Dynamic pain neural decoding and intervention system based on prefrontal temporal feature window

Abstract

The present invention relates to the technical field of medical health data information decoding, analysis, and processing, and specifically discloses a dynamic pain neural decoding and intervention system based on a prefrontal cortex temporal feature window, comprising: a signal acquisition module, a dynamic window control module, a phase prediction module, an intervention execution module, and an event-driven controller. The dynamic window control module intelligently adjusts the signal analysis window duration by real-time calculation of the energy ratio between the #imgabs0# frequency band and the #imgabs1# frequency band, accurately locking the transient feature window during the pain outbreak period; the phase prediction module achieves millisecond-level synchronization between the stimulation pulse and the rising edge of the action potential of the neural cluster based on a dual-factor mapping mechanism of individualized conduction velocity level and oscillation frequency; and the event-driven controller reduces power consumption to the level of a wearable device while maintaining a 10ms-level real-time response through a dynamic resource allocation strategy.

Description

Technical Field

[0001] The present invention relates to a dynamic pain nerve decoding and intervention system based on a prefrontal temporal feature window, belonging to the technical field of medical health data information decoding analysis and processing. Background Art

[0002] Pain neurological data decoding and intervention is an important research area in the field of medical and health information technology. By analyzing the temporal characteristics of prefrontal EEG signals, it enables objective assessment and precise intervention of pain states. Traditional systems typically use a fixed-duration analysis window (e.g., 1 second) to capture neural signals and trigger electrical stimulation intervention based on a preset frequency threshold. However, such methods have significant drawbacks in practical applications:

[0003] 1. Insufficient ability to capture transient pain features: The fixed window cannot adapt to the dynamic characteristics of theta-gamma cross-band coupling events in the prefrontal lobe (pain bursts typically last 0.3-0.8 seconds), resulting in key signals such as high-frequency oscillation phase transitions being submerged by smooth noise, and the feature loss rate exceeds 60%.

[0004] 2. Timing mismatch between intervention pulses and neural activity: Existing technologies rely on a unidirectional stimulus-response model, ignoring individual differences in nerve conduction velocity and fluctuations in oscillation frequency. This results in a 50-200ms delay between the stimulation pulse and the action potential of the target nerve cluster, missing the optimal time window for inducing synaptic plasticity and causing the intervention efficiency to decrease by more than 50%.

[0005] 3. System real-time performance and power consumption are difficult to achieve simultaneously: To improve accuracy, existing solutions often continuously run high-computing algorithms (such as real-time Fourier transforms), resulting in end-to-end latency exceeding 50ms and power consumption exceeding 5W, which cannot meet the needs of wearable devices and long-term monitoring scenarios. To address these issues, the industry has attempted to improve by optimizing signal processing algorithms (such as adaptive filtering) or increasing hardware computing power, but the former is still limited by a fixed window framework, while the latter increases power consumption and cost pressures.

[0006] See patent publication number CN111951958A, which discloses a pain data assessment method based on autoencoders and related components. This method uses a convolutional neural network (CNN) to extract features from laser-evoked potential (LEP) EEG data and employs deep separable convolutional layers to reduce the number of network parameters, thereby improving system efficiency. Although this method improves the accuracy of pain prediction by predicting pain levels through a machine learning model, it relies on static data feature extraction methods and cannot dynamically respond to transient changes in neural signals. This can lead to missed opportunities for precise intervention during dynamic pain interventions. Traditional neural network models also fail to fully account for individual differences, resulting in poor adaptability of the prediction model, particularly when dealing with fluctuations in different nerve conduction velocities and oscillation frequencies, which prevent precise matching. Summary of the Invention

[0007] The present invention provides a dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window. Its main purpose is to solve the problems of incomplete pain feature capture, low intervention efficiency and difficulty in clinical implementation caused by fixed analysis windows, phase delay and high power consumption in the existing technology.

[0008] To achieve the above objectives, the present invention provides a dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window, comprising: a signal acquisition module for acquiring prefrontal EEG signals in real time through an EEG device, and using an analog filter bank to separate Frequency band signal and Frequency band signal, where The frequency band is 4 to 8 Hz, The frequency band is 30 to 80 Hz;

[0009] A dynamic window control module is connected to the signal acquisition module and includes an energy ratio calculation unit for calculating the energy ratio of the signal. Frequency band signal and The frequency band signals are respectively integrated in a sliding time window to calculate the instantaneous energy ratio of the two; the window adjustment unit compresses the signal analysis window to a first preset time length when the instantaneous energy ratio exceeds a preset threshold, otherwise it is extended to a second preset time length, where the preset threshold is 1.2 to 1.8 times the average energy value of the frequency band, the first preset duration is 0.2-0.4 seconds, and the second preset duration is 0.8-1.6 seconds;

[0010] a phase prediction module, connected to the dynamic window control module, comprising: a frequency detection unit for extracting the dominant frequency of the current prefrontal oscillation signal; a delay mapping unit for matching a corresponding phase compensation time parameter from a pre-stored frequency delay comparison table with different conduction velocity levels based on the dominant frequency and the individual nerve conduction velocity level; the phase compensation time parameter is used to represent the lead time of the stimulation pulse triggering moment relative to the rising edge of the action potential of the target neural cluster;

[0011] an intervention execution module, connected to the phase prediction module, configured to trigger an electrical stimulation pulse during the dynamic window compression phase, and adjust the triggering timing of the stimulation pulse in advance to the rising edge of the action potential of the target neural cluster based on the phase compensation time parameter;

[0012] The event-driven controller connects the signal acquisition module, the dynamic window control module, the phase prediction module and the intervention execution module, and is configured as follows: when the energy ratio does not exceed the threshold, only the low-power operation of the signal acquisition module and the dynamic window control module is maintained; when the energy ratio exceeds the threshold, the phase prediction module and the intervention execution module are activated to the full-power mode, and the low-power state is restored after the window compression phase ends.

[0013] In a preferred embodiment, the energy ratio calculation unit is implemented by: Frequency band signal and The frequency band signals are full-wave rectified respectively; the rectified signals are smoothed by a first-order low-pass filter with a cutoff frequency of 2 Hz; the smoothed signals are calculated. Frequency band signal energy and The ratio of the signal energy in the frequency band.

[0014] In a preferred embodiment, the frequency delay comparison table is generated by the following rules: based on the big data of nerve conduction velocity of healthy people and chronic pain patients, the reference delay parameters corresponding to four different conduction velocity levels are preset, wherein the different conduction velocity levels include slow, medium, fast and ultrafast; the reference delay parameter is negatively correlated with the current oscillation frequency, and the delay decreases by 1.5ms for every 10Hz increase in frequency; the parameter is automatically calibrated after each intervention: if the actual phase deviation exceeds If the angle is radian, adjust the corresponding parameter in the table with a step size of ±2ms.

[0015] In a preferred embodiment, the power consumption control of the event-driven controller includes: in low power consumption mode, the signal acquisition module and the dynamic window control module occupy 5% of the chip resources, and the main frequency is reduced to 50MHz; in full power consumption mode, the phase prediction module and the intervention execution module occupy 95% of the chip resources, and the main frequency is increased to 200MHz.

[0016] In a preferred embodiment, the stimulation parameters of the intervention execution module meet the following requirements: the stimulation frequency is consistent with the current prefrontal cortex. The deviation of the oscillation frequency did not exceed ±5%; the stimulation intensity was mapped to a current range of 50 to 500 μA based on the input value of the pain level assessment scale through fuzzy logic rules.

[0017] In a preferred embodiment, the dynamic window control module implements window adjustment through hardware logic, specifically including: using the shift register group embedded in the FPGA to store the signal data in the current window; dynamically controlling the length of the shift register through a programmable counter to achieve continuous adjustment of the window length from 0.1 to 2 seconds.

[0018] In a preferred embodiment, the dynamic pain nerve decoding and intervention system also includes a safety control module for performing the following: monitoring the number of continuous high-frequency stimulations, and automatically switching to a baseline intervention mode if the pain level does not decrease after three stimulations. The stimulation frequency of the baseline intervention mode is 40 Hz and the stimulation intensity is 100 μA; when the EEG signal amplitude is detected to be greater than 200 μV, the stimulation output is immediately cut off and an alarm is triggered.

[0019] In a preferred embodiment, the signal acquisition module is compatible with the digital interface of Neuroscan equipment and BioSemi equipment, including: an SPI protocol conversion unit or an I2C protocol conversion unit for receiving raw EEG data from external devices; a signal resampling unit for uniformly converting the input signal to a 250Hz sampling rate.

[0020] In a preferred embodiment, the frequency delay comparison table includes the following mapping relationship: when the oscillation frequency is between 30 and 40 Hz and the conduction velocity level is medium, the corresponding compensation delay is 20 ms; when the oscillation frequency is between 40 and 60 Hz and the conduction velocity level is fast, the corresponding compensation delay is 15 ms; when the oscillation frequency is between 60 and 80 Hz and the conduction velocity level is ultrafast, the corresponding compensation delay is 10 ms.

[0021] Compared with the problems described in the background technology, the beneficial effects of the present invention are: Frequency band and Real-time dynamic calculation of frequency band energy ratio (non-fixed threshold) enables intelligent expansion and contraction of signal analysis window, directly responding to the unique pain coding of prefrontal cortex. Cross-band coupling events can accurately lock the transient feature window of 0.2-0.4 seconds during the pain outbreak period, overcome the smooth interference of traditional fixed windows on high-frequency oscillation phase transitions, and significantly improve the integrity of key signal capture; based on the dual-factor mapping mechanism of individualized conduction velocity level and oscillation frequency, combined with self-learning calibration of delay parameters, it can achieve precise timing matching between stimulation pulses and rising edges of action potentials of target neural clusters, breaking through the phase mismatch problem caused by nerve conduction delay in traditional interventions and ensuring the optimal induction window for synaptic plasticity; adopting hardware-level dynamic resource allocation strategy, it activates full-power computing modules only when pain characteristics appear, so that the system can maintain 10ms terminal While reducing the end-to-end delay, the overall power consumption is reduced to the level of wearable devices, achieving the goal of reconciling the technical contradiction between "high real-time" and "low power consumption" in medical equipment; integrating the objective characteristics of neural signals with the subjective input of clinical pain scores, dynamically mapping stimulation parameters through fuzzy logic, and automatically switching to the safe baseline mode when high-frequency intervention is ineffective, constructing a full-chain individualized adaptation logic from signal analysis to safe output, avoiding the risk of over-stimulation in parameter settings of traditional solutions; through standardized protocol conversion and signal resampling modules, it directly connects to the digital interface of existing medical equipment, and can achieve functional upgrades without hardware modification, meeting the low-cost and high-efficiency system iteration needs of medical institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a structural diagram of the dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window of the present invention;

[0023] Figure 2 A comparison diagram of nerve cluster stimulation triggering between the conventional method of the present invention and the method of the present invention;

[0024] Figure 3 This is a structural diagram of the dynamic window control and phase prediction module based on the Xilinx Zynq chip of the present invention.

[0025] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0027] The present invention provides a dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window. It includes: a signal acquisition module for acquiring prefrontal EEG signals in real time through an EEG device, and using an analog filter bank to separate the signals that have a greater impact on the decoding and intervention of the pain state. Frequency band signal and Frequency band signal, where The frequency band is 4 to 8 Hz, The frequency band is 30 to 80 Hz; a dynamic window control module connected to the signal acquisition module, including: an energy ratio calculation unit for Frequency band signal and The frequency band signals are respectively integrated in a sliding time window to calculate the instantaneous energy ratio of the two; the window adjustment unit compresses the signal analysis window to a first preset time length when the instantaneous energy ratio exceeds a preset threshold, otherwise it is extended to a second preset time length, where the preset threshold is The first preset duration is 1.2 to 1.8 times of the frequency band energy mean, the first preset duration is 0.2-0.4 seconds, and the second preset duration is 0.8-1.6 seconds; a phase prediction module, connected to the dynamic window control module, including: a frequency detection unit, used to extract the main frequency of the current prefrontal lobe oscillation signal; a delay mapping unit, according to the main frequency and the individual nerve conduction velocity level, matches the corresponding phase compensation time parameter from the pre-stored frequency delay comparison table with different conduction velocity levels, the phase compensation time parameter is used to characterize the advance time amount of the stimulation pulse triggering moment relative to the rising edge of the action potential of the target neural cluster; an intervention execution module , connected to the phase prediction module, used to trigger the electrical stimulation pulse in the dynamic window compression stage, and adjust the triggering timing of the stimulation pulse to the rising edge of the action potential of the target neural cluster based on the phase compensation time parameter; an event-driven controller, connected to the signal acquisition module, the dynamic window control module, the phase prediction module and the intervention execution module, configured as: when the energy ratio does not exceed the threshold, only maintain the low power consumption operation of the signal acquisition module and the dynamic window control module; when the energy ratio exceeds the threshold, activate the phase prediction module and the intervention execution module to the full power consumption mode, and restore the low power consumption state after the window compression stage ends.

[0028] In a preferred embodiment, the energy ratio calculation unit is implemented by: Frequency band signal and The frequency band signals are full-wave rectified respectively; the rectified signals are smoothed by a first-order low-pass filter with a cutoff frequency of 2 Hz; the smoothed signals are calculated. Frequency band signal energy and The frequency delay comparison table is generated by the following rules: based on the big data of nerve conduction velocity of healthy people and chronic pain patients, the baseline delay parameters corresponding to four different conduction velocity levels are preset, where the different conduction velocity levels include slow, medium, fast and ultra-fast; the baseline delay parameter is negatively correlated with the current oscillation frequency, and the delay decreases by 1.5ms for every 10Hz increase in frequency; the parameters are automatically calibrated after each intervention: if the actual phase deviation exceeds The corresponding parameters in the table are adjusted in ±2ms steps. The power consumption control of the event-driven controller includes: in low-power mode, the signal acquisition module and the dynamic window control module occupy 5% of the chip resources, and the main frequency is reduced to 50MHz; in full-power mode, the phase prediction module and the intervention execution module occupy 95% of the chip resources, and the main frequency is increased to 200MHz; the stimulation parameters of the intervention execution module meet the following requirements: the stimulation frequency is consistent with the current prefrontal cortex. The deviation of the oscillation frequency does not exceed ±5%; the stimulation intensity is mapped to a current range of 50 to 500 μA based on the input value of the pain level assessment scale through fuzzy logic rules; the dynamic window control module implements window adjustment through hardware logic, specifically including: using the shift register group embedded in the FPGA to store the signal data in the current window; and dynamically controlling the length of the shift register through a programmable counter to achieve continuous adjustment of the window length from 0.1 to 2 seconds.

[0029] In a preferred embodiment, the dynamic pain nerve decoding and intervention system also includes a safety control module for performing the following operations: monitoring the number of continuous high-frequency stimulations, and automatically switching to a baseline intervention mode if the pain level does not decrease after three stimulations. The baseline intervention mode has a stimulation frequency of 40 Hz and a stimulation intensity of 100 μA; immediately cutting off the stimulation output and triggering an alarm when the EEG signal amplitude is detected to exceed 200 μV; the signal acquisition module is compatible with the digital interface of Neuroscan equipment and BioSemi equipment, including: an SPI protocol conversion unit or an I2C protocol conversion unit for receiving raw EEG data from external devices; a signal resampling unit for uniformly converting the input signal to a 250 Hz sampling rate; and a frequency delay comparison table containing the following mapping relationships: when the oscillation frequency is between 30 and 40 Hz and the conduction velocity level is medium, the corresponding compensation delay is 20 ms; when the oscillation frequency is between 40 and 60 Hz and the conduction velocity level is fast, the corresponding compensation delay is 15 ms; when the oscillation frequency is between 60 and 80 Hz and the conduction velocity level is ultrafast, the corresponding compensation delay is 10 ms. To ensure stable operation within the power consumption requirements of wearable devices, the system integrates an intervention execution module with an event-driven controller, employing a dynamic power control strategy. When the energy ratio does not exceed a threshold, the system operates in low-power mode, using only the signal acquisition module and the dynamic window control module. The main frequency drops to 50 MHz, consuming only 5% of the total power. When the energy ratio exceeds a set threshold, the event-driven controller activates the phase prediction module and the intervention execution module, entering full-power mode and increasing the main frequency to 200 MHz. This ensures timely execution of pulse stimulation and optimal intervention accuracy. After the intervention is complete, the system automatically returns to a low-power state, ensuring optimal power consumption during long-term use. Furthermore, to enhance system safety, a safety control module is designed to monitor the number of consecutive high-frequency stimulations in real time. If the patient's pain level does not decrease after three stimulations, the system automatically switches to baseline intervention mode to prevent side effects caused by overstimulation. In addition, when the amplitude of the EEG signal exceeds 200μV, the system will immediately cut off the electrical stimulation and trigger an alarm to ensure patient safety. The system also uses an SPI or I2C protocol conversion unit and a signal resampling unit to ensure that data from different devices can be uniformly converted to a sampling rate of 250Hz, thereby achieving standardized operations and reducing modifications to existing equipment. These are all extended implementation methods that are known to ordinary technicians in this field.

[0030] Example 1: The signal acquisition module of this embodiment obtains the frontal lobe EEG signal in real time through the EEG device and divides the signal into Frequency band 4 to 8 Hz and There are two types of frequency bands: 30 to 80 Hz. In order to ensure high-quality signal acquisition, the system uses an analog filter bank to first separate the frequency bands and perform full-wave rectification on the separated signals to avoid signal distortion caused by noise interference. In the dynamic window control module, by calculating Frequency band and The instantaneous energy ratio of the frequency band signal is used to adjust the length of the analysis window. To ensure the accuracy of the analysis process, the system introduces a window length adaptive adjustment mechanism based on energy ratio calculation. and When the instantaneous energy ratio of the frequency band exceeds the preset threshold, the signal analysis window will be quickly compressed to a short-term window of 0.2 to 0.4 seconds. This helps to accurately capture transient features and avoid misjudgment caused by feature loss in traditional fixed windows.

[0031] Specifically, the energy ratio calculation is achieved through the following steps: Frequency band and The frequency band signals are full-wave rectified respectively, and the rectified signals are smoothed using a low-pass filter with a cutoff frequency of 2 Hz. The energy ratio of the smoothed signals is calculated:

[0032] ,

[0033] in, and Respectively Frequency band and Instantaneous energy of frequency band signal, energy ratio Directly affects the window duration adjustment strategy to ensure the responsiveness and accuracy of the signal analysis process to pain characteristics. Through this dynamic adjustment mechanism, the system can respond to sudden events during pain outbreaks in real time, avoiding the fixed duration window from missing key neural signals. Specifically, and Respectively Frequency band 4 to 8 Hz and The instantaneous energy of the signal in the frequency band 30 to 80 Hz. Indicates that within the current time window The energy of the frequency band signal is calculated as The band signal is full-wave rectified and smoothed using a low-pass filter with a cutoff frequency of 2 Hz, and its energy is calculated. Indicates that within the current time window The energy of the frequency band signal, the processing process and The frequency band is the same. Through this formula, we can deduce that the θ frequency band is The instantaneous energy ratio of the frequency band signal is used to dynamically adjust the duration of the signal analysis window based on this value. Specifically, when When the preset threshold is exceeded, the system compresses the window to 0.2 to 0.4 seconds to accurately capture the transient characteristics of pain.

[0034] At the same time, in order to ensure the efficiency and accuracy of the dynamic adjustment of the signal analysis window, the following implementation path can be adopted: First, the signal collected from the EEG device is Frequency band 4-8Hz and The frequency band 30-80Hz is separated. This process is achieved through an analog filter. The separated signal is full-wave rectified to obtain the instantaneous signal strength of each frequency band. Next, a first-order low-pass filter (cut-off frequency 2Hz) is used to smooth the rectified signal to remove high-frequency noise. Then, the following calculations are made: Frequency band and Energy of frequency band signal and The instantaneous energy ratio obtained through the above process is , as the basis for dynamically adjusting the signal analysis window length. If the value exceeds the preset threshold, the analysis window will be compressed to 0.2 to 0.4 seconds; if it is below the threshold, the window will be expanded to 0.8 to 1.6 seconds to adapt to different pain signal patterns.

[0035] The phase compensation module is based on a dual-factor mapping mechanism of individualized nerve conduction velocity and oscillation frequency. Each time the main frequency of the current prefrontal EEG signal is obtained from the frequency detection unit, the system calculates the corresponding phase compensation time based on the frequency and the individual's nerve conduction velocity using the frequency-delay comparison table. Specifically: Assuming the current oscillation frequency is Hz, the conduction velocity is "fast", then according to the frequency-delay comparison table, the system will find the corresponding compensation time, for example, 15ms. radians), the system will automatically adjust the phase compensation time, each adjustment step is 2ms, so as to ensure that each stimulation pulse matches the rising edge of the action potential of the target neural cluster. In order to solve the contradiction between real-time performance and low power consumption, a dynamic power consumption control mechanism can be adopted. The specific implementation steps are as follows: low power consumption mode: when the energy ratio is less than 100%, the system will automatically adjust the phase compensation time, each adjustment step is 2ms, so as to ensure that each stimulation pulse matches the rising edge of the action potential of the target neural cluster. When the threshold is not exceeded, the system only activates the signal acquisition module and the dynamic window control module, maintaining a low-power state. The main frequency is reduced to 50MHz, and resource utilization is only 5%. In this mode, the system continuously monitors signal changes to avoid unnecessary high power consumption. Full-Power Mode: When the energy ratio exceeds the threshold, the system activates the phase prediction module and the intervention execution module, entering full-power mode. The main frequency is increased to 200MHz, and resource utilization is reduced to 95%. In this mode, the system can quickly respond to pain signals and accurately trigger intervention pulses, ensuring timely and accurate treatment. This dynamic power control system can mobilize sufficient computing resources at critical moments while effectively reducing power consumption when signal analysis is less frequent, meeting the low-power and long-term operation requirements of wearable devices. The safety control module monitors the number of consecutive high-frequency stimulations in real time. If the pain level does not significantly decrease after three consecutive stimulations, the system automatically switches to the safe baseline intervention mode, which uses a stimulation frequency of 40Hz and an intensity of 100μA to avoid side effects caused by excessive stimulation. At the same time, when the amplitude of the EEG signal exceeds 200μV, the system will immediately cut off the electrical stimulation output and trigger an alarm to ensure the safety of the patient.

[0036] To address the phase mismatch problem caused by individual differences in nerve conduction velocity in existing technologies, this system has designed a personalized phase prediction mechanism. This mechanism, based on individualized nerve conduction velocity, combines a dual-factor mechanism of oscillation frequency and delay mapping to provide precise phase compensation. For example, the phase prediction module extracts the dominant frequency of the current prefrontal oscillation signal through a frequency detection unit and, based on the oscillation frequency and nerve conduction velocity level, retrieves the corresponding phase compensation time from a prestored frequency-delay comparison table. This phase compensation table is generated using extensive clinical data and a personalized conduction velocity model. After each intervention, if the actual phase deviation exceeds 𝜋 / 8 radians, the system automatically adjusts the corresponding compensation time. During this module's optimization process, the signal phase adjustment process no longer relies solely on fixed rules but instead incorporates real-time changes in individualized nerve conduction velocity and oscillation frequency, ensuring high synchronization between the intervention pulse and the action potential of the target neural population, thereby significantly improving the effectiveness of the intervention. These are all extended implementations known to those skilled in the art.

[0037] Example 2: See Figure 1 、 Figure 2 and Figure 3 , Figure 1 This is the structural diagram of the dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window of the present invention, showing the overall framework of the dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window. The system consists of multiple functional modules. The first is the signal acquisition module, which obtains EEG signals in real time through the EEG device and separates signals of different frequency bands through the analog filter group. Specifically, the signal is divided into Frequency band signal (4-8Hz) and Frequency band signal (30-80Hz). The signals of these two frequency bands are used for subsequent dynamic analysis and window adjustment. The output of the signal acquisition module is sent to the dynamic window control module, which includes Energy ratio calculation unit. This unit calculates Frequency band and The instantaneous energy ratio of the frequency band signal is calculated and the signal analysis window is dynamically adjusted based on the energy ratio. When the energy ratio exceeds a preset threshold, the window adjustment unit compresses the analysis window to a shorter duration, typically 0.2 to 0.4 seconds, to accurately capture the transient characteristics of pain. Otherwise, the analysis window is expanded to 0.8 to 1.6 seconds. After dynamic window adjustment, the signal is passed to the phase prediction module. Within this module, the frequency detection unit first extracts the dominant frequency of the prefrontal oscillation signal. This frequency, combined with the individualized nerve conduction velocity level, is then used by the delay mapping unit to determine the appropriate phase compensation time. This mechanism precisely adjusts the triggering timing of the stimulation pulse, ensuring precise synchronization with the rising edge of the action potential of the neural cluster. The signal, adjusted by the phase prediction module, then enters the intervention execution module, which generates the electrical stimulation pulse and adjusts its triggering timing in advance based on the phase compensation parameters, thereby achieving precise neural intervention. Finally, the system manages and controls the entire process through an event-driven controller. When the energy ratio does not exceed the threshold, the signal acquisition and dynamic window control modules maintain operation in low-power mode; when the energy ratio exceeds the threshold, the controller will activate the phase prediction module and the intervention execution module, enter the full-power mode, and restore the low-power state after the intervention. This not only improves the accuracy of pain signal capture, but also maintains high real-time performance under low-power conditions, providing an efficient and low-power solution for wearable devices.

[0038] Figure 2The comparison diagram of neural cluster stimulation triggering between the traditional method and the method of the present invention shows the comparison between the traditional method and the method of the present invention in neural cluster stimulation. In the traditional method, the stimulation signal is usually triggered by a fixed delay, and the delay time is relatively long, usually a pulse delay of 50ms. This fixed delay triggering method has a major defect and cannot accurately synchronize stimulation with neural activity, resulting in poor stimulation effect, especially in the accuracy and real-time performance of neural signal processing. In the technical solution of the present invention, the stimulation signal trigger adopts a more precise and flexible control method. Specifically, the present invention uses an advance compensation triggering technology, that is, through the real-time analysis of the neural signal, the stimulation signal is compensated in advance. This method can accurately trigger the electrical stimulation pulse when the deviation is less than ±π / 16 radians, ensuring that the stimulation pulse is completely synchronized with the action potential of the target neural cluster, thereby effectively improving the accuracy and efficiency of the intervention. And Figure 2 The neural clusters in Figure 3 demonstrate the application of the present invention at the neural population level. Comparison with traditional methods demonstrates significant improvements in accuracy, timeliness, and synchronization. This not only addresses the latency issues inherent in traditional methods, but also significantly enhances the personalization and precision of neural regulation, ensuring optimal pain intervention outcomes.

[0039] Figure 3 This diagram shows the structure and operation of the dynamic window control and phase prediction modules based on the Xilinx Zynq chip. First, the clinical EEG device receives signals via the SPI / I2C interface and communicates according to the SPI / I2C protocol. The input signals are then passed to the dynamic window control module, which has a 5% resource allocation and operates at a 50MHz frequency. Within this module, the signals are processed and, based on the dynamic window control algorithm, trigger corresponding event triggering mechanisms to accurately capture and analyze pain signals. After processing, the signals are output to the phase prediction module, which consumes 95% of the resources and operates at a 200MHz frequency. This module enables the system to predict the activity phase of neural clusters based on real-time changes in prefrontal cortex signals and perform phase compensation based on individualized conduction velocities, thereby improving the synchronization and accuracy of interventions. The diagram also shows the system's two operating modes: low-power mode (0.8V) and full-power mode (1.2V). In low-power mode, the system optimizes the use of computing resources to ensure low-power operation when signal processing is infrequent; in full-power mode, when higher-precision signal processing is required, the system can ensure accurate signal intervention and processing by increasing voltage and resource allocation.

[0040] Example 3: In this example, the frontal lobe EEG signal acquired by the signal acquisition module is separated into Frequency band 4-8Hz and After the frequency band 30-80Hz signal, the dynamic window control module is based on Frequency band and The energy ratio of the frequency band is used to adjust the duration of the analysis window. In the optimization scheme, the energy ratio calculation unit Frequency band and The frequency band signal is integrated over a sliding time window to calculate the instantaneous energy ratio. Specifically, we define the energy ratio as: ,in, and Respectively Frequency band and The instantaneous energy of the frequency band signal in the current window. Each time the energy ratio is calculated, the window size is dynamically adjusted according to the change of the energy ratio: when the energy ratio exceeds the preset upper threshold (for example, The signal analysis window will be compressed to a shorter duration (e.g. 0.2 to 0.4 seconds) if the energy ratio is lower than the set threshold, and the window will be extended to a longer duration (e.g. 0.8 to 1.6 seconds) to accommodate longer neural signal analysis.

[0041] The phase prediction module aims to eliminate timing mismatches caused by differences in nerve conduction velocity. To this end, this embodiment employs a dual-factor mapping mechanism combining individualized nerve conduction velocity and oscillation frequency. Unlike traditional techniques that employ simple fixed delay compensation, this system flexibly adjusts the phase compensation time based on the current oscillation frequency and individual nerve conduction velocity, ensuring millisecond-level synchronization between the stimulation pulse and the rising edge of the action potential of the target neural cluster. Specifically, the phase prediction module extracts the dominant frequency of the prefrontal oscillation signal using a frequency detection unit and searches for the corresponding phase compensation time using the following rules: a frequency-delay comparison table: Based on the nerve conduction velocity level (slow, medium, fast, ultrafast) and frequency, the corresponding phase compensation value is searched. For example, when the oscillation frequency is 40 Hz and the conduction velocity is "fast," the phase compensation is 10 ms; when the oscillation frequency is 60 Hz, the phase compensation is 5 ms. This mapping mechanism allows for phase compensation to be adjusted based on feedback after each intervention. Furthermore, after each electrical stimulation, the system self-calibrates phase deviation. If the actual deviation exceeds the set threshold (e.g., ±𝜋 / 8 radians), the system automatically adjusts the compensation time to ensure the accuracy of the next intervention.

[0042] To ensure long-term, stable operation of the system in a wearable device, this embodiment optimizes power consumption control. Using an event-driven controller, when the energy ratio does not exceed a threshold, the system only activates the signal acquisition module and dynamic window control module, maintaining a low-power mode (e.g., 5% resource utilization, 50MHz main frequency). When the energy ratio exceeds a set threshold, the phase prediction module and intervention execution module are activated, and the system switches to full-power mode (e.g., 95% resource utilization, 200MHz main frequency) to ensure precise timing of electrical stimulation pulses. During window adjustment, the system also implements a hardware-level dynamic resource allocation strategy to ensure real-time resource allocation during signal analysis, minimizing power consumption. After electrical stimulation is completed, the system returns to low-power mode, enabling long-term monitoring and low-power operation, ensuring patient comfort and wearability in daily life. Furthermore, considering patient safety, this embodiment introduces a safety control module to monitor the number of consecutive high-frequency stimulations. If pain levels do not decrease significantly after three consecutive stimulations, the system automatically switches to a safe baseline intervention mode, with a stimulation frequency of 40 Hz and an intensity of 100 μA. This measure avoids overstimulation and potential side effects. Furthermore, the signal acquisition module is designed to be compatible with existing medical equipment (such as Neuroscan and BioSemi), enabling seamless connection and receiving raw EEG data from external devices via SPI or I2C protocol conversion units. This design allows the system to quickly integrate with existing medical facilities without the need for additional hardware modifications, facilitating rapid deployment and application.

[0043] Example 4: In this example, the energy ratio It is through the frontal lobe EEG signals Frequency band 4-8Hz and The specific calculation process is as follows: Signal processing: The original EEG signal is separated into frequency bands by analog filter groups to obtain Frequency band signal and Frequency band signal Energy calculation: Full-wave rectification is performed on the separated signals, and the rectified signals are smoothed using a first-order low-pass filter with a cutoff frequency of 2 Hz to obtain the smoothed signal energy. Energy ratio definition: The energy ratio is defined as: ,in, and Respectively Frequency band and The instantaneous energy of the frequency band signal in the current time window, in units of power (for example, in μW). It is used to dynamically adjust the duration of the signal analysis window. By calculating the energy ratio, the system can respond to the transient characteristics of the pain outbreak in real time. When the preset threshold is exceeded, the system will shorten the signal analysis window to ensure that the key features of the pain outbreak period are captured and avoid misjudgment of traditional fixed window methods.

[0044] The present invention solves the problem that the traditional system cannot cope with the transient changes of pain characteristics by dynamically adjusting the signal analysis window length. Specifically, after the signal acquisition module obtains the frontal lobe EEG signal, the sliding time window is used to analyze the pain characteristics. Frequency band and The energy ratio of the frequency band signal is calculated. When the set threshold is exceeded, the system will compress the signal analysis window from the standard 0.8 to 1.6 seconds to 0.2 to 0.4 seconds in order to accurately capture the transient pain characteristics. When the threshold is below, the window expands to accommodate longer periods of neural signal analysis. This mechanism allows the system to flexibly respond to pain bursts of varying intensity and duration, ensuring that no critical pain signals are missed. This adjustment strategy also eliminates the smoothing interference of high-frequency oscillation phase transitions in fixed-duration window methods, effectively improving the capture rate of neural signals and significantly reducing feature loss.

[0045] In the phase prediction module of the present invention, a dual-factor mapping mechanism of individualized nerve conduction velocity and oscillation frequency can be used to accurately synchronize the stimulation pulse with the action potential of the target neural cluster. This mechanism extracts the dominant frequency of the prefrontal EEG signal through the frequency detection unit, and combines the individual's nerve conduction velocity level to find the corresponding phase compensation time from the pre-stored frequency-delay comparison table. Assuming that the current oscillation frequency is Hz, according to the different conduction velocities (such as slow, medium, fast, and ultrafast), find the corresponding phase compensation time (Unit is milliseconds), where the frequency is negatively correlated with the delay time. The specific mapping rule is: when the oscillation frequency is Hz, and the conduction velocity level is "fast", the compensation delay , when the oscillation frequency is Hz, and the conduction velocity grade is "slow", compensation delay After each electrical stimulation, the system will automatically calibrate the phase compensation parameters to ensure the accurate triggering of the next stimulation pulse, thereby ensuring the synchronization and accuracy of neural intervention. Radians, the system will set the step size Adjust the compensation parameters.

[0046] The core advantage of the system of the present invention lies in its dual optimization of real-time performance and low power consumption. When the signal acquisition module and dynamic window control module are working, the system operates in low-power mode, with a main frequency of only 50 MHz, consuming 5% of the total power. When the energy ratio exceeds the set threshold, the system activates the phase prediction module and the intervention execution module, enters full-power mode, and increases the main frequency to 200 MHz to ensure the precise execution of pulse stimulation. Through this dynamic power consumption control mechanism, the system can minimize power consumption without affecting the intervention effect, meeting the requirements of wearable devices. In addition, the system manages the operating status of each module through an event-driven controller, ensuring that high-power modules are only activated at critical moments, avoiding the waste of resources caused by long-term high-power operation.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window, characterized by: include: Signal acquisition module, used to obtain frontal lobe EEG signals in real time through EEG equipment and separate them using analog filter banks Frequency band signal and Frequency band signal, where The frequency band is 4 to 8 Hz, The frequency band is 30 to 80 Hz; A dynamic window control module is connected to the signal acquisition module and includes an energy ratio calculation unit for calculating the energy ratio of the signal. Frequency band signal and The frequency band signals are respectively integrated in a sliding time window to calculate the instantaneous energy ratio of the two; the window adjustment unit compresses the signal analysis window to a first preset time length when the instantaneous energy ratio exceeds a preset threshold, otherwise it is extended to a second preset time length, where the preset threshold is 1.2 to 1.8 times the average energy value of the frequency band, the first preset duration is 0.2-0.4 seconds, and the second preset duration is 0.8-1.6 seconds; a phase prediction module, connected to the dynamic window control module, comprising: a frequency detection unit for extracting the dominant frequency of the current prefrontal oscillation signal; a delay mapping unit for matching a corresponding phase compensation time parameter from a pre-stored frequency delay comparison table with different conduction velocity levels based on the dominant frequency and the individual nerve conduction velocity level; the phase compensation time parameter is used to represent the lead time of the stimulation pulse triggering moment relative to the rising edge of the action potential of the target neural cluster; an intervention execution module, connected to the phase prediction module, configured to trigger an electrical stimulation pulse during the dynamic window compression phase, and adjust the triggering timing of the stimulation pulse in advance to the rising edge of the action potential of the target neural cluster based on the phase compensation time parameter; The event-driven controller connects the signal acquisition module, the dynamic window control module, the phase prediction module and the intervention execution module, and is configured as follows: when the energy ratio does not exceed the threshold, only the low-power operation of the signal acquisition module and the dynamic window control module is maintained; when the energy ratio exceeds the threshold, the phase prediction module and the intervention execution module are activated to the full-power mode, and the low-power state is restored after the window compression phase ends.

2. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The energy ratio calculation unit is implemented in the following manner: Frequency band signal and The frequency band signals are full-wave rectified respectively; the rectified signals are smoothed by a first-order low-pass filter with a cutoff frequency of 2 Hz; the smoothed signals are calculated. Frequency band signal energy and The ratio of the signal energy in the frequency band.

3. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The frequency delay comparison table is generated by the following rules: based on the big data of nerve conduction velocity of healthy people and chronic pain patients, the baseline delay parameters corresponding to four different conduction velocity levels are preset, where the different conduction velocity levels include slow, medium, fast and ultra-fast; the baseline delay parameter is negatively correlated with the current oscillation frequency, and the delay decreases by 1.5ms for every 10Hz increase in frequency; the parameter is automatically calibrated after each intervention: if the actual phase deviation exceeds If the angle is radian, adjust the corresponding parameter in the table with a step size of ±2ms.

4. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The power consumption control of the event-driven controller includes: in low-power mode, the signal acquisition module and the dynamic window control module occupy 5% of the chip resources, and the main frequency is reduced to 50MHz; in full-power mode, the phase prediction module and the intervention execution module occupy 95% of the chip resources, and the main frequency is increased to 200MHz.

5. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The stimulation parameters of the intervention execution module meet the following requirements: the stimulation frequency is consistent with the current prefrontal cortex. The deviation of the oscillation frequency did not exceed ±5%; the stimulation intensity was mapped to a current range of 50 to 500 μA based on the input value of the pain level assessment scale through fuzzy logic rules.

6. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The dynamic window control module implements window adjustment through hardware logic, specifically including: using the shift register group embedded in the FPGA to store the signal data in the current window; dynamically controlling the length of the shift register through a programmable counter to achieve continuous adjustment of the window length from 0.1 to 2 seconds.

7. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The dynamic pain nerve decoding and intervention system also includes a safety control module, which is used to perform the following: monitoring the number of continuous high-frequency stimulations, and automatically switching to a baseline intervention mode if the pain level does not decrease after three stimulations. The stimulation frequency of the baseline intervention mode is 40 Hz and the stimulation intensity is 100 μA; when the EEG signal amplitude is detected to exceed 200 μV, the stimulation output is immediately cut off and an alarm is triggered.

8. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 1, characterized in that: The signal acquisition module is compatible with the digital interfaces of Neuroscan and BioSemi devices, and includes: an SPI protocol conversion unit or an I2C protocol conversion unit for receiving raw EEG data from external devices; and a signal resampling unit for uniformly converting the input signal to a 250Hz sampling rate.

9. The dynamic pain neural decoding and intervention system based on the prefrontal temporal feature window according to claim 3, characterized in that: The frequency delay comparison table contains the following mapping relationship: when the oscillation frequency is between 30 and 40 Hz and the conduction velocity level is medium, the corresponding compensation delay is 20 ms; when the oscillation frequency is between 40 and 60 Hz and the conduction velocity level is fast, the corresponding compensation delay is 15 ms; when the oscillation frequency is between 60 and 80 Hz and the conduction velocity level is ultra-fast, the corresponding compensation delay is 10 ms.

Images