of the corresponding connection in the reference brain network R.
7. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, Step five specifically includes: Local adjustment: at each virtual adjustment intensity In this case, only the relevant connections between candidate node i and other nodes are updated, while the connections between other nodes that are not involved in candidate node i remain at the original connection values in the brain network A to be evaluated. For any node The virtual adjusted connection strength is calculated according to the following formula: [ in,[ Indicates the virtual adjustment intensity of candidate node i. The virtual adjusted connection strength between node j and node j; This indicates the original connection strength of the corresponding connection in the brain network to be evaluated; This indicates the connection strength of the corresponding connection in the reference brain network; To maintain the symmetry of the connected matrix after virtual adjustment, the matrix elements at symmetrical positions are updated synchronously, and the elements on the main diagonal are set to zero.
8. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, In step six, the improvement rate uses the initial full-network deviation. Deviation of the entire network after adjustment of each virtual adjustment intensity calculate: in This is a preset, extremely small positive number, used to avoid zero denominators and improve the stability of numerical calculations.
9. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, In step seven, the area under the improvement rate curve is calculated using a numerical integration method; The normalized AUC score is obtained by dividing the area under the curve by the sum of the initial outlier and the smallest positive number of the corresponding candidate node, and is used to characterize the overall improvement of the entire network corresponding to a unit initial outlier.
10. The method for quantitative evaluation and ranking of transcranial electrical stimulation candidate targets according to claim 1, characterized in that, The output results in step eight include one or more of the following: candidate target labels, normalized AUC scores, and relative priority rankings, as well as, optionally, improvement rate curve data corresponding to each candidate target.
Citation Information
Patent Citations
Human brain transcranial stimulation circuit and human brain transcranial stimulation method
CN120285454A
