Trauma treatment system based on hrv monitoring and transcutaneous auricular vagus nerve stimulation

Abstract

This invention relates to the field of computer-aided diagnostic technology, and specifically discloses a psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation. The system eliminates the interference of time overlap on physiological indicators through HRV rhythm correction, eliminates the limitations of single indicator assessment through multi-dimensional data fusion, and achieves dynamic adaptation of intervention programs through subtyped taVNS treatment and closed-loop feedback, thereby reducing the bias in assessing the severity of psychological trauma and improving the accuracy of treatment.

Description

Technical Field

[0001] This invention relates to the field of computer-aided diagnostic technology, and in particular to a psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation. Background Technology

[0002] Psychological trauma is often accompanied by autonomic nervous system dysfunction and executive function impairment, severely reducing patients' quality of life. Transcutaneous vagus nerve stimulation (taVNS), as a non-invasive neuromodulation technique, has been proven to improve executive functions such as response inhibition and working memory in adults, with only mild, tolerable side effects such as mild tingling, making it a potential treatment for psychological trauma. For example, Chinese patent application CN104127193A describes a method that, by acquiring physiological information of the human body under different conditions and analyzing heart rate variability (HRV) characteristics under different conditions, assesses the balance of sympathetic and vagal nerve functions in the autonomic nervous system, establishes a quantitative assessment model for depression, and achieves rapid and objective assessment of the subject's depression level.

[0003] Among the aforementioned technical solutions, HRV monitoring and taVNS intervention in the treatment of psychological trauma have technical shortcomings. Since HRV (heart rate variability) is directly regulated by the body's diurnal rhythm, its core characteristics exhibit significant fluctuations in a 24-hour primary cycle and a 12-hour secondary cycle. From a physiological perspective, increased sympathetic nerve activity during the day leads to a higher proportion of low-frequency power in HRV and a lower standard deviation of the RR interval; at night, the parasympathetic nervous system dominates, resulting in a significant increase in high-frequency power in HRV and a greater dispersion of the RR interval. The amplitude of this rhythmic fluctuation is far greater than the HRV changes caused by stress. For example, in the study, the resting-state HRV low-frequency power was about 150 ms² at 8:30, and it rose to 220 ms² at 14:00 due to the peak of sympathetic nerve activity. However, the stress state at 8:30 only increased the low-frequency power to 180 ms². At this time, the difference in HRV between the resting-state at different times (70 ms²) far exceeded the difference between the resting-state and stress state at the same time (30 ms²), which directly caused the rhythmic fluctuation of HRV characteristics to overshadow the specific changes of the stress state.

[0004] Meanwhile, the rhythmic regulation of the human autonomic nervous system has a higher priority than acute stress. In the resting state, the rhythmic fluctuations of the autonomic nervous system drive a regular shift in HRV characteristics over time; however, stress only produces small fluctuations based on the rhythmic baseline of that time period. This results in overlapping and crossover of HRV characteristics between resting and stress states at different times. For example, the high-frequency power of stress-state HRV (reflecting parasympathetic activity) at 8:30 may be close to the high-frequency power of resting state at 14:00, and the mean resting-state RR interval at 22:30 may overlap with the mean stress-state interval at 8:30. This temporal overlap disrupts the correspondence between HRV characteristics and stress states, rendering the HRV threshold originally used to distinguish between resting and stress ineffective. If the HRV threshold of 8:30 is used to determine the state at 14:00, the normal resting state at 14:00 will be mistakenly identified as a stress state, and vice versa. Therefore, in the literature "Research on the Influence of HRV Rhythmality on Real-Time Stress Detection", a three-time-point Stroop color experiment was designed to collect 72 hours of HRV data from volunteers; anomalies were removed using a sliding window confidence ellipse, and 10-dimensional features in the time domain, frequency domain, and nonlinear domain were extracted; 24-hour / 12-hour cycles were identified by FFT, and the rhythmic trend was removed by smoothing priors, while retaining instantaneous stress fluctuations; the F-statistics before and after detrending were compared, and the XGBoost model was used for verification, which improved the difference in HRV features between the two states after detrending, improved the classification accuracy, and effectively distinguished the resting and stress state features that overlapped in different time periods. However, the study subjects were healthy volunteers without underlying pathological conditions such as psychological trauma, and their HRV rhythmicity was only a normal physiological rhythm. In contrast, clinical patients (such as those with psychological trauma) have autonomic dysfunction, and their HRV rhythm may be superimposed with pathological fluctuations. The baseline rhythms and stress response patterns of the two groups differed significantly. The stress in volunteers was induced in a short period of time by the Stroop color test, which is different from the long-term chronic stress state of patients, making it difficult to adapt the HRV stress fluctuation pattern to clinical patients. There was little individual variation among volunteers, while patients had differences in the severity of their condition and complications. A uniform detrending and classification model lacked adaptability to the individual pathological characteristics of patients.

[0005] In summary, due to the fundamental differences between healthy volunteers and clinical patients, when the differences in HRV characteristics between resting and stressed states at different time points exceed the differences in the states themselves and lead to temporal overlap in physiological data, it is easy to cause bias in the assessment of the severity of psychological trauma, making it difficult to accurately reflect the true stress state and reducing the accuracy of treatment. Therefore, there is an urgent need for a psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation that can improve the accuracy of treatment. Summary of the Invention

[0006] This invention provides a psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation, which can improve the accuracy of treatment.

[0007] To solve the above-mentioned technical problems, this application provides the following technical solution: A trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation includes: The data acquisition module is used to collect patient HRV data, behavioral data, and biomarker data; The data processing module is used to preprocess HRV data using the sliding window confidence ellipse algorithm to filter out abnormal fluctuations in the RR interval. Then, the preprocessed HRV data is analyzed in the frequency domain using fast Fourier transform to identify the rhythmic features of the 24-hour primary cycle and the 12-hour secondary cycle, and to obtain the rhythmic fluctuation function of HRV features in each dimension. Subsequently, a high-pass filtering model is constructed based on the smoothing prior method, with 24 hours as the cutoff period threshold, to remove the base trend term driven by the diurnal rhythm in the HRV data, retaining the instantaneous fluctuation component caused by stress, and obtaining the corrected HRV data. The timestamps of the corrected HRV data, behavioral data and biomarker data are simultaneously labeled for spatiotemporal alignment to obtain a standardized multidimensional dataset. The treatment plan evaluation module is used to retrieve standardized multidimensional datasets, perform noise reduction and feature extraction on the standardized multidimensional datasets to obtain a preprocessed feature set, and then input the preprocessed feature set into the HRV-cortisol coupling model for coupling correlation analysis to generate autonomic nervous system disorder classification data. Subsequently, the autonomic nervous system disorder classification data is substituted into a pre-trained mental trauma severity scoring model for quantitative calculation to obtain a trauma severity index that can characterize the condition. Based on the trauma severity index, a parameter matching algorithm is used to process the data to obtain a personalized taVNS treatment plan containing target potential / stimulation parameters. The execution feedback module is used to receive personalized taVNS plans, collect training completion and stimulus tolerance data, and send the training completion and stimulus tolerance data back to the plan evaluation module. The plan evaluation module is also used to modify the psychological trauma severity scoring model based on the training completion and stimulus tolerance data, and dynamically adjust the personalized taVNS treatment plan based on the training completion and stimulus tolerance data.

[0008] The basic principle and beneficial effects of this scheme are as follows: In this scheme, the abnormal fluctuations of the RR interval are first filtered out by the sliding window confidence ellipse algorithm. Then, the fast Fourier transform is used to accurately identify the rhythmic features of the 24-hour main cycle and the 12-hour secondary cycle in the HRV data, and obtain the rhythmic fluctuation function of the HRV features in each dimension. Subsequently, a high-pass filtering model is constructed based on the smoothing prior method. With 24 hours as the cutoff period threshold, the base trend term driven by the diurnal rhythm in the HRV data is stripped away, and only the instantaneous fluctuation component caused by stress is retained. Essentially, the physiological rhythm baseline is removed and stress-specific fluctuations are retained, eliminating the masking effect of autonomic nervous rhythmic regulation on HRV indicators at different times. This allows the corrected HRV data to accurately distinguish between resting and stress states at different times, avoiding misjudging the natural activation of the sympathetic nervous system as traumatic disorder or misjudging the nocturnal parasympathetic inhibition as normal rhythm.

[0009] Because uncorrected HRV data is out of sync with behavioral and biomarker data, such as the high-frequency power increase in resting HRV at night being easily confused with the increase in salivary cortisol under stress, this approach completely eliminates false positive or false negative biases in HRV indicators caused by rhythm interference by aligning the three types of data in time and space and combining cross-validation of behavioral and biomarker data. This allows the trauma severity index to truly reflect the patient's stress state. By precisely aligning calibrated HRV data with behavioral and biomarker data across time and context dimensions, a standardized multidimensional dataset is formed. The preprocessed feature set is first input into the HRV-cortisol coupling model to generate autonomic nervous system disorder classification data. This data is then integrated into a trauma severity scoring model to calculate an index, achieving multidimensional cross-validation of physiological indicators, behavioral performance, and biomarkers. Based on the trauma severity index, a personalized taVNS plan targeting the concha / cymba concha is generated using a parameter matching algorithm. Training completion and stimulation tolerance are fed back to dynamically adjust parameters such as taVNS stimulation intensity and pulse width, allowing the assessment model and treatment plan to iterate dynamically with the patient's condition. This avoids intervention inaccuracies caused by initial assessment biases and improves the precision and individualization of trauma treatment.

[0010] In summary, this approach eliminates the interference of time overlap on physiological indicators through HRV rhythm correction, overcomes the limitations of single-indicator assessment through multi-dimensional data fusion, and achieves dynamic adaptation of the intervention plan through subtyped taVNS treatment and closed-loop feedback, thereby reducing the bias in assessing the severity of psychological trauma and improving the accuracy of treatment.

[0011] Furthermore, the data processing module preprocesses the HRV data using a sliding window confidence ellipse algorithm to filter out abnormal fluctuations in the RR interval, including: Convert the RR interval sequences of the original HRV data into two-dimensional data pairs: (RR1, RR2), (RR2, RR3)…(RR n-1 ,RR n ); Among them, RR i Let be the i-th RR interval value; set the window size and window crossover number of the sliding window, and perform crossover sliding on the two-dimensional data sequence; calculate the covariance matrix of the two-dimensional data in each sliding window, and obtain the maximum eigenvalue λ1, the minimum eigenvalue λ2, and the data center (R). 1ρ R 2ρ Based on the pre-set chi-square distribution threshold corresponding to the confidence level, a confidence ellipse equation is constructed. [(xR 1ρ )² / λ1+(yR 2ρ [)² / λ²] = Chi-square distribution threshold; Two-dimensional data points located outside the confidence ellipse within the sliding window are identified as abnormal fluctuation points for their corresponding RR intervals and filtered out. This process is repeated until all HRV data are covered.

[0012] The beneficial effects are as follows: The algorithm combining sliding windows and confidence ellipses, by converting RR intervals into adjacent two-dimensional data pairs, focuses on the correlation between preceding and following heart rate intervals. Outliers caused by limb activity or equipment errors are often abrupt changes that deviate from the normal data distribution and clustering, while rhythmic fluctuations are continuous and regular changes that fall within the confidence ellipse. Targeted removal of abrupt RR interval changes caused by limb activity or equipment errors can distinguish between rhythmic fluctuations and abnormal fluctuations. The cross-sliding window achieves full data coverage detection, and the statistical boundary constructed by the chi-square distribution threshold can accurately screen out abrupt changes, reducing HRV data dispersion. The algorithm only removes isolated outliers outside the ellipse, without affecting the continuous rhythm-related data within the ellipse, fully preserving the 24-hour / 12-hour periodic fluctuation pattern of HRV and the small fluctuations caused by stress, avoiding the loss of confidence due to outliers. The signal distortion provides high-quality data for Fast Fourier Transform (FFT), ensuring the accuracy of period identification and preventing rhythmic features from being masked by outliers. This allows the processed data to effectively retain the true rhythmic features and stress-related fluctuations of HRV, avoiding interference from outliers in the FFT's identification of 24-hour / 12-hour periods. The confidence ellipse is constructed based on statistical parameters such as the covariance matrix and eigenvalues. Anomaly detection relies on the data's own distribution characteristics rather than manually set thresholds. The pre-set chi-square distribution threshold corresponding to the confidence level is statistically significant, ensuring a unified and objective anomaly detection standard. This avoids judgment bias caused by subjective experience and improves the reliability of the preprocessing results. The sliding window crossover enables full coverage detection of HRV data. The statistical characteristics of the confidence ellipse ensure the objectivity of anomaly detection and avoid subjective errors caused by manually set thresholds.

[0013] Furthermore, the data processing module preprocesses the HRV data using a sliding window confidence elliptic algorithm to filter out abnormal fluctuations in the RR interval. Then, it performs frequency domain analysis on the preprocessed HRV data using a fast Fourier transform to identify the rhythmic characteristics of the 24-hour primary cycle and the 12-hour secondary cycle, obtaining the rhythmic fluctuation functions of HRV characteristics in each dimension, including: The preprocessed HRV time-domain, frequency-domain, and nonlinear feature data are padded with zeros to lengths that are powers of 2. The time-domain signals of each HRV feature are converted into frequency-domain signals using Fast Fourier Transform, and the power spectral density corresponding to each frequency component is calculated. The frequency components with the highest and second-highest amplitudes in the power spectral density are selected. The frequency component with a period of 24 hours is determined to be the 24-hour main period, and the frequency component with a period of 12 hours is determined to be the 12-hour secondary period. Based on the least squares method, a second-order sine function is used. ; in, It is a constant. For amplitude, The angular frequency corresponding to the 24-hour main cycle is used to fit the time series data of HRV features in each dimension, and the rhythmic fluctuation function of HRV features in each dimension is obtained.

[0014] The beneficial effects are as follows: Compared to identifying the 24-hour / 12-hour cycle of HRV solely through FFT without zero-padding the preprocessed data and directly fitting the rhythm function based on the population data without processing each dimension of HRV characteristics (time domain, frequency domain, nonlinearity), this scheme first improves the FFT resolution by zero-padding, then selects the top two frequency components of the power spectral density according to the dimension to lock the cycle, and finally accurately fits the rhythm fluctuation function of each dimension through a second-order sine function, improving the cycle identification accuracy and avoiding the lack of dimension specificity in rhythm representation; zero-padding to an integer power of 2 eliminates the limitation of data length on the FFT frequency domain resolution, making the frequency component calculation more accurate, and the power spectral density amplitude sorting can selectively extract the rhythmic frequency with the highest proportion, eliminating the low-amplitude interference of random fluctuations; the second-order sine function fitting, combined with the main cycle and secondary cycle frequencies, perfectly matches the diurnal rhythm variation pattern of HRV characteristics, achieving accurate differentiation between rhythmic cycles and random fluctuations, providing a clear target for subsequent rhythm stripping.

[0015] By using FFT frequency domain analysis and power spectral density analysis, the 24-hour primary cycle and 12-hour secondary cycle of HRV characteristics in various dimensions can be accurately identified, distinguishing between rhythmic cycles and random fluctuations, providing a clear target for subsequent rhythm stripping. Based on the fitting of the second-order sine function, the diurnal rhythmic variation pattern of HRV characteristics can be fully matched, and the obtained rhythmic fluctuation function can accurately characterize the trend of rhythm's influence on HRV. The clear rhythmic cycle and accurate fluctuation function enable the smoothing prior method to selectively strip the rhythmic trend term, retaining stress-specific fluctuation components to the greatest extent, and effectively avoiding evaluation bias caused by time overlap.

[0016] Furthermore, the data processing module constructs a high-pass filtering model based on the smoothing prior method, using a 24-hour cutoff period threshold to remove the base trend term driven by the diurnal rhythm from the HRV data, retaining the instantaneous fluctuation component caused by stress, to obtain corrected HRV data, including: The preprocessed and FFT-analyzed HRV data Decomposed into stationary terms and cyclical trend items , ,in, Indicates stress-related transient fluctuations. This indicates that the circadian rhythm drives the base trend; Using a 24-hour cutoff period as the threshold, the corresponding angular frequency is calculated as the filter cutoff frequency, and a second-order discrete differential operator matrix is ​​constructed. ; Introducing regularization parameters By optimizing the objective function ; Where H is the unit observation matrix, For regression parameters, It is the preprocessed and FFT-analyzed raw HRV data, which includes two parts: a base trend term driven by circadian rhythms and a transient fluctuation component caused by stress. This can be understood as being based on parameters. The predicted HRV data obtained; These are the optimized regression parameter estimates, with subscripts... This indicates that the parameter is determined by the regularization parameter. Regulation; This indicates the search for parameters that minimize the objective function within the curly braces. That is, the regression parameters Perform a minimization solution; The residual sum of squares term measures "parameter-based" terms. The predicted values ​​obtained "and raw HRV data" The smaller the fitting error between the two values, the closer the predicted value is to the original data; To smooth the penalty term, measure the predicted value. The smaller the smoothness, the more gentle the fluctuation of the trend term; Solve for the estimated value of the trend term. ,in, It is an estimate of the base trend term driven by the diurnal rhythm in HRV data. These are the estimated regression parameters obtained by solving the objective function through prior optimization. Subtract the trend term estimate from the original HRV data. This yielded corrected HRV data that retained only the transient fluctuations caused by stress. ;in, This is the final corrected HRV data. It is an estimate of the base trend term driven by the diurnal rhythm.

[0017] The beneficial effects are as follows: the high-pass filtering model constructed by the smoothing prior method can specifically filter out the rhythmic base trend of the 24-hour main cycle and the 12-hour secondary cycle, avoiding rhythmic fluctuations from masking stress-specific changes and preventing the temporal overlap of HRV characteristics in different time periods; the combination of regularization parameters and second-order differential operators, while stripping away rhythmic trends, retains the transient fluctuation components of HRV related to psychological trauma to the greatest extent, preserving the true stress signal, so that the corrected HRV data can truly reflect the patient's stress state; the corrected HRV data eliminates the assessment bias caused by rhythmic interference, making the calculation of the psychological trauma severity index and the trauma subtype classification more consistent with the patient's true condition, ensuring the targeting of personalized taVNS treatment plans and improving treatment accuracy.

[0018] Furthermore, the data processing module synchronously labels and corrects the timestamps of HRV data, behavioral data, and biomarker data for spatiotemporal alignment, resulting in a standardized multidimensional dataset, including: A timestamp generation unit is configured for each of the corrected HRV data, behavioral data, and biomarker data to reduce the timestamp error of the corrected HRV data, behavioral data, and biomarker data to within a preset error threshold. The corrected HRV data, behavioral data, and biomarker data are segmented according to time series. Sliding windows are aligned according to preset time periods, and the corrected HRV feature statistics, cumulative behavioral data values, and instantaneous biomarker data values ​​within each sliding window are extracted. The Dynamic Time Warping (DTW) algorithm corrects time misalignment caused by data acquisition delays and removes isolated data points that exceed the sliding window time range. The aligned and corrected HRV data, behavioral data, and biomarker data are then structured and stored according to the sliding window-data type-feature parameter structure to generate a standardized multidimensional dataset, where each sliding window corresponds to a complete set of multidimensional data.

[0019] The beneficial effects are as follows: By using millisecond-level time synchronization and dynamic time warping algorithms, the time misalignment of data collected by different devices is eliminated, enabling the corrected HRV data to form a corresponding spatiotemporal correlation with behavioral data and biomarker data, avoiding assessment bias caused by asynchronous data; the division of time alignment windows and the removal of isolated data filter out random errors in the data collection process, ensuring that the dataset at each time point can comprehensively reflect the patient's physiological, behavioral, and biological overall state; integrating heterogeneous data into a structured format can efficiently extract coupled features and improve the accuracy of calculating the severity index of mental trauma and classifying subtypes.

[0020] Furthermore, the scheme evaluation module inputs the preprocessed feature set into the HRV-cortisol coupling model for coupling correlation analysis, generating autonomic nervous system disorder classification data, including: The core features of corrected HRV and salivary cortisol concentration features were extracted from the preprocessed feature set and normalized to the [0, 1] interval respectively. A dual-input, single-output HRV-cortisol coupling model was constructed, with the core features of the corrected HRV as the first input vector and the salivary cortisol concentration features as the second input vector. The temporal coupling strength of the two types of features was calculated by Pearson correlation coefficient, and the frequency domain co-fluctuation coefficient was calculated by cross-power spectral density analysis. By setting a coupling strength threshold and a synergistic fluctuation coefficient threshold, and combining the coupling strength and synergistic fluctuation coefficient, autonomic nervous system disorder classification data are generated. Strong coupling + high frequency synergistic corresponds to sympathetic hyperactivity, weak coupling + low frequency synergistic corresponds to parasympathetic inhibition, and moderate coupling + mixed synergistic corresponds to autonomic nervous system balance disorder.

[0021] The beneficial effects are as follows: through cross-dimensional coupling analysis of HRV and cortisol, the three subtypes of sympathetic hyperactivity, parasympathetic inhibition and balance disorder can be accurately distinguished, preventing insufficient treatment targeting caused by classification ambiguity; normalization eliminates the difference in characteristic dimensions, and the dual judgment criteria of coupling strength and synergistic fluctuation coefficient avoid classification errors caused by fluctuation of a single indicator.

[0022] Furthermore, the scheme evaluation module inputs the autonomic nervous system disorder classification data into a pre-trained trauma severity scoring model for quantitative calculation, obtaining a trauma severity index that characterizes the condition and corresponding trauma subtype data, including: The autonomic nervous system disorder classification data were converted into classification feature vectors, with sympathetic hyperactivity type corresponding to [1,0,0], parasympathetic inhibition type corresponding to [0,1,0], and autonomic nervous system balance disorder type corresponding to [0,0,1]. Extract the corrected HRV feature quantization value, normalized salivary cortisol concentration value, and behavioral data anomaly clustering score from the standardized multidimensional dataset, and concatenate them with the genotyping feature vector to form the model input vector; The trauma severity scoring model is a random forest model pre-trained using clinical data from trauma patients. The trauma severity index is calculated by substituting the input vector into the model and using decision tree ensemble voting. By combining the trauma severity index with the autonomic nervous system disorder classification, corresponding trauma subtype data are generated.

[0023] The beneficial effects are as follows: the random forest model trained on a large sample has strong fitting ability and anti-interference ability; by combining the input vector of subtyping features and multi-dimensional data, the calculation error of the trauma severity index is reduced; and by combining the two dimensions of severity level and autonomic nervous system disorder subtyping, the subtypes can be accurately classified.

[0024] Furthermore, the program evaluation module modifies the psychological trauma severity scoring model based on training completion and stimulus tolerance, and dynamically adjusts the personalized taVNS treatment plan based on training completion and stimulus tolerance, including: The training completion rate and stimulus tolerance are quantified and encoded, and mapped to feature weight correction coefficient and parameter adjustment coefficient, respectively. The decision tree feature weights of the trauma severity rating model are updated based on the correction coefficient. Among them, the training completion degree corresponds to the behavioral feature weights and the stimulus tolerance corresponds to the physiological feedback feature weights. After the update, the trauma severity index output by the model is recalculated. Dynamically adjust the personalized taVNS protocol: If the training completion rate is excellent and the tolerance is good, increase the stimulation intensity and prolong the stimulation duration; if the training completion rate is poor or the tolerance is poor, decrease the stimulation intensity and switch to a 30s on / 60s off cycle; if the training completion rate and tolerance are both average, keep the core parameters unchanged and optimize the pulse width; after adjustment, a new personalized taVNS treatment protocol is generated.

[0025] The beneficial effects are as follows: Using training completion and stimulus tolerance during treatment as real-time feedback signals corrects the model's feature weights, preventing a disconnect between the initial assessment and the patient's actual treatment response, reducing the update error of the trauma severity index, and improving assessment accuracy. Dynamic parameter adjustment based on feedback data ensures treatment effectiveness while reducing the risk of adverse reactions, thus improving the adaptability of the taVNS protocol.

[0026] Furthermore, the parameter matching algorithm of the personalized taVNS treatment plan dynamically matches stimulation parameters, including: for the subtype of response inhibition dysfunction, matching the concha cymba target point, first stimulation frequency, first stimulation intensity, first on / off cycle, and first pulse width; for the subtype of working memory dysfunction, matching the concha target point, second stimulation frequency, second stimulation intensity, continuous stimulation mode, and second pulse width. The stimulation parameters are dynamically bound to features in the corrected HRV data that reflect the correlation between response inhibition and working memory.

[0027] The beneficial effects are as follows: Since psychological trauma is often accompanied by executive function impairments such as response inhibition and working memory, targeted matching of taVNS parameters can improve the cognitive function targeting of taVNS parameters by utilizing the physiological mechanisms of taVNS in regulating frontal lobe nerve oscillations and enhancing frontotemporal lobe functional connectivity and implementing targeted interventions. By binding the features related to response inhibition / working memory in the corrected HRV with taVNS parameters, it can not only adapt to autonomic nervous system disorders but also accurately correspond to cognitive function impairments, reducing assessment bias and improving the accuracy of intervention.

[0028] Furthermore, the execution feedback module also includes a heart rate oscillation guidance unit and an electrode connection detection unit. The heart rate oscillation guidance unit provides at least one heart rate oscillation guidance mode and outputs guidance information to guide the patient to adjust their behavior. The on / off time of the guidance information is adapted to the on / off time of the personalized taVNS treatment plan. If the taVNS on / off time is less than twice the on / off time of the guidance mode, the patient is prompted to select one of the on / off times as the execution parameter. The electrode connection detection unit outputs a constant intensity detection current to the stimulation electrode and obtains the resistance value at both ends of the stimulation electrode. If the resistance value exceeds the preset range, an electrode connection abnormality prompt is output and the taVNS intervention is stopped. The intervention is resumed after the resistance value meets the preset range. The execution feedback module stops the taVNS intervention during the patient's inhalation phase and / or muscle contraction phase, and executes the personalized taVNS treatment plan during the exhalation phase and / or muscle extension phase.

[0029] The beneficial effects are as follows: by stimulating the patient's vagus nerve activity through the heart rate oscillation guidance mode, the taVNS intervention is carried out under the heart rate oscillation state, avoiding the stimulation effect being offset by the body's own inhibitory effect, and enhancing the positive regulatory effect of taVNS on response inhibition and working memory; the electrode connection detection unit monitors the electrode contact status in real time, avoiding the inability of the stimulation current to act on the vagus nerve due to abnormal skin impedance or the occurrence of safety risks, and combines the dynamic start and stop of stimulation with the respiratory and muscle activity status, so that taVNS is precisely coupled with the patient's physiological behavior, reducing ineffective stimulation and improving the targeting of intervention. Attached Figure Description

[0030] Figure 1This is a system block diagram of Example 1 of a trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation. Detailed Implementation

[0031] The following detailed description illustrates the specific implementation method: Example 1

[0032] This invention relates to a psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation, as shown in the attached document. Figure 1 As shown, the specific implementation process is as follows: The data acquisition module collects patient HRV data, behavioral data, and biomarker data. For example, wearable physiological sensors (such as chest strap heart rate monitors) collect electrocardiogram (ECG) or photoplethysmography (PPG) signals and continuously record heart rate interval (RR interval) data; wearable motion sensors (such as triaxial accelerometers and gyroscopes) capture patients' limb movements (muscle contraction / relaxation rhythms) in real time; saliva collection devices (such as saliva swabs) allow patients to collect saliva samples at fixed time points, and microfluidic chip detection technology is used to analyze cortisol concentration in real time; simultaneously, wearable device skin conductivity sensors assist in verifying stress states. The data is coupled with HRV and behavioral data according to time windows to form a standardized multidimensional dataset.

[0033] The data processing module preprocesses the HRV data using a sliding window confidence ellipse algorithm to filter out abnormal fluctuations in the RR intervals. It then converts the RR interval sequence of the original HRV data into two-dimensional data pairs (RR1, RR2), (RR2, RR3)...(RR...). n-1 ,RR n ), where RR i Let be the i-th RR interval value; set the window size and window crossover number of the sliding window, and perform crossover sliding on the two-dimensional data sequence; calculate the covariance matrix of the two-dimensional data in each sliding window, and obtain the maximum eigenvalue λ1, the minimum eigenvalue λ2, and the data center (R). 1ρ R 2ρ Based on the pre-set chi-square distribution threshold corresponding to the confidence level, a confidence ellipse equation [(xR) is constructed. 1ρ )² / λ1+(yR 2ρ [λ² / λ²] = Chi-square distribution threshold; Two-dimensional data outside the confidence ellipse within the sliding window are identified as abnormal fluctuation points for their corresponding RR intervals and filtered out. This process is repeated until all HRV data are covered. Removing abrupt changes in RR intervals caused by limb activity and equipment errors can distinguish between rhythmic fluctuations and abnormal fluctuations, reducing HRV data dispersion. It effectively preserves the true rhythmic characteristics and stress-related fluctuations of HRV, avoiding interference from outliers in the Fast Fourier Transform's identification of the 24-hour / 12-hour cycle.

[0034] The preprocessed HRV data is then analyzed in the frequency domain using Fast Fourier Transform (FFT) to identify the rhythmic characteristics of the 24-hour principal period and the 12-hour secondary period, obtaining the rhythmic fluctuation functions of each dimension of HRV features. The preprocessed HRV time-domain, frequency-domain, and nonlinear feature data are padded with zeros to an integer power of 2. FFT is then used to convert the time-domain signals of each dimension of HRV features into frequency-domain signals, and the power spectral density corresponding to each frequency component is calculated. The frequency components with the two largest amplitudes in the power spectral density are selected; the frequency component with a period of 24 hours is identified as the 24-hour principal period, and the frequency component with a period of 12 hours is identified as the 12-hour secondary period. Based on the least squares method, a second-order sine function is used... ; in, It is a constant. For amplitude, By fitting the time-series data of HRV features across various dimensions to the angular frequency corresponding to the 24-hour primary cycle, the rhythmic fluctuation functions of each HRV feature are obtained. This method can accurately pinpoint the 24-hour primary cycle and 12-hour secondary cycle of each HRV feature dimension, distinguishing between rhythmic cycles and random fluctuations. It can fully fit the diurnal rhythmic variation pattern of HRV features, and the obtained rhythmic fluctuation functions can accurately characterize the influence trend of rhythm on HRV. This allows the smoothing prior method to selectively remove the rhythmic trend term, retaining stress-specific fluctuation components to the greatest extent and effectively avoiding assessment bias caused by time overlap.

[0035] Subsequently, a high-pass filtering model was constructed based on the smoothing prior method. Using a 24-hour cutoff period threshold, the base trend term driven by diurnal rhythms was removed from the HRV data, retaining the transient fluctuation component caused by stress, thus obtaining corrected HRV data. The preprocessed and FFT-analyzed HRV data was then analyzed... Decomposed into stationary terms and cyclical trend items , ,in, Indicates stress-related transient fluctuations. The circadian rhythm drives the base trend; using a 24-hour cutoff period as the threshold, the corresponding angular frequency is calculated as the filter cutoff frequency, and a second-order discrete differential operator matrix is ​​constructed. ; Introducing regularization parameters By optimizing the objective function ; Where H is the unit observation matrix, For the regression parameters, solve for the estimated value of the trend term. Subtract the trend term estimate from the original HRV data. This yielded corrected HRV data that retained only the transient fluctuations caused by stress. The 24-hour primary cycle and 12-hour secondary cycle rhythmic base trends are filtered out to avoid rhythmic fluctuations masking stress-specific changes and to prevent temporal overlap of HRV characteristics across different time periods. While stripping away rhythmic trends, the transient fluctuation components of HRV related to psychological trauma are preserved to the greatest extent possible, retaining true stress signals so that the corrected HRV data can accurately reflect the patient's stress state. The corrected HRV data eliminates assessment bias caused by rhythmic interference, making the calculation of the psychological trauma severity index and the trauma subtype classification more consistent with the patient's actual condition, ensuring the targeting of personalized taVNS treatment plans and improving treatment accuracy.

[0036] The timestamps of the synchronized HRV data, behavioral data, and biomarker data are spatiotemporally aligned to obtain a standardized multidimensional dataset. A timestamp generation unit is configured for each of the synchronized HRV data, behavioral data, and biomarker data to reduce the timestamp errors of these data to within a preset error threshold. The corrected HRV data, behavioral data, and biomarker data are segmented according to time series. Sliding windows are aligned according to preset time periods, and the corrected HRV feature statistics, cumulative behavioral data values, and instantaneous biomarker values ​​within each sliding window are extracted. The Dynamic Time Warping (DTW) algorithm corrects for time misalignment caused by data acquisition delays and removes isolated data points outside the sliding window time range. The aligned corrected HRV, behavioral, and biomarker data are then structured and stored according to a sliding window-data type-feature parameter structure, generating a standardized multidimensional dataset. Each sliding window corresponds to a complete set of multidimensional data. This process eliminates time misalignment between data collected from different devices, establishing a corresponding spatiotemporal correlation between corrected HRV data, behavioral data, and biomarker data, avoiding assessment bias caused by asynchronous data. It also filters out random errors during data acquisition, ensuring that the dataset at each time point comprehensively reflects the patient's physiological, behavioral, and biological overall status.

[0037] The scheme evaluation module retrieves a standardized multidimensional dataset, performs noise reduction and feature extraction on the standardized multidimensional dataset to obtain a preprocessed feature set, and then inputs the preprocessed feature set into the HRV-cortisol coupling model for coupling correlation analysis to generate autonomic nervous system disorder classification data. Corrected HRV core features and salivary cortisol concentration features are extracted from the preprocessed feature set and normalized to the [0, 1] interval. A dual-input, single-output HRV-cortisol coupling model is constructed, with the corrected HRV core features as the first input vector and the salivary cortisol concentration features as the second input vector. The temporal coupling strength of the two types of features is calculated using the Pearson correlation coefficient, and the frequency domain co-fluctuation coefficient is calculated using cross-power spectral density analysis. Coupling strength thresholds and co-fluctuation coefficient thresholds are set, and combined with the coupling strength and co-fluctuation coefficient, autonomic nervous system disorder classification data is generated. Strong coupling + high-frequency co-fluctuation corresponds to sympathetic hyperactivity, weak coupling + low-frequency co-fluctuation corresponds to parasympathetic inhibition, and moderate coupling + mixed co-fluctuation corresponds to autonomic nervous system balance disorder.

[0038] Subsequently, the autonomic nervous system disorder classification data were substituted into a pre-trained trauma severity scoring model for quantification, yielding a trauma severity index and corresponding trauma subtype data that characterize the condition. The autonomic nervous system disorder classification data were converted into classification feature vectors: sympathetic hyperactivity corresponds to [1,0,0], parasympathetic inhibition corresponds to [0,1,0], and autonomic nervous system balance disorder corresponds to [0,0,1]. Corrected HRV feature quantification values, normalized salivary cortisol concentration values, and abnormal clustering scores of behavioral data were extracted from the standardized multidimensional dataset and concatenated with the classification feature vectors to form the model input vector. The trauma severity scoring model was a random forest model pre-trained using clinical data from trauma patients. The input vector was substituted into the model, and the trauma severity index was calculated through decision tree ensemble voting. Combining the trauma severity index with the autonomic nervous system disorder classification, corresponding trauma subtype data were generated. Random forest models trained on large samples have strong fitting ability and anti-interference ability. By combining subtyping features with input vectors of multi-dimensional data, the calculation error of the trauma severity index is reduced. Through the two-dimensional combination of severity level and autonomic nervous system disorder subtyping, subtypes can be accurately classified.

[0039] Based on the trauma severity index and corresponding trauma subtype data, a parameter matching algorithm is used to match the target point of the concha, the first stimulation frequency, the first stimulation intensity, the first on / off cycle, and the first pulse width for the subtype with response inhibition dysfunction; and the target point of the concha, the second stimulation frequency, the second stimulation intensity, the continuous stimulation pattern, and the second pulse width for the subtype with working memory dysfunction. The stimulation parameters are dynamically bound to the features reflecting the correlation between response inhibition and working memory in the corrected HRV data to obtain a personalized taVNS treatment plan containing target potential / stimulation parameters, as shown in Table 1.

[0040] Table 1. Parameter Matching Table for Personalized TaVNS Treatment Plans

[0041] The execution feedback module receives the personalized taVNS plan, collects training completion and stimulus tolerance data, and sends these data back to the plan evaluation module. For example, the wearable device's three-axis accelerometer and gyroscope capture the patient's limb movements (such as muscle contraction / relaxation amplitude) and respiratory rhythm in real time during treatment. This data is compared with the movement specifications (such as rhythm frequency and movement amplitude thresholds) in the preset treatment plan to calculate the movement compliance rate. Combined with the treatment duration and rhythm consistency score, a weighted quantification is used to obtain a training completion rate of 0-100%. Through the touchscreen and voice interaction module of the stimulation output device, patients can provide feedback on stimulus discomfort using a 1-5 level scale during and after treatment. Simultaneously, the wearable device collects skin conductance signals and heart rate fluctuation data (such as the magnitude of sudden heart rate increases during stimulation) to assist in verification, correct subjective scoring biases, and determine the final stimulus tolerance.

[0042] The treatment plan evaluation module modifies the trauma severity scoring model based on training completion and stimulus tolerance, and dynamically adjusts the personalized taVNS treatment plan accordingly. Training completion and stimulus tolerance are quantified and encoded, mapped to feature weight correction coefficients and parameter adjustment coefficients, respectively. Based on these correction coefficients, the decision tree feature weights of the trauma severity scoring model are updated. Training completion corresponds to behavioral feature weights, and stimulus tolerance corresponds to physiological feedback feature weights. After the update, the trauma severity index output by the model is recalculated. If training completion is excellent and tolerance is good, the stimulus intensity is increased and the stimulus duration is prolonged. If training completion is poor or tolerance is inadequate, the stimulus intensity is decreased and a 30s on / 60s off on / off cycle is switched. If both training completion and tolerance are average, the core parameters remain unchanged, and the pulse width is optimized. A new personalized taVNS treatment plan is generated after these adjustments.

[0043] Table 2 Comparison of TaVNS Treatment Protocol Compatibility Before and After Adjustment

[0044] As shown in Table 2, the training completion rate and stimulus tolerance during the treatment process are used as real-time feedback signals to correct the model feature weights, avoiding a disconnect between the initial assessment and the patient's actual treatment response, reducing the update error of the trauma severity index, and improving the accuracy of the assessment. Dynamic parameter adjustment based on feedback data ensures treatment effectiveness while reducing the risk of adverse reactions, thus improving the adaptability of the taVNS protocol. Example 2

[0045] The only difference from Embodiment 1 is that the execution feedback module further includes a heart rate oscillation guidance unit and an electrode connection detection unit. The heart rate oscillation guidance unit is used to provide at least one heart rate oscillation guidance mode and output guidance information to guide the patient to adjust their behavior. The on / off time of the guidance information is adapted to the on / off time of the personalized taVNS treatment plan. If the taVNS on / off time is less than twice the on / off time of the guidance mode, the patient is prompted to select one of the on / off times as the execution parameter. The electrode connection detection unit is used to output a constant intensity detection current to the stimulation electrode and obtain the resistance value at both ends of the stimulation electrode. If the resistance value exceeds the preset range, an electrode connection abnormality prompt message is output and the taVNS intervention is stopped. The intervention is resumed after the resistance value meets the preset range. The execution feedback module stops the taVNS intervention during the patient's inhalation phase and / or muscle contraction phase, and executes the personalized taVNS treatment plan during the exhalation phase and / or muscle extension phase. By using a heart rate oscillation-guided mode to stimulate the patient's vagal nerve activity, taVNS intervention is implemented under heart rate oscillation conditions, avoiding the stimulation effect being offset by the body's own inhibitory effect, and enhancing the positive regulatory effect of taVNS on response inhibition and working memory. The electrode connection detection unit monitors the electrode contact status in real time to avoid the stimulation current being unable to effectively act on the vagus nerve due to abnormal skin impedance or to avoid safety risks. Combined with the dynamic start and stop of stimulation based on respiratory and muscle activity status, taVNS is precisely coupled with the patient's physiological behavior, reducing ineffective stimulation and improving the targeting of intervention.

[0046] The above are merely embodiments of the present invention. The invention is not limited to the fields covered by these embodiments. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation, characterized in that, include: The data acquisition module is used to collect patient HRV data, behavioral data, and biomarker data; The data processing module is used to preprocess HRV data using the sliding window confidence ellipse algorithm to filter out abnormal fluctuations in the RR interval. Then, the preprocessed HRV data is analyzed in the frequency domain using fast Fourier transform to identify the rhythmic features of the 24-hour primary cycle and the 12-hour secondary cycle, and to obtain the rhythmic fluctuation function of HRV features in each dimension. Subsequently, a high-pass filtering model is constructed based on the smoothing prior method, with 24 hours as the cutoff period threshold, to remove the base trend term driven by the diurnal rhythm in the HRV data, retaining the instantaneous fluctuation component caused by stress, and obtaining the corrected HRV data. The timestamps of the corrected HRV data, behavioral data and biomarker data are simultaneously labeled for spatiotemporal alignment to obtain a standardized multidimensional dataset. The treatment plan evaluation module is used to retrieve standardized multidimensional datasets, perform noise reduction and feature extraction on the standardized multidimensional datasets to obtain a preprocessed feature set, and then input the preprocessed feature set into the HRV-cortisol coupling model for coupling correlation analysis to generate autonomic nervous system disorder classification data. Subsequently, the autonomic nervous system disorder classification data is substituted into a pre-trained mental trauma severity scoring model for quantitative calculation to obtain a trauma severity index that can characterize the condition. Based on the trauma severity index, a parameter matching algorithm is used to process the data to obtain a personalized taVNS treatment plan containing target potential / stimulation parameters. The execution feedback module is used to receive personalized taVNS plans, collect training completion and stimulus tolerance data, and send the training completion and stimulus tolerance data back to the plan evaluation module. The plan evaluation module is also used to modify the psychological trauma severity scoring model based on the training completion and stimulus tolerance data, and dynamically adjust the personalized taVNS treatment plan based on the training completion and stimulus tolerance data.

2. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 1, characterized in that, The data processing module preprocesses the HRV data using a sliding window confidence ellipse algorithm to filter out abnormal fluctuations in the RR interval, including: The RR interval sequences of the original HRV data were converted into two-dimensional data pairs. (RR1,RR2)、(RR2,RR3)…(RR n-1 ,RR n ); Among them, RR i Let be the i-th RR interval value; set the window size and window crossover number of the sliding window, and perform crossover sliding on the two-dimensional data sequence; calculate the covariance matrix of the two-dimensional data in each sliding window, and obtain the maximum eigenvalue λ1, the minimum eigenvalue λ2, and the data center (R). 1ρ R 2ρ Based on the pre-set chi-square distribution threshold corresponding to the confidence level, a confidence ellipse equation is constructed. [(xR 1ρ )² / λ1+(yR 2ρ [)² / λ²] = Chi-square distribution threshold; Two-dimensional data points located outside the confidence ellipse within the sliding window are identified as abnormal fluctuation points for their corresponding RR intervals and filtered out. This process is repeated until all HRV data are covered.

3. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 2, characterized in that, The data processing module preprocesses the HRV data using a sliding window confidence elliptic algorithm to filter out abnormal fluctuations in the RR interval. Then, it performs frequency domain analysis on the preprocessed HRV data using a fast Fourier transform to identify the rhythmic characteristics of the 24-hour primary cycle and the 12-hour secondary cycle, obtaining the rhythmic fluctuation functions of HRV characteristics in various dimensions, including: The preprocessed HRV time-domain, frequency-domain, and nonlinear feature data are padded with zeros to lengths that are powers of 2. The time-domain signals of each HRV feature are converted into frequency-domain signals using Fast Fourier Transform, and the power spectral density corresponding to each frequency component is calculated. The frequency components with the highest and second-highest amplitudes in the power spectral density are selected. The frequency component with a period of 24 hours is determined to be the 24-hour main period, and the frequency component with a period of 12 hours is determined to be the 12-hour secondary period. Based on the least squares method, a second-order sine function is used. ； in, It is a constant. For amplitude, The angular frequency corresponding to the 24-hour main cycle is used to fit the time series data of HRV features in each dimension, so as to obtain the rhythmic fluctuation function of HRV features in each dimension.

4. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 3, characterized in that, The data processing module constructs a high-pass filtering model based on a smoothing prior method. Using a 24-hour cutoff period threshold, it removes the base trend term driven by diurnal rhythms from the HRV data, retaining the instantaneous fluctuation component caused by stress, to obtain corrected HRV data, including: The preprocessed and FFT-analyzed HRV data Decomposed into stationary terms and cyclical trend items , ,in, Indicates stress-related transient fluctuations. This indicates that the circadian rhythm drives the base trend; Using a 24-hour cutoff period as the threshold, the corresponding angular frequency is calculated as the filter cutoff frequency, and a second-order discrete differential operator matrix is ​​constructed. ; Introducing regularization parameters By optimizing the objective function ； Where H is the unit observation matrix, For the regression parameters, solve for the estimated value of the trend term. ; Subtract the trend term estimate from the original HRV data. This yielded corrected HRV data that retained only the transient fluctuations caused by stress. .

5. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 4, characterized in that, The data processing module synchronously labels and corrects the timestamps of HRV data, behavioral data, and biomarker data, performing spatiotemporal alignment to obtain a standardized multidimensional dataset, including: A timestamp generation unit is configured for each of the corrected HRV data, behavioral data, and biomarker data to reduce the timestamp error of the corrected HRV data, behavioral data, and biomarker data to within a preset error threshold. The corrected HRV data, behavioral data, and biomarker data are segmented according to time series. Sliding windows are aligned according to preset time periods, and the corrected HRV feature statistics, cumulative behavioral data values, and instantaneous biomarker data values ​​within each sliding window are extracted. The Dynamic Time Warping (DTW) algorithm corrects time misalignment caused by data acquisition delays and removes isolated data points that exceed the sliding window time range. The aligned and corrected HRV data, behavioral data, and biomarker data are then structured and stored according to the sliding window-data type-feature parameter structure to generate a standardized multidimensional dataset, where each sliding window corresponds to a complete set of multidimensional data.

6. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 5, characterized in that, The scheme evaluation module inputs the preprocessed feature set into the HRV-cortisol coupling model for coupling correlation analysis, generating autonomic nervous system disorder classification data, including: The core features of corrected HRV and salivary cortisol concentration features were extracted from the preprocessed feature set and normalized to the [0, 1] interval respectively. A dual-input, single-output HRV-cortisol coupling model was constructed, with the core features of the corrected HRV as the first input vector and the salivary cortisol concentration features as the second input vector. The temporal coupling strength of the two types of features was calculated by Pearson correlation coefficient, and the frequency domain co-fluctuation coefficient was calculated by cross-power spectral density analysis. By setting a coupling strength threshold and a synergistic fluctuation coefficient threshold, and combining the coupling strength and synergistic fluctuation coefficient, autonomic nervous system disorder classification data are generated. Strong coupling + high frequency synergistic corresponds to sympathetic hyperactivity, weak coupling + low frequency synergistic corresponds to parasympathetic inhibition, and moderate coupling + mixed synergistic corresponds to autonomic nervous system balance disorder.

7. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 6, characterized in that, The scheme evaluation module inputs the autonomic nervous system disorder classification data into a pre-trained trauma severity scoring model for quantitative calculation, obtaining a trauma severity index that characterizes the condition and corresponding trauma subtype data, including: The autonomic nervous system disorder classification data were converted into classification feature vectors, with sympathetic hyperactivity type corresponding to [1,0,0], parasympathetic inhibition type corresponding to [0,1,0], and autonomic nervous system balance disorder type corresponding to [0,0,1]. Extract the corrected HRV feature quantization value, normalized salivary cortisol concentration value, and behavioral data anomaly clustering score from the standardized multidimensional dataset, and concatenate them with the genotyping feature vector to form the model input vector; The trauma severity scoring model is a random forest model pre-trained using clinical data from trauma patients. The trauma severity index is calculated by substituting the input vector into the model and using decision tree ensemble voting. By combining the trauma severity index with the autonomic nervous system disorder classification, corresponding trauma subtype data are generated.

8. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 7, characterized in that, The program evaluation module modifies the psychological trauma severity scoring model based on training completion and stimulus tolerance, and dynamically adjusts the personalized taVNS treatment plan based on training completion and stimulus tolerance, including: The training completion rate and stimulus tolerance are quantified and encoded, and mapped to feature weight correction coefficient and parameter adjustment coefficient, respectively. The decision tree feature weights of the trauma severity rating model are updated based on the correction coefficient. Among them, the training completion degree corresponds to the behavioral feature weights and the stimulus tolerance corresponds to the physiological feedback feature weights. After the update, the trauma severity index output by the model is recalculated. Dynamically adjust the personalized taVNS protocol: If the training completion rate is excellent and the tolerance is good, increase the stimulation intensity and prolong the stimulation duration; if the training completion rate is poor or the tolerance is poor, decrease the stimulation intensity and switch to a 30s on / 60s off cycle; if the training completion rate and tolerance are both average, keep the core parameters unchanged and optimize the pulse width; after adjustment, a new personalized taVNS treatment protocol is generated.

9. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 8, characterized in that, The parameter matching algorithm of the personalized taVNS treatment plan dynamically matches stimulation parameters, including: for the subtype of response inhibition dysfunction, matching the concha cymba target point, first stimulation frequency, first stimulation intensity, first on / off cycle, and first pulse width; for the subtype of working memory dysfunction, matching the concha target point, second stimulation frequency, second stimulation intensity, continuous stimulation pattern, and second pulse width. The stimulation parameters are dynamically bound to features in the corrected HRV data that reflect the correlation between response inhibition and working memory.

10. The psychological trauma treatment system based on HRV monitoring and transcutaneous vagus nerve stimulation according to claim 9, characterized in that, The execution feedback module further includes a heart rate oscillation guidance unit and an electrode connection detection unit. The heart rate oscillation guidance unit provides at least one heart rate oscillation guidance mode and outputs guidance information to guide the patient to adjust their behavior. The on / off time of the guidance information is adapted to the on / off time of the personalized taVNS treatment plan. If the taVNS on / off time is less than twice the on / off time of the guidance mode, the patient is prompted to select one of the on / off times as the execution parameter. The electrode connection detection unit outputs a constant intensity detection current to the stimulation electrode and obtains the resistance value at both ends of the stimulation electrode. If the resistance value exceeds the preset range, an electrode connection abnormality prompt message is output and the taVNS intervention is stopped. The intervention is resumed after the resistance value meets the preset range. The execution feedback module stops taVNS intervention during the patient's inhalation and / or muscle contraction phases, and executes a personalized taVNS treatment plan during the exhalation and / or muscle extension phases.

Images