Chronic pain relieving method based on virtual reality and time interference nerve regulation

Abstract

The invention discloses a chronic pain relieving method based on virtual reality and time interference nerve regulation, and relates to the technical field of nerve regulation and virtual reality medical treatment, and the method comprises the steps: constructing an individual function vulnerability quantitative model of a multi-dimensional index; generating a dynamic adaptive virtual reality scene synchronized with the pain rhythm and emotion based on the model; configuring double-frequency time interference electric field parameters of targeted thalamus and anterior clasp back; performing millisecond-level synchronization on cognitive emotion events in neural regulation and virtual reality, and performing closed-loop adjustment on intervention parameters based on gamma wave band phase amplitude coupling strength; the pain evolution trend is predicted through the long-short-term memory network, and the subsequent intervention strategy is optimized. According to the invention, multi-dimensional precise targeted intervention on chronic pain is realized, the safety, individual suitability and long-term curative effect of treatment are remarkably improved, and the problems of poor suitability and difficult maintenance of curative effect in chronic pain intervention of old people are solved.

Description

Technical Field

[0001] This invention relates to the fields of neuromodulation and virtual reality medical technology, and in particular to a method for relieving chronic pain based on virtual reality and time-interference neuromodulation. Background Technology

[0002] With the advancement of technology, chronic pain has gradually become a core health issue affecting the quality of life and functional independence of the elderly. Chronic pain in the elderly not only manifests as persistent or intermittent physical discomfort but is also often accompanied by cognitive decline, mood disorders, and decreased motor function, forming a complex, multidimensional pathological state. Current clinical interventions mainly include pharmacological analgesia, physical therapy, and psychological support. However, long-term use of opioids or nonsteroidal anti-inflammatory drugs (NSAIDs) can easily lead to liver and kidney toxicity, gastrointestinal bleeding, and central nervous system side effects. Traditional non-pharmacological therapies have significant limitations in terms of individualized adaptation, depth of neuroplasticity regulation, and long-term efficacy maintenance. Therefore, there is an urgent need for a non-invasive intervention paradigm that can integrate multimodal physiological feedback, precisely target pain neural circuits, and possess dynamic adaptability to meet the comprehensive needs of the elderly for safety, compliance, and functionally integrated rehabilitation.

[0003] Virtual reality technology, due to its immersive perception reconstruction and behavioral guidance capabilities, has been initially explored for pain distraction and emotion regulation. However, existing pain intervention systems based on virtual reality mostly focus on attention shifting mechanisms, providing only one-way sensory stimulation through static or preset scenarios. They lack real-time analysis and closed-loop regulation of individual pain neurodynamic characteristics. Especially in the elderly, due to the degeneration of sensory gating function, weakened prefrontal cortex regulation ability, and abnormally active default mode network, it is difficult to effectively inhibit the ascending transmission of pain through the spinothalamic pathway by simply relying on external visual immersion. It is also impossible to reshape the pain emotion encoding of the cortical-limbic system. The lack of dynamic coupling between virtual reality content and the user's neurophysiological state in the time dimension leads to the inability of intervention intensity and timing to match the significantly fluctuating pain rhythm and diurnal neural oscillation characteristics of the elderly.

[0004] Current technologies face three fundamental contradictions in integrating neuromodulation with virtual reality. First, while traditional transcranial electrical stimulation (TCS) or magnetic stimulation can modulate the excitability of specific brain regions, their spatial resolution is limited and they cannot precisely synchronize with cognitive-emotional events in virtual scenarios, making it difficult to intervene in the phase-amplitude coupling process of pain-related neural oscillations on a millisecond-scale timescale. Second, existing systems lack a comprehensive assessment system for intrinsic function in the elderly, failing to incorporate multidimensional biomarkers such as gait stability, autonomic nervous system responses, cortisol rhythms, and resting-state functional connectivity into the intervention parameter generation logic, causing the modulation strategy to deviate from individual functional reserves and vulnerability thresholds. Third, while temporal intervention, as an emerging deep non-invasive neuromodulation technique, can theoretically achieve high-frequency electric field focusing on deep pain nodes such as the thalamus or anterior cingulate cortex, a synergistic enhancement mechanism has not yet been established between it and virtual reality-induced gamma-band neural synchronization, resulting in a significant decrease in modulation efficacy due to the lack of context-driven neural resonance. Summary of the Invention

[0005] The purpose of this invention is to provide a method for relieving chronic pain based on virtual reality and time-interference neuromodulation, in order to solve the problems of poor adaptability and difficulty in maintaining efficacy in the intervention of chronic pain in the elderly.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A chronic pain relief method based on virtual reality and time-interference neuromodulation includes the following specific steps: Step S1: Establish a multidimensional functional comprehensive assessment system for elderly patients with chronic pain. Collect individual gait stability data, heart rate variability data, skin conductance response data, and salivary cortisol concentration time series data through wearable sensor networks. Combine resting-state functional magnetic resonance imaging to obtain the functional connectivity strength index between the default mode network and the anterior cingulate cortex. Construct an individual functional vulnerability quantitative model containing at least 15 dimensions. Step S2: Generate a dynamic adaptive virtual reality intervention scene. Based on the individual functional vulnerability quantification model, use a generative adversarial network architecture to render an immersive environment that changes synchronously with the individual's pain rhythm and emotional state in real time. Scene elements include dynamic gradients of natural landscapes, interactive cognitive tasks, and multi-sensory stimulation fusion modules, with an update frequency of no less than 30 Hz. Step S3: Configure time-interference neural modulation parameters. Based on the individual functional vulnerability quantification model and real-time virtual reality scene content, calculate the dual-frequency cross-field parameters targeting the ventroposterolateral nucleus of the thalamus and the anterior cingulate cortex. The carrier frequencies are set to 2000 Hz and 2010 Hz, respectively. The modulation depth is dynamically adjusted from 10% to 90%. The stimulation intensity is 1.5 mA to 2.5 mA per square centimeter. Step S4: Perform closed-loop neural modulation and virtual reality synchronous intervention, apply the time interference electric field to the key cognitive and emotional events in the virtual reality scene for millisecond-level time locking, monitor the phase amplitude coupling strength of the γ-band neural oscillation in real time through a 64-lead EEG system, and dynamically adjust the interaction difficulty between the electric field focusing area and the virtual scene based on the proportional-integral-differential controller. Step S5: Evaluate the long-term intervention effect and update the strategy. After each intervention session, recalculate the changes in key indicators in the individual functional vulnerability quantification model, use a long short-term memory neural network to predict the evolution trend of pain intensity in the next 3 months, and optimize the virtual reality scene parameters and time interference modulation mode for the next cycle based on this.

[0007] Preferably, in step S1, the wearable sensor network includes a 3-axis accelerometer and gyroscope embedded in the insole, with a sampling frequency of 100 Hz, used to calculate the gait cycle variation coefficient; the chest-attached heart rate monitoring module uses photoplethysmography to measure the low-frequency and high-frequency power spectral density in heart rate variability; the wrist-worn skin conductivity sensor has a measurement range of 0 to 100 micro-Siemens and an accuracy of ±0.1 micro-Siemens; and the salivary cortisol is collected using a passive flow saliva collector, collected 5 times daily at fixed times, and the concentration is quantified by enzyme-linked immunosorbent assay (ELISA) with a time resolution of 30 minutes.

[0008] Preferably, the resting-state functional magnetic resonance imaging uses a 3 Tesla magnetic resonance scanner, a gradient echo planar imaging sequence, a repetition time of 2000 ms, an echo time of 30 ms, a voxel size of 3 mm x 3 mm x 3 mm, and a scan duration of 8 minutes. Subsequently, independent component analysis is used to extract the default mode network, and the functional connection strength between it and the anterior cingulate cortex is calculated using the Pearson correlation coefficient, with the connection strength threshold set to 0.3.

[0009] Preferably, in step S2, the generator of the generative adversarial network architecture contains 12 transposed convolutional layers, the discriminator contains 8 convolutional layers, the latent space dimension is 512, the training dataset contains more than 100,000 labeled natural scene images and corresponding physiological signal time series data, the scene rendering engine supports real-time ray tracing, the resolution reaches 3840 x 2160 pixels, and the latency is less than 20 milliseconds.

[0010] Preferably, the multi-sensory stimulation fusion module includes physically based 3D audio rendering, supports binaural time difference and intensity difference localization, and has a sound pressure level dynamic range of 30 to 80 dB; tactile feedback is achieved through a linear resonant actuator with a vibration frequency range of 50 to 250 Hz and adjustable intensity; olfactory stimulation uses a digital microfluidic chip to control the release concentration and mixing ratio of four basic odor molecules, with a response time of less than 100 milliseconds.

[0011] Preferably, in step S3, the dual-frequency cross electric field is applied through four circular electrodes with a diameter of 2 cm. The electrode arrangement follows the international 10-20 system, specifically located at points C3, C4, Fz, and Pz. Conductive gel is used between the electrodes and the scalp, and the impedance is maintained below 5 kΩ. The electric field distribution is calculated using the finite element method with a mesh count exceeding 1 million and a solution accuracy of 0.1 mm.

[0012] Preferably, the modulation depth is dynamically adjusted based on real-time EEG characteristics. When the phase amplitude coupling strength of the γ band is less than 15 percent of the baseline level, the modulation depth is increased by 5 percent; when the coupling strength is more than 20 percent of the baseline level, the modulation depth is decreased by 5 percent; the adjustment period is 10 seconds.

[0013] Preferably, the millisecond-level time locking in step S4 is achieved through a dedicated synchronization module. This module integrates a high-precision clock with a jitter of less than 1 millisecond. The synchronization signal simultaneously triggers the waveform output of the event marker and the time interference stimulator in the virtual reality engine, ensuring that the time difference between the visual stimulus onset and the rising edge of the electric field modulation envelope is within ±5 milliseconds.

[0014] Preferably, the proportional-integral-derivative controller has a proportional coefficient of 0.8, an integration time of 2 seconds, a derivative time of 0.1 seconds, and a control objective of maintaining the phase amplitude coupling strength of the γ-band within a 95 percent confidence interval of the individualized target range. The target range is set to the average value plus or minus 1.5 standard deviations based on the baseline measurement value.

[0015] Preferably, in step S5, the long short-term memory neural network contains two hidden layers, each with 128 units. The input features are 15-dimensional functional assessment indicators, pain visual simulation scores, and medication use logs from the past 10 intervention sessions. The output is a predicted value of daily pain intensity for the next 90 days, with a root mean square error of less than 0.5. The strategy update is based on the difference between the predicted results and the preset functional improvement target, and adjusts the cognitive load level of the virtual reality scene and the stimulation intensity parameters of the time intervention in the next cycle through a Bayesian optimization algorithm.

[0016] Preferably, the method further includes establishing an individualized pain neural signature database, storing more than 1,000 hours of multimodal physiological data, brain imaging data and behavioral data for each user; based on this database, graph convolutional networks are used to identify key neural circuit topological features related to the progression of chronic pain, with a feature dimension of 256 and an identification accuracy of more than 92%.

[0017] Preferably, the virtual reality scene includes an adaptive cognitive training task based on reinforcement learning. The task difficulty is dynamically adjusted according to the user's real-time performance. The adjustment strategy adopts the ε-greedy algorithm, where the ε value is 0.1. The reward function comprehensively considers the task completion speed, accuracy, and physiological stress response indicators.

[0018] Preferably, the synergistic effect of the time-interference neural modulation and virtual reality is quantified by calculating the cross-frequency coupling index, which is defined as the modulation index between the theta oscillation phase induced by virtual reality and the γ oscillation amplitude of the time-interference modulation. A modulation index greater than 0.2 is considered to be effective synergy and triggers the solidification and preservation of the intervention parameters.

[0019] Preferably, the method requires a two-week safety adaptation period before implementation. During this period, the intensity of the time interference stimulus is gradually increased from 0.5 mA per square centimeter to the target intensity, and the sensory load of the virtual reality scene is also gradually increased from the lowest level, while any adverse reactions are closely monitored.

[0020] Compared with the prior art, the beneficial technical effects of the present invention are as follows: This invention constructs a closed-loop intervention framework by deeply integrating temporal interference neuromodulation, dynamic adaptive virtual reality, and comprehensive functional assessment of the elderly, achieving multidimensional and precise targeted relief of chronic pain in the elderly. This method can analyze and regulate individual pain neurodynamic characteristics in real time, synchronizing neuromodulation with cognitive and emotional events in the virtual environment at the millisecond level. It effectively intervenes in pain-related neural oscillations and dynamically optimizes strategies based on long-term functional assessment, thereby significantly improving the safety, adaptability, and long-term efficacy maintenance of the intervention. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall technical solution architecture of the chronic pain relief method based on virtual reality and time-interference neuromodulation proposed in this invention; Figure 2 This is a schematic diagram of the core principle framework of the synergistic effect of time-interference neural modulation and dynamic adaptive virtual reality proposed in this invention. Detailed Implementation

[0022] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the present invention and not to limit the present invention. For those skilled in the art, the present invention can be practiced without some of these specific details. The following description of the embodiments is merely to provide a better understanding of the present invention by illustrating examples of the invention.

[0023] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0024] In the embodiments of the present invention, the same reference numerals denote the same components, and for the sake of brevity, detailed descriptions of the same components are omitted in different embodiments. It should be understood that the thickness, length, width, and other dimensions of various components in the embodiments of the present invention shown in the accompanying drawings, as well as the overall thickness, length, width, and other dimensions of the integrated device, are merely illustrative and should not constitute any limitation on the present invention; the term "multiple" in the present invention refers to two or more (including two).

[0025] In the current context, chronic pain has become a core health problem affecting the quality of life and functional independence of the elderly. Chronic pain in the elderly not only manifests as persistent or intermittent physical discomfort, but is also often accompanied by cognitive decline, mood disorders, and decreased motor function, forming a complex multidimensional pathological state. Existing clinical interventions have significant limitations in terms of individualized adaptation, depth of neuroplasticity modulation, and long-term efficacy maintenance. To address these technical problems, this invention proposes a closed-loop intervention framework that deeply integrates temporal interference neuromodulation, dynamic adaptive virtual reality, and comprehensive geriatric functional assessment. This framework achieves multidimensional and precise targeted relief of chronic pain in the elderly and is applied to a chronic pain relief method based on virtual reality and temporal interference neuromodulation. Please refer to Figure 1 The overall technical architecture of this invention includes a multi-dimensional functional comprehensive evaluation module, a dynamic adaptive virtual reality generation module, a time-interference neuromodulation module, a closed-loop synchronous execution module, and a long-term effect evaluation and strategy update module. These modules work collaboratively to form a complete, data-driven, real-time feedback neuromodulation closed-loop system.

[0026] Please refer to Figure 2 The synergistic principle of time-interference neural modulation and dynamic adaptive virtual reality lies in the precise alignment of external sensory stimulation events with the internal electric field modulation envelope through millisecond-level time locking, thereby inducing and enhancing the phase-amplitude coupling of specific neural oscillations, and achieving precise intervention on pain-related neural circuits.

[0027] Step S1: Establish a multidimensional functional comprehensive assessment system for elderly patients with chronic pain. This involves collecting individual gait stability data, heart rate variability data, skin conductance response data, and salivary cortisol concentration time-series data using a wearable sensor network. Resting-state functional magnetic resonance imaging (fMRI) is used to obtain the functional connectivity strength index between the default mode network and the anterior cingulate cortex, constructing a quantitative model of individual functional vulnerability encompassing at least 15 dimensions. Specifically, the wearable sensor network includes a 3-axis accelerometer and gyroscope embedded in the insole, with a sampling frequency of 100 Hz, used to calculate the gait cycle variability coefficient; a chest-attached heart rate monitoring module employs photoplethysmography (PPG) to measure the low-frequency and high-frequency power spectral densities in heart rate variability; a wrist-worn skin conductance sensor with a measurement range of 0 to 100 microsiemens and an accuracy of ±0.1 microsiemens; and salivary cortisol collection using a passive saliva collector, collected five times daily at fixed time points, with concentration quantified by enzyme-linked immunosorbent assay (ELISA) at a time resolution of 30 minutes. Furthermore, resting-state functional magnetic resonance imaging (fMRI) was performed using a 3 Tesla MRI scanner with a gradient echo planar imaging sequence. The repetition time was 2000 ms, the echo time was 30 ms, the voxel size was 3 mm x 3 mm x 3 mm, and the scan duration was 8 minutes. Subsequently, independent component analysis was used to extract the default mode network, and the functional connectivity strength with the anterior cingulate cortex was calculated using the Pearson correlation coefficient, with a connectivity strength threshold set at 0.3. The construction process of this 15-dimensional individual functional vulnerability quantification model involves the preprocessing, feature extraction, and standardization of the raw physiological signals. For example, after wavelet denoising, the coefficients of variation of parameters such as step length, gait speed, and double support phase time were calculated from the gait data; heart rate variability data were decomposed into low-frequency (0.04 to 0.15 Hz) and high-frequency (0.15 to 0.4 Hz) power spectral densities using fast Fourier transform, and the LF / HF ratio was calculated; peak amplitude, rise time, and recovery slope were extracted from the skin conductance response data; and the area under the curve and peak time offset were extracted from the salivary cortisol data by fitting a diurnal rhythm curve. All features are normalized to the 0-1 range and input into a pre-trained multilayer perceptron, which outputs a comprehensive vulnerability score, which directly determines the initial strength and complexity of subsequent interventions.

[0028] Step S2 generates a dynamically adaptive virtual reality intervention scene. Based on an individual functional vulnerability quantification model, a generative adversarial network (GAN) architecture is used to render an immersive environment that changes in real time in sync with the individual's pain rhythm and emotional state. Scene elements include dynamic gradients of natural landscapes, interactive cognitive tasks, and a multi-sensory stimulation fusion module, with an update frequency of no less than 30 Hz. Specifically, the generator of the GAN architecture contains 12 transposed convolutional layers, the discriminator contains 8 convolutional layers, the latent space dimension is 512, the training dataset contains more than 100,000 labeled natural scene images and corresponding physiological signal time-series data, the scene rendering engine supports real-time ray tracing, a resolution of 3840 x 2160 pixels, and a latency of less than 20 milliseconds. The multi-sensory stimulation fusion module includes physically based 3D audio rendering, supporting binaural time difference and intensity difference localization, with a sound pressure level dynamic range of 30 to 80 dB; tactile feedback is achieved through linear resonant actuators, with a vibration frequency range of 50 to 250 Hz and adjustable intensity; olfactory stimulation uses a digital microfluidic chip to control the release concentration and mixing ratio of four basic odor molecules, with a response time of less than 100 milliseconds. The execution process begins by reading the current emotional state label and pain rhythm phase from an individual's functional vulnerability quantification model. The generator network samples from the latent space based on these labels and generates matching scene textures, lighting conditions, and object layouts. For example, when a user is detected to be in a high-anxiety state, the system renders a tranquil forest scene, accompanied by soothing birdsong, the tactile sensation of a breeze rustling through leaves, and a faint pine scent. The interactive cognitive task is based on the adaptive mechanism of reinforcement learning. The task difficulty is dynamically adjusted according to the user's real-time performance. The adjustment strategy adopts the ε-greedy algorithm, where ε is 0.1. The reward function comprehensively considers the task completion speed, accuracy, and physiological stress response indicators. The entire scene rendering pipeline has been highly optimized to ensure that the end-to-end latency from physiological signal input to final image output is strictly controlled within 20 milliseconds to maintain the user's immersion and the system's real-time performance.

[0029] Step S3 involves configuring time-interference neuromodulation parameters. Based on an individual functional vulnerability quantification model and real-time virtual reality scene content, dual-frequency cross-field parameters targeting the ventroposterolateral nucleus of the thalamus and the anterior cingulate cortex are calculated. The carrier frequencies are set to 2000 Hz and 2010 Hz, respectively. The modulation depth is dynamically adjusted from 10% to 90%, and the stimulation intensity is 1.5 mA to 2.5 mA per square centimeter. Specifically, the dual-frequency cross-field is applied through four circular electrodes with a diameter of 2 cm. The electrode arrangement follows the international 10-20 system, specifically located at points C3, C4, Fz, and Pz. Conductive gel is used between the electrodes and the scalp, maintaining an impedance below 5 kΩ. The electric field distribution is calculated using the finite element method, with a mesh size exceeding 1 million and a solution accuracy of 0.1 mm. The dynamic adjustment of modulation depth is based on real-time EEG characteristics. When the gamma band phase amplitude coupling strength is below 15% of the baseline level, the modulation depth increases by 5%; when the coupling strength is above 20% of the baseline level, the modulation depth decreases by 5%. The adjustment period is 10 seconds. The core of this step lies in precisely calculating electrode locations and current intensities to create the necessary interference envelope in deep brain regions. The system first loads the individual's head MRI structural images, performs tissue segmentation (distinguishing between scalp, skull, cerebrospinal fluid, gray matter, and white matter), and then constructs a finite element mesh model containing over one million elements. Based on this model, the Poisson equation is solved to simulate the electric field distribution generated under different electrode combinations. Through iterative optimization algorithms, the electrode current configuration that precisely focuses the maximum electric field on the ventroposterolateral nucleus of the thalamus and the anterior cingulate cortex is found. The initial stimulation intensity is set based on a comprehensive score in the individual's functional vulnerability quantification model; the higher the vulnerability, the lower the initial intensity, typically starting from 1.5 mA per square centimeter. Modulation depth, as a key dynamic parameter, has its adjustment logic built into the closed-loop controller, ensuring that the intensity of neuromodulation always matches the user's real-time neural state.

[0030] Step S4 involves executing closed-loop neural modulation and synchronous intervention with virtual reality. A time-interference electric field is applied to key cognitive and emotional events in the virtual reality scene, achieving millisecond-level time locking. A 64-lead EEG system monitors the phase-amplitude coupling strength of the gamma-band neural oscillations in real time, and dynamically adjusts the interaction difficulty between the electric field focusing area and the virtual scene based on a proportional-integral-derivative (PID) controller. Specifically, millisecond-level time locking is achieved through a dedicated synchronization module. This module integrates a high-precision clock with jitter less than 1 millisecond. The synchronization signal simultaneously triggers the waveform output of the event marker and the time-interference stimulator in the virtual reality engine, ensuring that the time difference between the visual stimulus onset and the rising edge of the electric field modulation envelope is within ±5 milliseconds. The PID controller has a proportional gain of 0.8, an integration time of 2 seconds, and a derivative time of 0.1 seconds. The control objective is to maintain the gamma-band phase-amplitude coupling strength within a 95% confidence interval of the individualized target range, which is set to ±1.5 standard deviations of the average value based on baseline measurements. This step is the core execution link of the entire closed-loop system. A 64-lead EEG system continuously acquires data at a sampling rate of 1000 Hz. The data stream is filtered by a real-time filter bank to separate the gamma-band signal. The phase-amplitude coupling strength is calculated using the modulation index method, which calculates the mutual information between the low-frequency theta-band phase and the high-frequency gamma-band amplitude. The calculated coupling strength value is used as a feedback signal input to a proportional-integral-differential controller (PIDC). The controller has two outputs: one is to send instructions to the temporal interference stimulator to fine-tune the current ratio of the four electrodes, thereby moving the electric field focus; the other is to send instructions to the virtual reality engine to adjust the difficulty parameters of the current cognitive task. This bidirectional adjustment mechanism ensures that the intervention process is always in an optimal neural resonance state. The synergistic effect of temporal interference neuromodulation and virtual reality is quantified by calculating the cross-frequency coupling index, which is defined as the modulation index between the theta oscillation phase induced by virtual reality and the gamma oscillation amplitude regulated by temporal interference. A modulation index greater than 0.2 is considered effective synergy and triggers the solidification and storage of intervention parameters for direct retrieval in subsequent sessions.

[0031] Step S5 involves evaluating the long-term intervention effect and updating the strategy. After each intervention session, the changes in key indicators in the individual functional vulnerability quantification model are recalculated. A long short-term memory neural network is used to predict the evolution trend of pain intensity over the next three months, and based on this, the virtual reality scene parameters and temporal interference modulation mode for the next cycle are optimized. Specifically, the long short-term memory neural network contains two hidden layers, each with 128 units. The input features are 15-dimensional functional assessment indicators, pain visual simulation scores, and medication use logs from the past 10 intervention sessions. The output is the predicted value of daily pain intensity for the next 90 days, with a root mean square error of less than 0.5. The strategy update is based on the difference between the predicted results and the preset functional improvement target, and the cognitive load level of the virtual reality scene and the stimulation intensity parameters of the temporal interference are adjusted using a Bayesian optimization algorithm. In addition, the method also includes establishing an individualized pain neural signature database, storing more than 1,000 hours of multimodal physiological data, brain imaging data, and behavioral data for each user. Based on this database, a graph convolutional network is used to identify key neural circuit topological features related to the progression of chronic pain, with a feature dimension of 256 and an accuracy rate of more than 92%. This step constitutes the system's long-term learning and evolutionary capabilities. After each session, the system automatically triggers an evaluation process, comparing various physiological indicators before and after the intervention and calculating the improvement rate. This data, along with the user's subjective pain score, is fed into a long short-term memory neural network for training and prediction. If the prediction results indicate a future trend of worsening pain, the Bayesian optimization algorithm will proactively increase the intensity of the intervention in the next cycle, such as increasing the cognitive challenge of the virtual reality scene or raising the upper limit of the modulation depth of the temporal intervention. The personalized pain neural signature database is not only the foundation for strategy updates but also a valuable resource for conducting group-level research. Graph convolutional networks, through the analysis of massive brain functional connectivity graphs in the database, can discover new biomarkers related to chronic pain. These new discoveries can feed back into and upgrade individual functional vulnerability quantification models, forming a virtuous cycle of continuous evolution.

[0032] In this embodiment, a 72-year-old female patient with chronic pain caused by knee osteoarthritis for 5 years was included. During a 2-week safety adaptation period, the temporal interference stimulation intensity was gradually increased from 0.5 mA / cm² to 2.0 mA / cm², and the sensory load of the virtual reality scene was gradually increased from the lowest level to the intermediate level. No adverse reactions were observed during the adaptation period. In the formal intervention phase, the system generated a virtual scene themed "Tranquil Garden" based on the patient's individual functional vulnerability quantification model. In a typical 30-minute intervention session, when the patient successfully completed a cognitive task, the system detected a brief decrease in the gamma-band phase amplitude coupling strength. The millisecond synchronization module immediately triggered the temporal interference stimulator, applying an electric field envelope with a modulation depth of 60% at the moment of task completion. The 64-lead EEG system monitored that the coupling strength recovered and stabilized within the target range within 5 seconds. After the session, the system recorded a 12% reduction in the coefficient of variation of gait cycle and an earlier peak time of salivary cortisol by 45 minutes, indicating an improvement in autonomic function. After 8 weeks of intervention, their visual analog scale score for pain dropped from 7 to 3, and their daily activities improved significantly.

[0033] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for chronic pain relief based on virtual reality and time- dependent neuromodulation, characterized in that, Comprise the following steps: Step S1: Establish a multi-dimensional functional comprehensive evaluation system for elderly patients with chronic pain, collect individual gait stability data, heart rate variability data, skin conductance response data and salivary cortisol concentration time series data through a wearable sensor network, combine resting state functional magnetic resonance imaging to obtain default mode network and prefrontal cortex functional connection strength indicators, and construct an individual functional vulnerability quantification model; Step S2: Generate a dynamic adaptive virtual reality intervention scene, based on the individual functional vulnerability quantification model, use a generative adversarial network architecture to render an immersive environment that changes in synchronization with the individual's pain rhythm and emotional state in real time, scene elements include natural landscape dynamic gradient, interactive cognitive tasks and multi-sensory stimulation fusion module; Step S3: Configure time interference neural regulation parameters, calculate the double-frequency cross electric field parameters of the target thalamus ventral posterior lateral nucleus and prefrontal cortex according to the individual functional vulnerability quantification model and real-time virtual reality scene content; Step S4: Perform closed-loop neural regulation and virtual reality synchronous intervention, millisecond-level time locking of time interference electric field application and key cognitive emotional events in the virtual reality scene, real-time monitoring of the phase amplitude coupling strength of gamma band neural oscillations through a 64-channel electroencephalogram system, and dynamic adjustment of the electric field focusing area and virtual scene interaction difficulty based on a proportional integral derivative controller; Step S5: Evaluate the long-term intervention effect and update the strategy, recalculate the key indicator change in the individual functional vulnerability quantification model after each intervention session, use a long short-term memory neural network to predict the future 3-month pain intensity evolution trend, and optimize the virtual reality scene parameters and time interference modulation mode for the next cycle based on this.

2. The chronic pain relief method based on virtual reality and time interference neural regulation according to claim 1, characterized in that: The wearable sensor network includes a 3-axis accelerometer and gyroscope embedded in the insole; the chest-attached heart rate monitoring module uses the photoplethysmography technology; the wrist-mounted skin conductance sensor measures a range of 0 to 100 microsiemens; Salivary cortisol collection uses a passive flow saliva collector, and the concentration is quantified by enzyme-linked immunosorbent assay.

3. The chronic pain relief method based on virtual reality and time interference neural regulation according to claim 1, characterized in that: The resting state functional magnetic resonance imaging uses a 3 Tesla magnetic resonance scanner, a gradient echo planar imaging sequence, a repetition time of 2000 milliseconds, an echo time of 30 milliseconds, a voxel size of 3 mm x 3 mm x 3 mm, and a scanning time of 8 minutes; the default mode network is extracted by independent component analysis, and the functional connection strength of the prefrontal cortex is calculated by the Pearson correlation coefficient.

4. The chronic pain relief method based on virtual reality and time interference neural regulation according to claim 1, characterized in that: The generator of the generative adversarial network architecture contains 12 transpose convolutional layers, the discriminator contains 8 convolutional layers, and the latent space dimension is 512; the scene rendering engine supports real-time ray tracing, the resolution reaches 3840x2160 pixels, and the end-to-end delay is less than 20 milliseconds.

5. The method of claim 1, wherein the method further comprises: adjusting the cognitive load of the virtual reality scene based on the pain intensity prediction value. The multi-sensory stimulation fusion module includes physical-based 3D audio rendering; tactile feedback is realized through linear resonant actuators; olfactory stimulation uses a digital microfluidic chip to control the release concentration and mixing ratio of four basic odor molecules.

6. The method of claim 1, wherein the method further comprises: adjusting the cognitive load of the virtual reality scene based on the pain intensity prediction value. The double-frequency cross electric field is applied through four circular electrodes with a diameter of 2 centimeters, the electrode positioning follows the international 10-20 system, and is specifically located at the C3, C4, Fz and Pz points; conductive gel is used between the electrode and the scalp, and the impedance is maintained below 5 kilohms; the electric field distribution is calculated by the finite element method.

7. The method of claim 1, wherein the method further comprises: adjusting the modulation depth based on the real-time gamma band phase-amplitude coupling strength, when the strength is lower than 15% of the baseline level, the modulation depth is increased by 5%; when it is higher than 20% of the baseline level, the modulation depth is reduced by 5%.

8. The method of claim 1, wherein the method further comprises: achieving millisecond-level time locking through a dedicated synchronization module, the synchronization signal triggers the virtual reality engine event marker and the time interference stimulator waveform output, ensuring that the time difference between the visual stimulation onset and the rising edge of the electric field modulation envelope is within ±5 milliseconds; the proportional coefficient of the proportional-integral-derivative controller is 0.8, the integral time is 2 seconds, and the differential time is 0.1 second, and the control target is to maintain the gamma band phase-amplitude coupling strength within 95% of the confidence interval of the individualized target range, which is the baseline average value ±1.5 standard deviations.

9. The method of claim 1, wherein the method further comprises: the long short-term memory neural network contains 2 hidden layers, each layer has 128 units, the input features include 15-dimensional functional evaluation indicators, pain visual analog scale scores and drug use logs of the past 10 intervention sessions, the output is the pain intensity prediction value for each day within the next 90 days, and the root mean square error is less than 0.5; the strategy update adjusts the cognitive load level of the virtual reality scene and the time interference stimulation intensity parameters in the next period through the Bayesian optimization algorithm.

10. The method of claim 1, wherein the method further comprises: ​ ​ ​ Also included is the establishment of an individualized pain neural signature database storing over 1000 hours of multimodal physiological data, brain imaging data, and behavioral data for each user; based on this database, graph convolution networks are used to identify key neural circuit topological features related to chronic pain progression.

Images