Stimulation coil positioning system and method based on transcranial magnetic stimulation therapy

Abstract

The invention relates to the technical field of medical instruments, and discloses a transcranial magnetic stimulation therapy-based stimulation coil positioning system and method.The system comprises a biological signal sensing module and a positioning decision module, the biological signal sensing module acquires multi-modal physiological data including cortical electric signals, hemodynamic data and magnetoencephalogram data in real time, and the positioning decision module determines whether the cortical electric signals, the hemodynamic data and the magnetoencephalogram data are in real time; and a target spot positioning instruction is generated through dynamic noise suppression and feature fusion of the modeling processing layer. The modeling processing layer comprises a signal preprocessing module and a collaborative optimization module, the signal preprocessing module carries out space-time segmentation, anomaly filtering and standardized registration on original data, and the collaborative optimization module realizes frequency domain denoising, anatomical correlation modeling and multi-dimensional fusion of multi-modal data through a biological noise filtering layer, a target point correction layer and a positioning execution layer. The method comprises the steps of data acquisition, noise suppression, feature fusion and positioning instruction generation. Through multi-modal data fusion and dynamic modeling, the positioning accuracy and clinical adaptability are improved, and the method is suitable for nerve regulation personalized treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and in particular to a stimulation coil positioning system and method based on transcranial magnetic stimulation therapy. Background Art

[0002] Transcranial Magnetic Stimulation (TMS), a non-invasive neuromodulation technology, has demonstrated significant application value in the treatment of mental illness and neurological rehabilitation. Its core principle is to generate a time-varying magnetic field through a stimulation coil, which penetrates the skull to stimulate neurons in the cerebral cortex, thereby regulating neural activity. However, one of the key challenges currently facing TMS technology in clinical applications is the accuracy and reliability of the stimulation coil positioning.

[0003] From the current state of technological development, traditional TMS localization primarily relies on anatomical imaging (such as MRI) for target planning. However, this positioning method based on static anatomical information has significant limitations. On the one hand, individual differences in brain structure and function lead to deviations between anatomical targets and actual functional areas. For example, neural activity patterns at the same anatomical location can vary significantly between different patients. On the other hand, brain function is dynamic, and the distribution of neural activity in the same patient changes during different physiological states (such as wakefulness, sleep, and medication effects) or treatment cycles. Static anatomical localization is difficult to adapt to these dynamic changes.

[0004] Among existing positioning technologies, although some attempts have been made to combine electrophysiological signals (such as EEG) or hemodynamic signals (such as fNIRS) for functional localization, multimodal data fusion methods are still immature. For example, single-modal signals are susceptible to noise interference (such as EEG signals are susceptible to artifacts such as muscle electrical activity and eye movements), and the temporal and spatial registration accuracy between different modal signals is insufficient (such as the temporal resolution difference between electrical signals and hemodynamic signals is several orders of magnitude), resulting in large errors in positioning results. In addition, existing systems lack effective modeling of the dynamic adaptability of treatment scenarios, making it difficult to perform personalized positioning optimization for different disease types (such as depression, Parkinson's disease) or individual pathological characteristics (such as lesion location, abnormal neural connections).

[0005] In clinical applications, positioning errors may lead to poor treatment effects or increased risk of side effects. For example, in dorsolateral prefrontal TMS treatment for depression, if the stimulation target deviates from the functionally active area, it may not effectively regulate emotion-related neural circuits, thereby reducing efficacy; while overstimulation of non-target areas may cause adverse reactions such as headaches and epilepsy. At the same time, the existing positioning process usually requires manual adjustment of the coil position, relying on the physician's experience and lacking automated and intelligent positioning strategies. This leads to a low degree of standardization in the treatment process and makes it difficult to promote and apply.

[0006] From the analysis of technology evolution trends, with the development of neuroimaging technology, biosensor technology and computational intelligence algorithms, multimodal physiological data fusion and dynamic positioning have become important development directions of TMS technology. However, how to achieve efficient noise suppression, precise spatiotemporal alignment and personalized modeling of multimodal data remains a technical problem that needs to be solved urgently. In existing studies, there is insufficient analysis of the frequency domain correlation characteristics of different modal signals, making it difficult to effectively separate biological noise from real neural signals; in terms of anatomical and functional correlation modeling, there is a lack of in-depth analysis of the topological structure of functional nodes in brain regions, resulting in inaccurate target compensation mechanisms; in addition, existing systems do not fully consider the impact of physiological delay effects (such as the time delay between neural activity and hemodynamic response) on positioning accuracy, which further restricts the improvement of positioning accuracy. Summary of the Invention

[0007] The object of the present invention is to provide a stimulation coil positioning system and method based on transcranial magnetic stimulation therapy to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a stimulation coil positioning system based on transcranial magnetic stimulation therapy, the system comprising:

[0009] A biosignal sensing module for acquiring multimodal physiological data from the patient's head region in real time, the multimodal physiological data comprising a first acquisition sequence corresponding to cortical electrical signal data, a second acquisition sequence corresponding to hemodynamic data, and a third acquisition sequence corresponding to magnetoencephalogram data. The cortical electrical signal data comprises first neural activity data generated by a high-density electrode array and second metabolic dynamics data collected by an optical sensor.

[0010] A positioning decision module is used to perform dynamic noise suppression processing on the multimodal physiological data and input the data into the modeling processing layer for feature fusion, and generate a target positioning instruction based on the output result of the modeling processing layer;

[0011] The modeling processing layer includes a signal preprocessing module and a collaborative optimization module, wherein the signal preprocessing module is used to perform spatiotemporal segmentation of the original physiological data stream and filter abnormal signals, and the collaborative optimization module is obtained by joint training based on historical physiological data and historical anatomical data of multiple treatment scenarios; the collaborative optimization module includes a biological noise filtering layer, a target correction layer and a positioning execution layer connected in sequence.

[0012] Preferably, the biological noise filtering layer is used to perform frequency domain correlation processing on different acquisition sequences in the original physiological data stream to generate denoised physiological feature data; the target correction layer is used to model the anatomical correlation relationship between the denoised physiological feature data corresponding to each acquisition sequence to generate compensated target field data; the positioning execution layer is used to perform multi-dimensional fusion based on the compensated target field data and the denoised physiological feature data to generate target positioning instructions.

[0013] Preferably, the modeling of the anatomical correlation relationship between the denoised physiological characteristic data corresponding to each acquisition sequence to generate compensated target field data includes:

[0014] A topology optimization algorithm is used to identify key functional nodes in the denoised physiological characteristic data, and a target compensation sequence corresponding to each acquisition sequence is determined based on the brain region type corresponding to each key functional node;

[0015] The deviation coefficient between the functional nodes of the same brain region in the target compensation sequences corresponding to any two acquisition sequences is calculated, and the compensation target field data between the any two acquisition sequences is generated based on the deviation coefficient.

[0016] Preferably, the calculation of the deviation coefficient between functional nodes in the same brain region in the target compensation sequences corresponding to any two acquisition sequences includes:

[0017] When the number of functional nodes in the target compensation sequences corresponding to any two acquisition sequences is inconsistent, virtual node interpolation is performed based on the brain region parameters corresponding to the terminal functional node in the one with the smaller number of functional nodes, and the deviation coefficient between the functional nodes in the same brain region is calculated based on the interpolated data.

[0018] Preferably, the signal preprocessing module is specifically used to:

[0019] Dividing the first acquisition sequence, the second acquisition sequence, and the third acquisition sequence with equal precision according to a preset brain area map to generate standardized first neural data, standardized second metabolic data, and standardized third magnetic image data;

[0020] The standardized first neural data and the standardized second metabolic data are aligned in real time using a dynamic graph clustering method, and the standardized third magnetic image data are noise-optimized using a steady-state frequency domain analysis method, and a first alignment sequence, a second alignment sequence, and a third alignment sequence are output; wherein the first alignment sequence includes the aligned first neural activity data and the aligned second metabolic dynamic data.

[0021] Preferably, the signal preprocessing module is further used for:

[0022] calculating a functional offset coefficient between the registered first neural activity data and the registered second metabolic dynamics data within a historical treatment cycle;

[0023] Predicting the expected metabolic value of the registered second metabolic dynamic data in the real-time treatment cycle according to the functional shift coefficient and the brain region parameter of the registered first neural activity data in the real-time treatment cycle;

[0024] Target optimization data is generated based on the registered second metabolic dynamic data and its expected metabolic value, and the acquisition sequence corresponding to the target optimization data is used as the first registration sequence.

[0025] Preferably, the target correction layer specifically includes:

[0026] a function mapping unit, configured to perform functional network tracing on each acquisition sequence in the denoised physiological feature data, so as to extract a corresponding function propagation chain from each acquisition sequence;

[0027] The brain region matching unit is used to align the functional propagation chain extracted from each acquisition sequence with the corresponding denoised physiological feature data in time and space to generate compensated target field data.

[0028] Preferably, the target correction layer further comprises:

[0029] The signal compensation unit is used to perform physiological delay effect elimination processing on the compensated target field data.

[0030] Preferably, the positioning execution layer specifically includes:

[0031] A multi-dimensional collaborative unit, comprising a plurality of positioning collaborative nodes, each of which is connected to each acquisition sequence in the compensated target field data and the denoised physiological characteristic data through parameter configuration;

[0032] a dynamic parameter optimization unit, configured to iteratively adjust the parameter configuration using a dynamic parameter optimization algorithm to minimize the error between the target positioning instruction and the actual function distribution;

[0033] The functional abnormality identification unit is used to locate the lesion area based on the compensated target field data and the denoised physiological characteristic data, and generate a target positioning instruction.

[0034] Preferably, the present invention further includes a method for positioning a stimulation coil based on transcranial magnetic stimulation therapy, the method comprising the following steps:

[0035] Step 1: Using a biosignal sensing module to acquire multimodal physiological data of the patient's head region in real time, the multimodal physiological data includes a first acquisition sequence corresponding to cortical electrical signal data, a second acquisition sequence corresponding to hemodynamic data, and a third acquisition sequence corresponding to magnetoencephalogram data. The cortical electrical signal data includes first neural activity data generated by a high-density electrode array and second metabolic dynamic data collected by an optical sensor.

[0036] Step 2: transmitting the multimodal physiological data to a positioning decision module, and having the positioning decision module perform dynamic noise suppression processing on the multimodal physiological data;

[0037] Step 3: Input the multimodal physiological data after dynamic noise suppression processing into the modeling processing layer in the positioning decision module for feature fusion, wherein the signal preprocessing module of the modeling processing layer performs spatiotemporal segmentation and abnormal signal filtering on the original physiological data stream, and the collaborative optimization module of the modeling processing layer is jointly trained based on historical physiological data and historical anatomical data of multiple treatment scenarios. The collaborative optimization module includes a biological noise filtering layer, a target correction layer, and a positioning execution layer connected in sequence;

[0038] Step 4: Based on the output results of the modeling processing layer, the positioning decision module generates a target positioning instruction to achieve the positioning of the stimulation coil based on transcranial magnetic stimulation treatment.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] In terms of multimodal data acquisition and preprocessing, the biosignal perception module acquires multimodal physiological data including cortical electrical signals, hemodynamic data, and magnetoencephalography data in real time. The cortical electrical signals are combined with optical sensors through a high-density electrode array to synchronously collect neural activity data and metabolic dynamics data, providing more comprehensive functional information for positioning. The signal preprocessing module divides the multimodal data with equal precision through preset brain area maps, realizes the standardization of data of different modalities, and solves the problems of insufficient information content of traditional single-modal data and inconsistent scale of multimodal data. At the same time, a dynamic graph clustering method is used to perform real-time registration of electrophysiological and metabolic data, and steady-state frequency domain analysis is combined to optimize the noise of magnetoencephalography data, effectively improving the spatiotemporal consistency and signal-to-noise ratio of the data, laying the foundation for subsequent feature fusion.

[0041] At the level of noise suppression and feature fusion, the biological noise filtering layer of the collaborative optimization module denoises different acquisition sequences through frequency domain correlation processing, which can specifically suppress specific noise in each modal signal (such as electromyographic noise in cortical electrical signals and motion artifacts in hemodynamic data). Compared with traditional filtering methods, it can better retain the frequency domain characteristics of real neural activity. The target correction layer identifies key functional nodes through a topology optimization algorithm, generates a target compensation sequence based on the type of brain region, and solves the problem of inconsistent numbers of functional nodes in different modal data through deviation coefficient calculation and virtual node interpolation processing, thereby achieving accurate modeling of anatomical association relationships. In addition, the functional propagation chain is extracted and spatiotemporally aligned through the functional mapping unit, and the physiological delay effect is eliminated in combination with the signal compensation unit, which further improves the spatiotemporal accuracy of the target compensation and ensures that the positioning results match the real-time neural activity distribution.

[0042] In terms of positioning decision-making and dynamic optimization, the multi-dimensional collaborative unit of the positioning execution layer implements parallel processing and parameter configuration of multimodal data through multiple positioning collaborative nodes. The dynamic parameter optimization unit minimizes positioning errors based on an iterative algorithm, enabling the system to adapt to individual differences and dynamic physiological changes within the same patient. The functional abnormality identification unit combines compensated target field data with denoised physiological characteristic data to localize lesions. This is not only suitable for precise stimulation of healthy brain areas, but also allows personalized adjustments for abnormal functional distribution in pathological conditions, broadening the system's clinical application scenarios (such as neurodegenerative diseases and mental disorders).

[0043] From a clinical application perspective, the system achieves a shift from static anatomical positioning to dynamic functional positioning. By integrating multi-dimensional physiological information with historical treatment data, it can generate personalized target positioning instructions for different patients and at different stages of treatment, thereby improving the effectiveness of TMS treatment. For example, in the treatment of depression, precise positioning of neural circuits related to emotion regulation can significantly improve efficacy and reduce ineffective stimulation; in the treatment of Parkinson's disease, dynamic tracking and positioning of brain areas related to motor control can optimize the improvement of motor symptoms. At the same time, the automated positioning process reduces reliance on physician experience, improves the standardization of the treatment process, and is conducive to the clinical promotion of TMS technology.

[0044] In terms of technological innovation, the system combines topology optimization algorithms, dynamic parameter optimization algorithms, and multimodal physiological data fusion technology for the first time, constructing a multi-level collaborative optimization model that includes biological noise filtering, target correction, and positioning execution, breaking through the bottlenecks of traditional positioning methods in dynamic adaptability and multimodal fusion efficiency. By modeling and compensating for physiological delay effects, the time difference between neural activity and physiological response is resolved, making the positioning results more in line with the real-time requirements of neural regulation. In addition, the application of technologies such as virtual node interpolation and functional offset coefficient prediction further enhances the system's ability to process complex physiological data and ensures the reliability of positioning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a working principle diagram of the stimulation coil positioning system and method based on transcranial magnetic stimulation therapy according to the present invention;

[0046] Figure 2 This is the workflow diagram of the collaborative optimization module in the target localization system;

[0047] Figure 3 Flowchart for identifying key functional nodes for topology optimization algorithm;

[0048] Figure 4 Flowchart for data standardization and registration of the signal preprocessing module;

[0049] Figure 5 Flowchart for data generation for target optimization based on functional shift coefficients. DETAILED DESCRIPTION

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] See also Figure 1 - Figure 5 The present invention relates to a stimulation coil positioning system based on transcranial magnetic stimulation therapy, comprising a biological signal sensing module and a positioning decision module. The system comprises:

[0052] The biosignal sensing module acquires multimodal physiological data of the patient's head area in real time. The multimodal physiological data includes a first acquisition sequence corresponding to cortical electrical signal data, a second acquisition sequence corresponding to hemodynamic data, and a third acquisition sequence corresponding to magnetoencephalogram data. The cortical electrical signal data includes first neural activity data generated by a high-density electrode array and second metabolic dynamic data collected by an optical sensor.

[0053] After the positioning decision module performs dynamic noise suppression processing on the multimodal physiological data, it inputs the data into the modeling processing layer for feature fusion and generates target positioning instructions based on the output results of the modeling processing layer.

[0054] The modeling processing layer includes a signal preprocessing module and a collaborative optimization module. The signal preprocessing module is used to perform spatiotemporal segmentation on the original physiological data stream and filter abnormal signals. The collaborative optimization module is obtained through joint training based on historical physiological data and historical anatomical data of multiple treatment scenarios, and includes a biological noise filtering layer, a target correction layer, and a positioning execution layer connected in sequence.

[0055] The present invention will be further described below in conjunction with Examples 1 to 5:

[0056] Example 1:

[0057] In the stimulation coil positioning system based on transcranial magnetic stimulation therapy, the biological noise filtering layer, the target correction layer and the positioning execution layer constitute the core processing link of the collaborative optimization module. Taking the physiological data processing of patients receiving transcranial magnetic stimulation therapy as an example, the biological noise filtering layer performs frequency domain correlation processing on the first acquisition sequence corresponding to the cortical electrical signal data, the second acquisition sequence corresponding to the hemodynamic data, and the third acquisition sequence corresponding to the magnetoencephalogram data. For example, the signals of different acquisition sequences are decomposed in the frequency domain to identify the energy distribution differences of each sequence in a specific frequency band (such as α wave and β wave band). The high-frequency noise or power frequency interference across sequences is removed through frequency domain correlation analysis to generate denoised physiological characteristic data.

[0058] The target correction layer performs anatomical correlation modeling on each denoised acquisition sequence. Taking the first and second acquisition sequences as examples, the key functional nodes in the denoised physiological feature data are first identified through a topology optimization algorithm. For example, the neural activity peak node of the dorsolateral prefrontal cortex is identified in the first acquisition sequence (cortical electrical signals), and the blood oxygen change node of the same brain region is identified in the second acquisition sequence (hemodynamics). These nodes are identified as key functional nodes, and target compensation sequences are generated for each sequence based on the brain region type (such as functional area, communication area).

[0059] When calculating the deviation coefficient of functional nodes in the same brain region in the target compensation sequences of two acquisition sequences (e.g., the first acquisition sequence contains 10 functional nodes and the third acquisition sequence contains 8 functional nodes, and both involve the postcentral parietal gyrus, the brain region parameters (e.g., three-dimensional coordinates, functional attributes) corresponding to the terminal functional node of the third acquisition sequence (e.g., the eighth node in the postcentral parietal gyrus) are used as the basis, and two virtual nodes are inserted at the end of the corresponding brain region in the first acquisition sequence to ensure the same number of nodes in both sequences. Subsequently, the deviation coefficient is generated by calculating parameters such as the spatial position difference and signal intensity difference of each node in the same brain region (e.g., the postcentral parietal gyrus). Based on this coefficient, the compensation target field data between the two sequences are constructed to reflect the compensation relationship between the anatomical differences of the different modal data.

[0060] The positioning execution layer integrates the compensated target field data with the denoised physiological characteristic data in multiple dimensions. For example, parameters such as the neural activity intensity of the first acquisition sequence, the hemodynamic response amplitude of the second acquisition sequence, and the magnetic field distribution phase of the third acquisition sequence are integrated through weighted summation or tensor fusion. Ultimately, a target positioning instruction containing information such as three-dimensional spatial coordinates and stimulation intensity threshold is generated, providing a precise basis for the physical positioning of the stimulation coil.

[0061] Example 2:

[0062] In the stimulation coil positioning system based on transcranial magnetic stimulation therapy, the signal preprocessing module processes the collected multimodal physiological data. For the first acquisition sequence corresponding to the cortical electrical signal data, the second acquisition sequence corresponding to the hemodynamic data, and the third acquisition sequence corresponding to the magnetoencephalogram data, the signal preprocessing module divides these three acquisition sequences with equal precision according to the preset brain area map. For example, the preset brain area map divides the brain into several functional areas. The signal preprocessing module maps the first acquisition sequence, the second acquisition sequence, and the third acquisition sequence to these functional areas according to the same accuracy standard, thereby generating standardized first neural data, standardized second metabolic data, and standardized third magnetogram data.

[0063] Next, the signal preprocessing module uses a dynamic graph clustering method to perform real-time registration of the standardized first neural data and the standardized second metabolic data. The dynamic graph clustering method takes into account the dynamic changes in time and space of neural activity and metabolic processes, and accurately matches the two. At the same time, the steady-state frequency domain analysis method is used to optimize the noise of the standardized third magnetic image data. The steady-state frequency domain analysis method can identify and remove noise interference generated by the external environment or the device itself in the third magnetic image data, and output the first registration sequence, the second registration sequence and the third registration sequence, where the first registration sequence contains the registered first neural activity data and the registered second metabolic dynamic data.

[0064] The signal preprocessing module also calculates the functional offset coefficient of the first neural activity data after registration and the second metabolic dynamic data after registration within the historical treatment cycle. For example, by analyzing the changing relationship between the two types of data in multiple historical treatment cycles, the functional offset law between them is determined. Then, based on the functional offset coefficient and the brain region parameters of the first neural activity data after registration in the real-time treatment cycle, the expected metabolic value of the second metabolic dynamic data after registration in the real-time treatment cycle is predicted. For example, based on the functional offset coefficient and the real-time brain region parameters, a mathematical model is established to infer the expected metabolic value. Finally, target optimization data is generated based on the second metabolic dynamic data after registration and its expected metabolic value, and the acquisition sequence corresponding to the target optimization data is used as the first registration sequence.

[0065] Example 3:

[0066] In the stimulation coil positioning system based on transcranial magnetic stimulation therapy, the target correction layer processes the denoised physiological characteristic data through the functional mapping unit, brain region matching unit, and signal compensation unit. Take the first acquisition sequence corresponding to the cortical electrical signal and the second acquisition sequence corresponding to the hemodynamic data as an example:

[0067] The functional mapping unit traces the functional network for each acquisition sequence. For example, for the first acquisition sequence (containing neural activity data), by analyzing the firing patterns and synchrony of neurons, the neural conduction pathway from the prefrontal cortex to the parietal cortex was traced, extracting the functional transmission chain of "dorsolateral prefrontal cortex → precentral gyrus → parietal angular gyrus". For the second acquisition sequence (containing metabolic dynamics data), blood oxygen level-dependent (BOLD) signal analysis was used to trace the blood perfusion change path corresponding to the aforementioned neural conduction pathway, extracting the functional transmission chain of "blood oxygen consumption in the dorsolateral prefrontal cortex → vasodilation in the precentral gyrus → increased oxyhemoglobin concentration in the parietal angular gyrus".

[0068] The brain region matching unit performs spatiotemporal alignment of the extracted functional propagation chains with the corresponding denoised physiological feature data. For example, the peak time point of neural activity in the dorsolateral prefrontal cortex in the first acquisition sequence (t = 150ms) is temporally aligned with the start time point of blood oxygen drop in the same brain region in the second acquisition sequence (t = 160ms). At the same time, the spatial coordinates of the central pregyrus in the two sequences (x = 20mm, y = -15mm, z = 55mm) are spatially matched to generate compensated target field data reflecting the spatiotemporal correlation between neural activity and metabolic response, which contains the signal conduction delay parameters and spatial correspondence of each brain region.

[0069] The signal compensation unit eliminates physiological delay effects on the compensated target field data. For example, because there is a physiological delay of approximately 1-5 seconds between neural activity and metabolic responses, the time axis of the metabolic data in the angular gyrus of the parietal lobe (collected at t = 200ms) and the corresponding metabolic data (collected at t = 2200ms) are shifted forward by 2000ms using a time offset correction algorithm, aligning the two in time. This eliminates the interference of physiological delay on target positioning and ultimately outputs the corrected compensated target field data for subsequent processing.

[0070] Example 4:

[0071] In the stimulation coil positioning system based on transcranial magnetic stimulation therapy, the positioning execution layer generates target positioning instructions through a multi-dimensional collaboration unit, a dynamic parameter optimization unit, and a functional abnormality identification unit. For example, the first acquisition sequence corresponding to cortical electrical signals, the second acquisition sequence corresponding to hemodynamic data, and the third acquisition sequence corresponding to magnetoencephalography data are processed:

[0072] The multidimensional collaboration unit consists of multiple positioning collaboration nodes, each of which is connected to each acquisition sequence in the compensated target field data and denoised physiological characteristic data through parameter configuration. For example, the first positioning collaboration node is connected to the neural activity intensity parameter (N1) of the first acquisition sequence, the blood oxygen change rate parameter (H2) of the second acquisition sequence, and the magnetic field gradient parameter (M3) of the third acquisition sequence. The parameter configuration is represented by the node's weight coefficients (w1, w2, w3) for each parameter.

[0073] The dynamic parameter optimization unit iteratively adjusts the parameter configuration through the dynamic parameter optimization algorithm. Assuming that the optimization goal is to minimize the error between the target positioning instruction and the actual function distribution, the error function is defined as:

[0074]

[0075] Among them, E represents the total error, n is the number of sample points, m is the number of acquisition sequences, T i is the actual function distribution coordinate value of the i-th sample point, S i,j is the characteristic value of the jth acquisition sequence at the i-th sample point, w j is the weight coefficient of the jth acquisition sequence. w is updated iteratively through gradient descent and other algorithms j , so that E is gradually reduced and the parameter configuration is optimized.

[0076] The functional abnormality recognition unit locates the lesion area based on the compensated target field data and the denoised physiological characteristic data. For example, by analyzing the abnormal high-frequency oscillation of neural activity in a certain brain area in the first acquisition sequence (such as continuous discharge exceeding 30Hz), the abnormal increase in blood perfusion in the corresponding brain area in the second acquisition sequence (such as the blood oxygen change rate exceeding the threshold H th ), and abnormal magnetic field fluctuations in the region during the third acquisition sequence (such as gradient values ​​exceeding M th ), comprehensively judge the brain area as an abnormal functional area, and generate target positioning instructions containing the coordinates of the area and stimulation priority, providing a basis for the precise positioning of the stimulation coil.

[0077] Example 5:

[0078] In the stimulation coil positioning system based on transcranial magnetic stimulation therapy, the positioning execution layer generates target positioning instructions through a multi-dimensional collaboration unit, a dynamic parameter optimization unit, and a functional abnormality identification unit. Taking the processing of cortical electrical signals (first acquisition sequence), hemodynamic data (second acquisition sequence), and magnetoencephalography data (third acquisition sequence) as an example:

[0079] The multidimensional collaborative unit contains multiple positioning collaborative nodes, each of which is connected to a different acquisition sequence through parameter configuration. For example, the first positioning collaborative node connects the neural activity frequency parameters (such as beta wave power value) of the first acquisition sequence and the magnetic field gradient parameters (such as x-axis gradient intensity) of the third acquisition sequence. The second node connects the blood oxygen saturation change rate of the second acquisition sequence and the neuronal synchronization index of the first acquisition sequence. Each node performs a preliminary fusion of the input parameters using a preset weight matrix (such as a neural activity frequency weight of 0.6 and a magnetic field gradient weight of 0.4) to form a local positioning feature.

[0080] The dynamic parameter optimization unit iteratively adjusts the parameter configuration through a dynamic parameter optimization algorithm (such as the particle swarm optimization algorithm). For example, the initial weight matrix is ​​set to [0.5, 0.5] (the weights of neural activity and hemodynamic data are equal). During the iteration process, the algorithm automatically adjusts the weight parameters of each node based on the spatial error between the target positioning instruction and the actual functional distribution (such as the deviation between the stimulation coil positioning coordinates and the lesion center displayed by the magnetic resonance imaging). If it is found that the neural activity data has a greater impact on the positioning accuracy, the algorithm will gradually increase the weight of the neural activity parameter to 0.7 and adjust the weight of the hemodynamic parameter to 0.3 until the error converges to the preset threshold.

[0081] The functional abnormality identification unit locates the lesion area based on the compensated target field data and denoised physiological characteristic data. For example, by analyzing the abnormal high-frequency discharge signal (frequency>30Hz) of the left temporal lobe cortex in the first acquisition sequence, the abnormal low blood perfusion (blood oxygen decrease>20%) in the second acquisition sequence, and the corresponding abnormal magnetic field fluctuations in the third acquisition sequence (magnetic field intensity standard deviation>5nT), a comprehensive judgment is made on the presence of an epileptic-like discharge lesion in the left temporal lobe cortex. Combined with the fusion features output by the multi-dimensional collaborative unit, a target positioning instruction containing the three-dimensional coordinates of the lesion center (x=-48mm, y=-12mm, z=20mm) and the recommended value of stimulation intensity (1.2T) is generated to guide the stimulation coil to align with the area for precise treatment.

[0082] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0083] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A stimulation coil positioning system based on transcranial magnetic stimulation therapy, characterized in that: include: A biosignal sensing module for acquiring multimodal physiological data from the patient's head region in real time, the multimodal physiological data comprising a first acquisition sequence corresponding to cortical electrical signal data, a second acquisition sequence corresponding to hemodynamic data, and a third acquisition sequence corresponding to magnetoencephalogram data. The cortical electrical signal data comprises first neural activity data generated by a high-density electrode array and second metabolic dynamics data collected by an optical sensor. A positioning decision module is used to perform dynamic noise suppression processing on the multimodal physiological data and input the data into the modeling processing layer for feature fusion, and generate a target positioning instruction based on the output result of the modeling processing layer; The modeling processing layer includes a signal preprocessing module and a collaborative optimization module, wherein the signal preprocessing module is used to perform spatiotemporal segmentation of the original physiological data stream and filter abnormal signals, and the collaborative optimization module is jointly trained based on historical physiological data and historical anatomical data of multiple treatment scenarios; The collaborative optimization module comprises a biological noise filtering layer, a target correction layer and a positioning execution layer which are connected in sequence.

2. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 1, characterized in that: The biological noise filtering layer is used to perform frequency domain correlation processing on different acquisition sequences in the original physiological data stream to generate denoised physiological feature data; The target correction layer is used to model the anatomical correlation between the denoised physiological feature data corresponding to each acquisition sequence and generate compensated target field data; The positioning execution layer is used to generate target positioning instructions based on multi-dimensional fusion of compensated target field data and denoised physiological feature data.

3. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 2, characterized in that: The step of modeling the anatomical correlation between the denoised physiological feature data corresponding to each acquisition sequence to generate compensated target field data includes: A topology optimization algorithm is used to identify key functional nodes in the denoised physiological characteristic data, and a target compensation sequence corresponding to each acquisition sequence is determined based on the brain region type corresponding to each key functional node; The deviation coefficient between the functional nodes of the same brain region in the target compensation sequences corresponding to any two acquisition sequences is calculated, and the compensation target field data between the any two acquisition sequences is generated based on the deviation coefficient.

4. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 3, characterized in that: The calculation of the deviation coefficient between functional nodes in the same brain region in the target compensation sequences corresponding to any two acquisition sequences includes: When the number of functional nodes in the target compensation sequences corresponding to any two acquisition sequences is inconsistent, virtual node interpolation is performed based on the brain region parameters corresponding to the terminal functional node in the one with the smaller number of functional nodes, and the deviation coefficient between the functional nodes in the same brain region is calculated based on the interpolated data.

5. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 1, characterized in that: The signal preprocessing module is specifically used for: Dividing the first acquisition sequence, the second acquisition sequence, and the third acquisition sequence with equal precision according to a preset brain area map to generate standardized first neural data, standardized second metabolic data, and standardized third magnetic image data; The standardized first neural data and the standardized second metabolic data are aligned in real time using a dynamic graph clustering method, and the standardized third magnetic image data are noise-optimized using a steady-state frequency domain analysis method, and a first alignment sequence, a second alignment sequence, and a third alignment sequence are output; wherein the first alignment sequence includes the aligned first neural activity data and the aligned second metabolic dynamic data.

6. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 5, characterized in that: The signal preprocessing module is further used for: calculating a functional offset coefficient between the registered first neural activity data and the registered second metabolic dynamics data within a historical treatment cycle; Predicting the expected metabolic value of the registered second metabolic dynamic data in the real-time treatment cycle according to the functional shift coefficient and the brain region parameter of the registered first neural activity data in the real-time treatment cycle; Target optimization data is generated based on the registered second metabolic dynamic data and its expected metabolic value, and the acquisition sequence corresponding to the target optimization data is used as the first registration sequence.

7. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 2, characterized in that: The target correction layer specifically includes: a function mapping unit, configured to perform functional network tracing on each acquisition sequence in the denoised physiological feature data, so as to extract a corresponding function propagation chain from each acquisition sequence; The brain region matching unit is used to align the functional propagation chain extracted from each acquisition sequence with the corresponding denoised physiological feature data in time and space to generate compensated target field data.

8. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 7, characterized in that: The target correction layer further comprises: The signal compensation unit is used to perform physiological delay effect elimination processing on the compensated target field data.

9. The stimulation coil positioning system based on transcranial magnetic stimulation therapy according to claim 2, characterized in that: The positioning execution layer specifically includes: A multi-dimensional collaborative unit, comprising a plurality of positioning collaborative nodes, each of which is connected to each acquisition sequence in the compensated target field data and the denoised physiological characteristic data through parameter configuration; a dynamic parameter optimization unit, configured to iteratively adjust the parameter configuration using a dynamic parameter optimization algorithm to minimize the error between the target positioning instruction and the actual function distribution; The functional abnormality identification unit is used to locate the lesion area based on the compensated target field data and the denoised physiological characteristic data, and generate a target positioning instruction.

10. A method for positioning a stimulation coil based on transcranial magnetic stimulation therapy, characterized in that: The following steps are involved: Step 1: Using a biosignal sensing module to acquire multimodal physiological data of the patient's head region in real time, the multimodal physiological data includes a first acquisition sequence corresponding to cortical electrical signal data, a second acquisition sequence corresponding to hemodynamic data, and a third acquisition sequence corresponding to magnetoencephalogram data. The cortical electrical signal data includes first neural activity data generated by a high-density electrode array and second metabolic dynamic data collected by an optical sensor. Step 2: transmitting the multimodal physiological data to a positioning decision module, and having the positioning decision module perform dynamic noise suppression processing on the multimodal physiological data; Step 3: Input the multimodal physiological data after dynamic noise suppression processing into the modeling processing layer in the positioning decision module for feature fusion, wherein the signal preprocessing module of the modeling processing layer performs spatiotemporal segmentation and abnormal signal filtering on the original physiological data stream, and the collaborative optimization module of the modeling processing layer is jointly trained based on historical physiological data and historical anatomical data of multiple treatment scenarios. The collaborative optimization module includes a biological noise filtering layer, a target correction layer, and a positioning execution layer connected in sequence; Step 4: Based on the output results of the modeling processing layer, the positioning decision module generates a target positioning instruction to achieve the positioning of the stimulation coil based on transcranial magnetic stimulation treatment.

Images