Evaluation system based on intracranial magnetic stimulation treatment effect of intracranial depression

Abstract

The invention provides an evaluation system based on a transcranial magnetic stimulation treatment effect of refractory depression, and belongs to the field of transcranial magnetic stimulation treatment evaluation. The problem that the objectivity and interpretability of an evaluation result are insufficient due to the fact that multi-modal data are not effectively integrated in the prior art is solved. The system comprises a neural navigation module which is used for determining an individualized transcranial magnetic stimulation target spot through a resting state functional magnetic resonance imaging technology and a remote isolation effect principle; the periodic treatment module is used for carrying out cranial magnetic stimulation frequency setting by adopting an alternate theta burst stimulation scheme; the data acquisition and processing module is used for acquiring multi-modal data before and after treatment and performing feature extraction; and the treatment evaluation module is used for establishing a multi-modal fusion curative effect prediction model through a support vector machine, and the model is used for evaluating the treatment effect and predicting the symptom relieving degree after being trained by a training data set. The method is mainly used in data evaluation field.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of transcranial magnetic stimulation treatment effect evaluation, and particularly relates to a transcranial magnetic stimulation treatment effect evaluation system based on refractory depression. BACKGROUND

[0002] Treatment-Resistant Depression (TRD) is a severe mental illness characterized by persistent low mood and high suicide rate, and the drug treatment remission rate is less than 30%. As a non-invasive neural regulation technology, repetitive transcranial magnetic stimulation (rTMS) is widely used in TRD treatment, but its efficacy evaluation system has significant defects. At present, the clinical mainly relies on self-reporting scale such as Hamilton Depression Scale (HAMD) and doctor's experience for subjective judgment, the evaluation dimension is single, the misjudgment rate is high, there is lack of objective biomarker support, and it is difficult to accurately reflect the improvement degree of neural function.

[0003] In recent years, research has found that the changes of brain function characteristics (such as brain network connection strength and neural oscillation pattern) before and after rTMS treatment can be used as objective indicators for efficacy prediction. However, although functional magnetic resonance imaging (fMRI) can provide high spatial resolution brain activity information, it has problems such as high data redundancy, complex analysis, high equipment cost and low ecological validity, which limits its clinical applicability. In contrast, resting-state electroencephalogram (EEG) and near-infrared spectroscopy (fNIRS) technology have the advantages of non-invasiveness, high temporal resolution and convenient operation. EEG can capture millisecond-level neural electrical activity, and fNIRS can monitor the dynamic blood oxygen metabolism in target areas such as prefrontal cortex in real time, and the two have complementarity in revealing neural regulation mechanisms. However, existing technologies have not effectively integrated multi-modal data, nor have they established a quantitative evaluation model based on machine learning, resulting in insufficient objectivity and interpretability of the evaluation results.

[0004] In addition, the traditional rTMS treatment system has two bottlenecks: first, the stimulation target positioning depends on the electroencephalogram 10-20 system or population template coordinates, and the positioning accuracy is usually greater than 5mm, making it difficult to achieve individualized precise navigation; second, the standard treatment course requires 4-6 weeks, which cannot meet the rapid intervention needs of patients with suicide risk. Although research has confirmed that functional connectivity "remote effect" can indirectly regulate deep brain areas, existing technologies lack an optimization scheme that combines individualized functional connectivity navigation with efficient stimulation mode, and have not formed a systematic solution from target positioning, treatment parameter optimization to multi-modal efficacy evaluation. SUMMARY

[0005] Therefore, the present application aims to provide a transcranial magnetic stimulation treatment effect evaluation system for refractory depression, so as to solve the problem that the prior art has not effectively integrated multi-modal data, resulting in insufficient objectivity and explainability of the evaluation results.

[0006] To achieve the above object, the present application adopts the following technical solution: a transcranial magnetic stimulation treatment effect evaluation system for refractory depression, comprising: A neural navigation module for determining individualized transcranial magnetic stimulation target points through resting-state functional magnetic resonance imaging technology and distant effect principle; A periodic treatment module for setting transcranial magnetic stimulation frequency using an alternating theta burst stimulation protocol; A data acquisition and processing module for acquiring multi-modal data before and after treatment and extracting features, the multi-modal data including clinical scale data, resting-state electroencephalogram data, and resting-state near-infrared spectroscopy data; A treatment evaluation module for building a multi-modal fusion treatment effect prediction model using support vector machines, the model being used to evaluate treatment effect and predict symptom relief degree after training by a training data set.

[0007] Further, a preferred mode is also proposed, wherein the neural navigation module comprises: A target recognition unit for determining deep brain regions related to depression as control targets; A functional connectivity analysis unit for finding the strongest stimulation target points in the stimulable cortical region functionally connected to the target brain region based on resting-state functional connectivity maps; A navigation execution unit for optical tracking to achieve target positioning.

[0008] Further, a preferred mode is also proposed, wherein the deep brain region is the pregenual cingulate gyrus, the stimulable cortical region is the left dorsolateral prefrontal cortex within a depth of 2-4 cm, and the functional connectivity analysis uses the Pearson correlation coefficient method to calculate the whole brain functional connectivity strength.

[0009] Further, a preferred mode is also proposed, wherein the periodic treatment module specifically uses Stanford accelerated intelligent neuromodulation therapy, comprising: An alternating stimulation unit for alternating 5 times of accelerated intermittent theta burst stimulation and 5 times of accelerated continuous theta burst stimulation per day; A parameter setting unit for controlling the stimulation parameters to be 50Hz triplet pulse sequence, with a total pulse number of 18000 times / day; An intensity control unit for adjusting the stimulation intensity to be 80% of the resting motor threshold.

[0010] Further, a preferred mode is proposed, wherein the accelerated intermittent theta burst stimulation lasts for 10 minutes each time, the accelerated continuous theta burst stimulation lasts for 2 minutes each time, the continuous treatment cycle lasts for 5 days, and the total number of pulses reaches 90000.

[0011] Further, a preferred mode is proposed, wherein the data acquisition and processing module comprises: A brain electrical acquisition unit is configured to acquire 5-10 minutes of eye-closed resting state data at a sampling rate of 1000 Hz by using a 128-lead EEG cap. A near-infrared acquisition unit is configured to monitor the hemoglobin dynamics of the frontal-parietal network. A feature extraction unit is configured to calculate the EEG power spectral density, phase locking value, hemoglobin concentration mean value, and network functional connectivity index based on the acquired resting state data and hemoglobin dynamics.

[0012] Further, a preferred mode is proposed, wherein the feature extraction unit specifically performs: EEG feature extraction: calculate the relative power of δ, θ, α, β, and γ bands, and the phase locking value. Near-infrared feature extraction: calculate the slope of the frontal lobe oxygenated hemoglobin concentration change and the global efficiency index of the network functional connectivity matrix. Cross-modal feature fusion: perform spatio-temporal coupling analysis on the α wave power and the slope of the frontal lobe hemoglobin concentration change.

[0013] Further, a preferred mode is proposed, wherein the treatment evaluation module comprises: A model training unit is configured to train a support vector machine classification model and a regression model by using the leave-one-out cross-validation method. A feature selection unit is configured to filter the baseline brain function features with significant differences as model inputs by using the T-test. An effect verification unit is configured to calculate the area under the ROC curve and the symptom improvement prediction correlation coefficient.

[0014] Further, a preferred mode is proposed, wherein the support vector machine classification model uses a linear kernel function, the penalty coefficient C=10, and the input features include the baseline θ / α ratio, the slope of the frontal lobe hemoglobin concentration, and the HAMD score corresponding to the clinical scale. The support vector machine regression model uses an RBF kernel function, the penalty coefficient C=15, and the output is the continuous prediction value of the symptom improvement rate.

[0015] Further, a preferred mode is proposed, wherein the remote effect principle is specifically implemented as: By stimulating the left DLPFC target, the neural activity of sgACC is indirectly regulated through the structural-functional connection pathway. Target points are screened according to the standard of Pearson correlation coefficient r>=0.6 of functional connection strength.

[0016] Compared with the prior art, the beneficial effects of the present application are: 1、The evaluation system proposed in the present application breaks through the subjective limitations of traditional single scale evaluation by integrating clinical scales (HAMD, C-SSRS), resting state electroencephalogram (EEG) spectral features, near-infrared spectroscopy (fNIRS) network functional connections and cross-modal biomarkers, and constructing a multi-dimensional evaluation index library. Clinical verification shows that the evaluation prediction AUC value of the joint model reaches 0.89, and the sensitivity and specificity are improved to 85% and 82% respectively, which is significantly better than single scale evaluation.

[0017] 2、The evaluation system proposed in the present application quantifies key indicators such as EEG alpha wave power, intuitively reflects prefrontal nerve excitability, integrates fNIRS prefrontal HbO fluctuation amplitude and global network efficiency, establishes a linear regression model of symptom relief degree and biomarker, and captures the dynamic evolution rule of brain function characteristics by comparing baseline data before treatment, and multi-time point data during and after treatment, realizes early prediction and long-term prognosis evaluation of treatment effect.

[0018] 3、The individualized neural navigation module based on the principle of distant effect of the evaluation system proposed in the present application improves the target positioning accuracy to <2mm (traditional method >5mm), combines the high-frequency burst stimulation mode (a-iTBS / a-cTBS alternation) to shorten the treatment cycle from the conventional 30 days to 5 days, and the simultaneously collected EEG-fNIRS dynamic data can quantitatively verify the changes of nerve excitation / inhibition state of the target area and the functional connection of the deep brain area (such as sgACC), and provide real-time biological feedback for efficacy evaluation. The finally formed evaluation system not only realizes the efficacy prediction before treatment (the prediction efficiency R² of the baseline period alpha / theta power ratio combined with HbO slope is 0.76), but also can dynamically track the neural plasticity change trajectory through the comparison of multi-time point (1 week / 1 month) longitudinal data, provide data support for optimizing individualized treatment plan, and the overall evaluation efficiency is more than 3 times higher than that of the traditional method. BRIEF DESCRIPTION OF DRAWINGS

[0019] The drawings that form a part of this application are used to provide a further understanding of the application, and the illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A schematic diagram of an evaluation system for the treatment effect of transcranial magnetic stimulation for refractory depression according to the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict, and the described embodiments are only a part of the embodiments of the present application, not all the embodiments.

[0021] Embodiment one, see Figure 1 This embodiment is described. The evaluation system for the treatment effect of transcranial magnetic stimulation for refractory depression described in this embodiment comprises: The neural navigation module is used to determine the individualized transcranial magnetic stimulation target point through resting-state functional magnetic resonance imaging technology and the principle of remote effect; The periodic treatment module is used to set the transcranial magnetic stimulation frequency by adopting an alternating theta burst stimulation scheme; The data acquisition and processing module is used to acquire multi-modal data before and after treatment and perform feature extraction, wherein the multi-modal data includes clinical scale data, resting-state electroencephalogram data and resting-state near-infrared spectroscopy data; The treatment evaluation module is used to build a multi-modal fusion treatment effect prediction model through a support vector machine, and the model is used to evaluate the treatment effect and predict the symptom relief degree after being trained by a training data set.

[0022] The present embodiment significantly improves the objectivity and prediction accuracy of the treatment effect of rTMS for refractory depression by constructing a multi-modal biomarker evaluation system. First, in the evaluation dimension, the limitations of traditional single scale evaluation are broken through, and multi-dimensional data such as resting-state electroencephalogram (EEG) high-frequency oscillation characteristics, near-infrared (fNIRS) prefrontal lobe blood oxygen metabolism dynamics, and clinical scales are integrated to construct an evaluation model containing multiple quantitative indicators such as power spectral density, functional connectivity strength, and network topology attributes, and the objectivity of the efficacy evaluation is improved to the biomarker level. Secondly, in the evaluation method, support vector machine (SVM) algorithm is used for feature fusion and dimension reduction processing of multi-modal data, and through leave-one-out cross-validation (LOOCV), it is confirmed that the recognition accuracy of the combined model for treatment responders (AUC=0.89) is about 40% higher than that of single scale evaluation, and the prediction error of symptom improvement (MAE=1.82 points) is significantly lower than that of clinical experience judgment (MAE≥3.5 points). In addition, the individualized neural navigation module based on the principle of remote effect improves the target positioning accuracy to <2mm (traditional method >5mm), and combined with the high-frequency burst stimulation mode (a-iTBS / a-cTBS alternation), the treatment cycle is shortened from the conventional 30 days to 5 days, and the simultaneously collected EEG-fNIRS dynamic data can quantitatively verify the changes of neural excitation / inhibition state in the target area and the functional connectivity of deep brain areas (such as sgACC), providing real-time biological feedback for efficacy evaluation. The finally formed evaluation system not only realizes the prediction of treatment effect before treatment (the prediction performance R²=0.76 of the baseline period α / θ power ratio combined with HbO slope), but also can dynamically track the neural plasticity change trajectory through the comparison of multi-time point (1 week / 1 month) longitudinal data, providing data support for optimizing individualized treatment plan, and the overall evaluation efficiency is more than 3 times higher than that of the traditional method.

[0023] Embodiment two, the present embodiment is a further limitation of the evaluation system for the treatment effect of transcranial magnetic stimulation for refractory depression according to embodiment one, the neural navigation module comprises: a target recognition unit for determining a deep brain area related to depression as a control target; a functional connectivity analysis unit for finding the strongest stimulation target in the stimulable cortical area functionally connected to the target brain area based on the resting-state functional connectivity map; a navigation execution unit for optical tracking to achieve target positioning.

[0024] Embodiment three, the present embodiment is a further limitation of the evaluation system for the treatment effect of transcranial magnetic stimulation for refractory depression according to embodiment two, the deep brain area is the subgenual anterior cingulate gyrus, the stimulable cortical area is the region within 2-4 cm depth of the left dorsolateral prefrontal cortex, and the functional connectivity analysis uses Pearson correlation coefficient method to calculate the whole brain functional connectivity strength.

[0025] Embodiment four, the embodiment is one kind based on the further limitation of the treatment effect evaluation system of refractory depression transcranial magnetic stimulation described in embodiment one, the cycle treatment module specifically adopts Stanford accelerated intelligent neuromodulation therapy, including: Alternating stimulation unit, for implementing 5 times accelerated intermittent theta burst stimulation and 5 times accelerated continuous theta burst stimulation alternately every day; Parameter setting unit, for controlling the stimulation parameter to be 50Hz triplet pulse sequence, and the total pulse number reaches 18000 times / day; Intensity control unit, for adjusting the stimulation intensity to be 80% resting motor threshold.

[0026] Embodiment five, the embodiment is one kind based on the further limitation of the treatment effect evaluation system of refractory depression transcranial magnetic stimulation described in embodiment four, the accelerated intermittent theta burst stimulation lasts for 10 minutes each time, the accelerated continuous theta burst stimulation lasts for 2 minutes each time, the continuous treatment cycle is 5 days, and the total pulse number reaches 90000 times.

[0027] Embodiment six, the embodiment is one kind based on the further limitation of the treatment effect evaluation system of refractory depression transcranial magnetic stimulation described in embodiment one, the data acquisition and processing module includes: Electroencephalogram acquisition unit, for acquiring 5-10 minute closed-eye resting state data by using 128-lead electroencephalogram cap with 1000Hz sampling rate; Near-infrared acquisition unit, for monitoring the hemoglobin dynamics change of frontal-parietal network; Feature extraction unit, for calculating electroencephalogram power spectral density, phase locking value, hemoglobin concentration mean value and functional connection index according to the acquired resting state data and hemoglobin dynamics change.

[0028] Embodiment seven, the embodiment is one kind based on the further limitation of the treatment effect evaluation system of refractory depression transcranial magnetic stimulation described in embodiment six, the feature extraction unit specifically executes: Electroencephalogram feature extraction: calculate the relative power of δ, θ, α, β and γ frequency band; Near-infrared feature extraction: calculate the slope of frontal lobe oxygenated hemoglobin concentration change, global efficiency of network functional connection matrix and other graph analysis indexes; Cross-modal feature fusion: time-space coupling analysis is performed on the alpha wave power and the slope of frontal lobe hemoglobin concentration change.

[0029] Embodiment eight, the embodiment is one kind based on the further limitation of the treatment effect evaluation system of refractory depression transcranial magnetic stimulation described in embodiment one, the treatment evaluation module includes: a model training unit configured to train a support vector machine classification model and a support vector machine regression model by using a leave-one-out cross-validation method; a feature selection unit configured to filter baseline brain function features with significant differences as model inputs by using a T-test; an effect verification unit configured to calculate an area under a ROC curve and a symptom improvement prediction correlation coefficient.

[0030] In the ninth implementation, the support vector machine classification model adopts a linear kernel function, a penalty coefficient C=10, and input features including a baseline θ / α ratio, a prefrontal lobe hemoglobin concentration slope, and a HAMD score corresponding to a clinical scale. The support vector machine regression model adopts an RBF kernel function, a penalty coefficient C=15, and an output of a continuous prediction value of a symptom improvement rate.

[0031] In the tenth implementation, the remote effect principle is specifically implemented as follows: The neural activity of the sgACC is indirectly regulated through a structural-functional connection path by stimulating the left DLPFC target point. The target point is selected according to a standard of a Pearson correlation coefficient r≥0.6.

[0032] In the eleventh implementation, the system for evaluating the treatment effect of transcranial magnetic stimulation on refractory depression is specifically implemented as follows, and is also used to explain the second to tenth implementations. The system for evaluating the treatment effect of transcranial magnetic stimulation on refractory depression comprises a neural navigation module, a periodic treatment module, a data acquisition module, and a treatment evaluation module.

[0033] The neural navigation module is configured to determine an individualized transcranial magnetic stimulation target point by using a resting-state functional magnetic resonance imaging technology and a remote effect principle, and comprises the following steps. The structural item and functional item data of the resting-state functional magnetic resonance of the patient with depression are acquired, and then a functional connection map of a target brain region (such as the subgenual cingulate gyrus) and the whole brain is calculated. For example, the target coordinates of the subgenual cingulate gyrus (sgACC) are x=-24, y=-18, and z=-18. The point with the strongest functional connection (such as x=-44, y=40, and z=29) in the left prefrontal cortex region (within a depth of 2-4 cm) that can be directly stimulated by TMS is searched for in the left prefrontal cortex region (within a depth of 2-4 cm) that can be directly stimulated by TMS.

[0034] The effective stimulation depth of TMS is about 2-4 cm, which limits TMS to directly stimulate the surface region of the brain, and the stimulation of deeper structures (such as the hippocampus) needs to rely on the "remote effect" of functional connection. The principle of remote effect refers to the fact that more distant brain regions can be affected by the stimulation site through structural or functional connections, which means that although TMS cannot directly stimulate deep brain regions, it can indirectly affect the activities of these deep regions by stimulating the surface regions that have strong functional connections with them. The operation process of rTMS stimulation guided by resting-state functional connection first identifies the target, such as identifying the target deep brain region related to depression, such as subgenual anterior cingulate cortex (sgACC), and then finding the local maximum point with the strongest functional connection with the target in the left dorsolateral prefrontal cortex (DLPFC) as the stimulation target. The TMS positioning method guided by resting-state functional connection achieves indirect regulation of deep brain regions through "remote effect", overcoming the depth limitation of TMS stimulation.

[0035] The periodic treatment module is used for transcranial magnetic stimulation frequency setting with an alternating theta burst stimulation scheme, comprising: In this embodiment, the alternating mode in Stanford Accelerated Smart Neuromodulation Therapy is used, that is, a-iTBS (accelerated intermittent Theta Burst Stimulation) and a-cTBS (accelerated continuous Theta Burst Stimulation) are used alternately. The specific parameters are 50Hz triplets for 5 days, stimulation is performed once every hour, 10 treatments per day (18000 pulses / day), 5 a-iTBS and 5 a-cTBS respectively, from the internal frequency 50Hz, the internal number is 3, the interval frequency is 5Hz, and the MNI coordinates in the left DLPFC (-44, 40 and 29) and the right DLPFC (44, 40 and 29) are accurately positioned under neural navigation. Continuous 5 days (total 90000 pulses), the stimulation intensity is 80% resting motor threshold (RMT), and the efficacy of theta burst stimulation (TBS) can be improved by acceleration, interval delivery of stimulation, increase in total pulse number and individualized target positioning.

[0036] The data acquisition module is used for acquiring multi-modal data before and after treatment and extracting features, and the multi-modal data includes clinical scale data, resting-state electroencephalogram data and resting-state near-infrared spectroscopy data, including: Data acquisition parameter setting, wherein the electroencephalogram data acquisition device includes an electroencephalogram acquisition instrument, a 128-lead electroencephalogram cap and an amplifier, the positions of the electrodes are placed according to the international standard lead 10-20 system standard, the reference electrode is Cz, the EEG parameters are: sampling rate 1000 Hz (capture high frequency oscillation), impedance ≤10 kΩ, record 5-10 minutes of closed-eye resting state; key frequency band: δ (1-4 Hz), θ (4-8 Hz), α (8-13 Hz), β (13-30 Hz), γ (30-80 Hz); fNIRS parameters: sampling rate ≥10 Hz, light source-detector distance 25-30 mm, wavelength 690 nm / 830 nm, monitor frontal lobe / default mode network (DMN) target area HbO / HbR changes; the sampling rate is 10 Hz; Before and after transcranial magnetic stimulation treatment, the patient's clinical symptom scale (such as Hamilton Depression Scale (HAMD-17) and Columbia Suicide Severity Rating Scale (C-SSRS)) needs to be collected, and PHQ-9 item score is used to evaluate depressive symptoms at the end of 10 times of stimulation per day. The electroencephalogram and near-infrared data of 5-10 minutes of closed-eye resting state are also collected, and the EEG and fNIRS time stamps are aligned using a hardware trigger (such as LabStreamingLayer) in a synchronous scheme, and then EEGlab and homer2 are used for data preprocessing. The main processes of EEG preprocessing include removing power frequency noise (50 / 60 Hz notch filtering); ICA removes eye movement / EMG artifacts; segmentation (non-overlapping 2s epoch); the main processes of fNIRS preprocessing include motion artifact correction (such as Wavelet filtering); band-pass filtering (0.01-0.1 Hz); optical path correction (DPF=6); Feature extraction is performed on the preprocessed data, including: Extracting electroencephalogram feature value indicators: power spectral density (calculating the relative power of each frequency band, such as α / θ ratio), phase lock value (PLV) or weighted phase lag index (wPLI) to evaluate the synchronization of brain regions; fNIRS features include hemoglobin dynamics: HbO / HbR mean, fluctuation amplitude, global efficiency, local efficiency, node degree, clustering coefficient and other graph theory indicators representing network topology properties, and the calculation formula is as follows: Power spectral density: , wherein, X ( f ) is the Fourier transform of the signal x ( t ), which represents the complex form of the signal in the frequency domain, and X ( f )∣2 is the power density of the signal at frequency f ; Phase lock value: , wherein,Φx t ) and Φy t represent the instantaneous phase of the signals x and y at time point t , i is the imaginary unit, exp [ i Φx t −Φy t ]) is the complex exponential representation of the phase difference; the sum of the complex exponentials for all time points is taken, and the absolute value is taken, and then divided by the total number of time points N, to obtain the PLV value; HbO / HbR mean, representing the mean of the concentrations of oxygenated hemoglobin and deoxygenated hemoglobin, respectively:

[0037] wherein HbO(t) and HbR(t) represent the concentrations of oxygenated hemoglobin and deoxygenated hemoglobin at time point t, respectively, and N is the total number of time points, i.e., the number of data points; Global efficiency: wherein represents the shortest path length from node i to j , the higher the value, the stronger the network information transmission efficiency; Local efficiency:

[0038] wherein represents the shortest path length (or distance) between node j and node k, and N i represents the number of neighbors of node i: i.e., the total number of nodes directly connected to node i (excluding node i itself); Clustering coefficient:

[0039] wherein E is the actual number of edges between neighbor nodes, k is the number of neighbor nodes, and k(k-1) is the maximum possible number of edges between neighbor nodes.

[0040] A treatment evaluation module is configured to build a multi-modal fusion treatment effect prediction model by using a support vector machine, wherein the model is trained by a training data set and is used to evaluate the treatment effect and predict the symptom relief degree, and the treatment evaluation module comprises: ​​​​​​The paired t-test / Wilcoxon test is used to compare the changes in the extracted features before and after treatment, and the effect size (Cohen's d) is calculated. If the patient responds, the depression score decreases, the alpha power significantly increases, especially in the left DLPFC target area, the HbO fluctuation amplitude decreases, especially in the frontal-limbic system pathway, and the non-responders are the opposite; in addition, the brain topography (EEG power change) and connection matrix (fNIRS functional connection difference) can be visualized and analyzed. Input features and use filter method for dimensionality reduction (T-test), such as baseline EEG (such as theta / alpha ratio) + fNIRS (frontal HbO slope) + clinical scale (HAMD score); The kernel function of the support vector machine classifier is a linear kernel function, and the penalty coefficient C is set to 10; the support vector machine classification model is used to evaluate the result type predicted by the evaluation model, and the penalty coefficient C is set to 15; specific including: accuracy, sensitivity and specificity; and using precision, recall, F1 score, accuracy to evaluate the evaluation model, and drawing ROC curve, the calculation formula is as follows: Accuracy:

[0041] Sensitivity: Sensitivity=TP / (FN+TP) Specificity: Specificity=TN / (FP+TN) Precision: Precision=TP / (FP+TP) TP, TN, FP, FN represent the number of true positive, true negative, false positive and false negative samples, respectively; Based on the leave-one-out cross-validation method (select one sample as the test set each time, and the remaining data as the training set, repeat the process until each sample has been tested once) and use the validation dataset to validate the trained evaluation model, including: The indicators for validating the support vector machine regression model are mean absolute error MAE, mean square error MSE, determination coefficient R2, correlation coefficient R, and explained variance score (EVS); after obtaining the predicted results of depression rTMS treatment, calculate the Pearson correlation coefficient R between the predicted Hamilton Depression Scale score after treatment and the actual scale score after treatment minus the pre-treatment score; R is the rTMS efficacy prediction model based on MDD's multi-modal data, which uses the changes in pre-treatment multi-modal data to predict the degree of relief of depressive symptoms in patients with depression after rTMS treatment, and the calculation formula is as follows: Mean Absolute Error MAE: MAE=

[0042] Mean Squared Error (MSE): MSE=

[0043] Coefficient of Determination (R) 2 : R 2 =

[0044] where y i is the true value (observed value) of the i-th sample, i is the predicted value of the i-th sample; Correlation Coefficient (R): R=

[0045] where Cov(X,Y) represents the covariance of random variables X and Y, and the covariance is used to measure the overall error of two random variables, represents the standard deviation of random variable X; Explainable Variance Score (EVS): EVS=

[0046] where, represents the variance of the residual, and Var(y) represents the variance of the actual observed value of the dependent variable.

[0047] Compared with the positioning accuracy of the traditional rTMS >5mm (electroencephalogram 10-20), the positioning accuracy of the evaluation system proposed in the application is <2mm (optical tracking), and the treatment cycle is 5 days, which is much smaller than that of the traditional rTMS (30 days), and at the same time, the evaluation dimension of the application is based on multi-modal (scale, electroencephalogram index, near-infrared index), and the evaluation accuracy is higher.

[0048] The specific embodiments of the application disclosed above are only used to help explain the application. The specific embodiments do not describe all the details, nor limit the application to the specific embodiments described. According to the content of the specification, many modifications and changes can be made. The specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application.

Claims

1. An evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression, characterized in that: include: A neuronavigation module for identifying individualized transcranial magnetic stimulation targets using resting-state functional magnetic resonance imaging and the principle of remote effect; Cycle therapy module for setting the frequency of transcranial magnetic stimulation using an alternating theta burst stimulation protocol; A data acquisition and processing module is used to collect multimodal data before and after treatment and perform feature extraction. The multimodal data includes clinical scale data, resting-state EEG data, and resting-state near-infrared spectroscopy data; The treatment evaluation module is used to construct a multimodal fusion efficacy prediction model through a support vector machine. After being trained with a training data set, the model is used to evaluate the treatment effect and predict the degree of symptom relief.

2. The evaluation system for the treatment effect of refractory depression based on transcranial magnetic stimulation according to claim 1, characterized in that: The neuronavigation module includes: a target recognition unit, which identifies deep brain regions associated with depression as targets for regulation; Functional connectivity analysis unit, used to find stimulation targets with the strongest functional connectivity with the target brain area in the stimulatable cortical area based on the resting-state functional connectivity map; Navigation execution unit, used for optical tracking to achieve target positioning.

3. The evaluation system for the treatment effect of refractory depression based on transcranial magnetic stimulation according to claim 2, characterized in that: The deep brain region is the subgenual anterior cingulate gyrus, the stimulatable cortical area is the area within a depth of 2-4 cm in the left dorsolateral prefrontal cortex, and the functional connectivity analysis uses the Pearson correlation coefficient method to calculate the whole-brain functional connectivity strength.

4. The evaluation system for the treatment effect of refractory depression based on transcranial magnetic stimulation according to claim 1, characterized in that: The cycle treatment module specifically adopts Stanford Accelerated Intelligent Neuromodulation Therapy, including: an alternating stimulation unit for alternately administering 5 accelerated intermittent theta burst stimulations and 5 accelerated continuous theta burst stimulations per day; Parameter setting unit, used to control the stimulation parameters to 50 Hz triplet pulse sequence, with a total pulse number of 18,000 times / day; Intensity control unit, used to adjust the stimulation intensity to 80% of the resting motor threshold.

5. The evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression according to claim 4, characterized in that: The accelerated intermittent theta burst stimulation lasted for 10 minutes each time, and the accelerated continuous theta burst stimulation lasted for 2 minutes each time. The continuous treatment cycle was 5 days, and the total number of pulses reached 90,000 times.

6. The evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression according to claim 1, characterized in that: The data acquisition and processing module includes: EEG acquisition unit, used to collect 5-10 minutes of eyes-closed resting state data using a 128-lead EEG cap at a sampling rate of 1000 Hz; a near-infrared acquisition unit to monitor changes in hemoglobin dynamics in the prefrontal-parietal network; The feature extraction unit is used to calculate the EEG power spectral density, phase locking value, hemoglobin concentration mean and functional connectivity index based on the collected resting state data and hemoglobin dynamic changes.

7. The evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression according to claim 6, characterized in that: The feature extraction unit specifically performs: EEG feature extraction: calculation of relative power and phase lock value of δ, θ, α, β, and γ frequency bands; Near-infrared feature extraction: Calculate the slope of change in prefrontal oxyhemoglobin concentration and the global efficiency index of the network functional connectivity matrix; Cross-modal feature fusion: spatiotemporal coupling analysis of alpha wave power and the slope of prefrontal hemoglobin concentration changes.

8. The evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression according to claim 1, characterized in that: The treatment assessment module includes: A model training unit, used for training a support vector machine classification model and a support vector machine regression model by using a leave-one-out cross-validation method; A feature selection unit is used to screen baseline brain function features with significant differences as model inputs through T-test; The effect verification unit is used to calculate the area under the ROC curve and the correlation coefficient for symptom improvement prediction.

9. The evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression according to claim 8, characterized in that: The support vector machine classification model uses a linear kernel function with a penalty coefficient C=10, and the input features include the baseline θ / α ratio, the slope of the prefrontal hemoglobin concentration, and the HAMD score corresponding to the clinical scale; The support vector machine regression model adopts the RBF kernel function, the penalty coefficient C=15, and the output is the continuous prediction value of the symptom improvement rate.

10. The evaluation system for the therapeutic effect of transcranial magnetic stimulation for refractory depression according to claim 1, characterized in that: The remote effect principle is specifically implemented as follows: By stimulating the left DLPFC target, the neural activity of sgACC is indirectly regulated through the structural-functional connection pathway; Targets were screened based on the functional connectivity strength using the Pearson correlation coefficient r≥0.6.

Images