A frozen gait neural feature recognition and stimulation parameter feedback regulation system based on gait cycle analysis

Abstract

The application discloses a freezing gait neural feature recognition and stimulation parameter feedback regulation system based on gait cycle analysis, which comprises a data synchronous acquisition module, a gait cycle division module, a freezing related abnormal electrophysiological feature recognition module and a gait phase specificity stimulation parameter feedback regulation module. The system takes the gait cycle as an analysis unit, divides the cycle into a support phase, a swing phase and a double support phase; takes the standing state local field potential feature and the gait cycle mean as double baseline references, carries out standardized analysis on the cortex power and the cortex-thalamic bottom nucleus coherence, recognizes the freezing risk related abnormal neural features in the non-freezing stable walking stage; and generates and switches different frequency band thalamic bottom nucleus deep brain electrical stimulation parameters according to the joint determination result of the current gait phase and the abnormal neural features. The application expands the freezing risk recognition and intervention window to the non-freezing stable walking stage, and realizes the objective quantization optimization of the stimulation parameters.

Description

Technical Field

[0001] This invention relates to the field of neurophysiological signal analysis and deep brain stimulation technology, specifically to a frozen gait neural feature recognition and stimulation parameter feedback regulation system and method based on gait cycle analysis, which is used for freezing risk identification in Parkinson's disease patients with frozen gait during the non-frozen stable walking phase and for gait phase-specific optimization adjustment of deep brain stimulation parameters in the subthalamic nucleus. Background Technology

[0002] Parkinson's disease (PD) is a common neurodegenerative disease affecting the elderly. Typical symptoms include resting tremor, rigidity, bradykinesia, and postural and gait disturbances. Regarding gait disturbances, patients typically exhibit slowed walking speed, shortened stride length, increased gait frequency, and increased gait variability, gradually developing frozen gait (FOG) as the disease progresses. Current clinical studies show that the incidence of gait disturbances increases significantly with disease duration, reaching 58% in patients with a disease duration of over 10 years and exceeding 80% in those with a disease duration of over 15 years. Falls caused by gait disturbances have become the leading cause of disability and the fourth leading cause of death in PD patients. Therefore, effectively identifying and intervening in gait disturbances, especially frozen gait, in PD patients has become a pressing technical problem to be solved in this field.

[0003] Currently, there is still a lack of ideal treatments for gait disorders in Parkinson's disease (PD). While levodopa, as a first-line drug for PD, can improve gait speed and stride length, long-term use can easily lead to "on / off" phenomena and dyskinesia, affecting gait stability. Subthalamic nucleus (STN)-deep brain stimulation (DBS) is a commonly used surgical treatment for mid-to-late-stage PD. Traditional high-frequency STN-DBS (130-185Hz) can effectively relieve core motor symptoms such as tremor and rigidity and reduce drug side effects; however, its improvement in gait disorders varies significantly among individuals, with some patients even experiencing symptom exacerbation. Low-frequency stimulation (60-90Hz) helps improve gait, but its effect on overall motor symptoms is weak, making long-term use alone difficult, and its long-term efficacy is uncertain. Existing DBS treatment regimens have limited effectiveness in improving gait disorders, and the selection of stimulation parameters lacks objective basis.

[0004] Although existing studies have reported many electrophysiological indicators that reflect FOG, the construction of individualized neural modulation or closed-loop modulation schemes still faces bottlenecks. First, studies on the electrophysiological characteristics of FOG mostly focus on the transient critical state (about 1.5s) at or before freezing occurs, and such a short time window poses a high technical challenge to precise modulation. Second, although some studies have found high phase-amplitude coupling PAC in the beta-gamma band of the primary motor cortex in FOG patients during stable walking, PAC analysis usually requires a relatively long time window of 10-30s to achieve stable calculation. Third, there is currently a lack of behavioral indicators that can reflect the risk of freezing during the stable walking phase.

[0005] Gait cycles, as the fundamental spatiotemporal unit of human walking, accurately reflect the dynamic evolution of rhythmic gait patterns. Existing studies have confirmed that gait cycles are a key indicator distinguishing between non-freezing gait (nFOG) and FOG patients; FOG patients exhibit prolonged support phases and shortened swing phases. The classic "threshold model" theory also indicates a progressive shortening of stride length over three consecutive gait cycles before freezing occurs. Recent animal experiments have further revealed that, compared to holistic measurements such as whole-body velocity, single-limb gait phase-locking analysis is more sensitive in revealing phase-locking imbalances in D1 / D2 MSNs caused by dopamine depletion. These results suggest that conventional full gait cycle averaging analysis may mask abnormalities in dynamic neural activity during critical phases of gait alternation, and the identification of FOG-related biomarkers urgently requires more sensitive gait phase analysis methods.

[0006] In summary, the existing technology has at least the following defects: (i) the effective time window for freezing gait electrophysiological markers is too short, making it difficult to achieve precise real-time control; (ii) there is a lack of reliable behavioral and neurophysiological indicators that reflect the risk of freezing during the stable walking phase without freezing; (iii) the STN-DBS stimulation parameters mainly rely on clinical experience for trial-and-error adjustment, lacking quantitative optimization methods based on objective neural characteristics.

[0007] Therefore, it is necessary to provide a system and method that can stably identify freezing risk during the non-freezing stable walking phase and output gait phase-specific STN-DBS stimulation parameters accordingly, in order to overcome the above-mentioned technical defects. Summary of the Invention

[0008] To address the technical problems in the prior art, this invention uses gait cycle as the basic analysis unit and aims to provide a system and method that can stably identify freezing risk during the non-freezing stable walking phase and output gait phase-specific STN DBS stimulation parameters.

[0009] In a first aspect, the present invention provides a frozen gait neural feature recognition and stimulation parameter feedback control system based on gait cycle analysis, comprising:

[0010] The data synchronization acquisition module is configured to acquire three-dimensional gait kinematic data and local field potential (LFP) signals of the bilateral subthalamic nuclei (STN) and target cortical brain regions, which are synchronously recorded during the subject's walking process. The gait kinematic data and the LFP signals are aligned in the time dimension.

[0011] The gait cycle segmentation module is configured to take the initial heel strike of one side to the second heel strike of the same side as a complete gait cycle, and to divide each gait cycle into a support phase, a swing phase, and a double support phase within the support phase;

[0012] The module for identifying abnormal electrophysiological features related to freezing is configured to use the LFP features in the standing state and the mean of the gait cycle as two types of baseline references to perform a standardized analysis of the cortical power and cortical-STN coherence in the complete gait cycle and its various phases, and to identify abnormal neural features related to freezing risk in the non-freezing stable walking phase based on the standardized analysis.

[0013] The gait phase-specific stimulation parameter feedback control module is configured to receive the current gait phase output by the gait cycle division module and the abnormal neural feature recognition result output by the freeze-related abnormal electrophysiological feature recognition module, and generate and switch STN deep brain electrical stimulation parameters of different frequency bands based on the joint judgment result of the two.

[0014] In the system, the data synchronization acquisition module is connected to the gait cycle segmentation module and the freeze-related abnormal electrophysiological feature recognition module to provide gait kinematic data and local field potential signals; the gait cycle segmentation module is connected to the freeze-related abnormal electrophysiological feature recognition module and the gait phase-specific stimulation parameter feedback control module to provide gait cycle segmentation results and the current gait phase; the freeze-related abnormal electrophysiological feature recognition module is connected to the gait phase-specific stimulation parameter feedback control module to provide abnormal neural feature recognition results.

[0015] In one implementation, the gait cycle segmentation module can be configured to take the first left heel strike (LHS) to the next left heel strike (LHS2) as a complete gait cycle. Based on five key events—LHS, right toe lift-off (RTO), right heel strike-off (RHS), left toe lift-off (LTO), and LHS2—the support phase is further subdivided into a first double support phase (DS1), a single support phase (SS), and a second double support phase (DS2), with DS1 and DS2 being merged into the double support phase.

[0016] In a preferred embodiment, the gait cycle segmentation module is further configured to divide the gait trials collected during the drug off state into frozen trials (FOGT) and non-frozen trials (nFOGT), wherein the frozen trials are trials in which at least one freezing episode occurs, and the non-frozen trials are trials in which no freezing episode occurs.

[0017] In a further implementation, the freeze-related abnormal electrophysiological feature identification module can be configured to compare the non-freezing walking gait cycle in the freeze trial with the walking gait cycle in the non-freezing trial within the same subject. By comparing the differences in gait features and electrophysiological features between the two types of trials, freeze-related abnormal neural features can be identified, thereby extending the freeze risk identification window from the instantaneous critical state of freeze onset to the non-freezing stable walking stage.

[0018] In a further implementation, the gait cycle segmentation module can also be configured to: perform turning segment exclusion processing and freeze event detection processing before performing gait cycle segmentation, and retain only the gait cycle data during straight walking and non-freezing segments for subsequent electrophysiological feature recognition.

[0019] In one implementation, the gait cycle segmentation module can also be configured to calculate the following gait parameters for each gait cycle: gait space parameters, including stride length, stride speed, and stride velocity; gait time parameters, including cadence, stride duration, and stride length; and gait cycle component parameters, including double support time, single support time, support time, and swing time, wherein the gait cycle component parameters characterize the proportion of the corresponding phase duration to the total duration of the gait cycle; wherein the parameters related to stride length and stride frequency use the mean of two-sided data to reduce the influence of one-sided differences.

[0020] In one implementation, the freeze-related abnormal electrophysiological feature identification module may include a parallel full gait cycle analysis unit and a gait phase-specific analysis unit. The full gait cycle analysis unit uses the mean LFP power spectrum of the subject in the standing state as a baseline to perform z-score normalization on cortical power and cortical-STN coherence within a complete gait cycle, revealing the overall energy and connectivity changes during walking relative to the static standing state. The gait phase-specific analysis unit uses the mean gait cycle as a baseline to perform z-score normalization on cortical power and cortical-STN coherence within the support phase, swing phase, and double support phase, to reveal the relative dynamic changes between different phases within the same gait cycle.

[0021] Specifically, a correlation analysis is performed between the output of the full gait cycle analysis unit and the gait space parameters, and a correlation analysis is performed between the output of the gait phase specificity analysis unit and the gait cycle component parameters. Both correlation analyses are performed after averaging across gait cycles and trials. The integrated correlation output is ultimately used as a correlation map between abnormal gait parameters and abnormal electrophysiological characteristics related to the risk of freezing, providing a decision-making basis for the feedback regulation module.

[0022] Furthermore, the freeze-related abnormal electrophysiological feature identification module can be configured to perform the following preprocessing and time-frequency analysis steps: sequentially performing irrelevant channel deletion, bandpass filtering, power frequency notch filtering, stimulus-related artifact and harmonic notch filtering, downsampling, adjacent electrode bipolar reference, and artifact removal on the acquired LFP data; performing time-frequency analysis on the preprocessed LFP data using wavelet transform, and synchronizing it with gait key events in time, dividing it into complete gait cycles and their respective phases; aligning the key event time points of each gait cycle through linear time deformation, and performing time resampling using cubic spline interpolation to ensure waveform smoothness and time consistency.

[0023] Specifically, the freeze-related abnormal electrophysiological feature identification module can divide the analysis frequency band into four bands: theta, alpha, lowbeta, and highbeta, and use the lowbeta power and theta coherence as the target frequency band for freeze-related abnormal neural features. The cortical-STN coherence can be calculated by performing continuous wavelet transform (CWT) on paired ipsilateral cortical electrode and STN electrode signals, and outputting a coherence value in the range of 0 to 1 after smoothing. The subsequent processing of the coherence data is the same as that of the power spectrum analysis.

[0024] In some embodiments, the freeze risk-related abnormal neural features include at least one or more of the following features: (1) a phase-dependent fluctuation abnormality of lowbeta power in the primary somatosensory cortex (hereinafter sometimes referred to as the S1 brain region or S1); (2) an abnormality of lowbeta power in the superior parietal lobule (hereinafter sometimes referred to as the SPL brain region or SPL) during the double support phase; and (3) an abnormality of the inverse regulation of S1-STN theta coherence.

[0025] Specifically, feature (1) is characterized by an anomalous fluctuation pattern in lowbeta power, which is significantly reduced during the support phase and increased during the swing phase. Feature (2) is characterized by a significant decrease in lowbeta power during the double support phase. It should be noted that the superior parietal lobule (SPL) plays a role in spatial perception, proprioceptive integration, and motor planning in gait control. Its lowbeta power decreases during the double support phase (i.e., the critical biomechanical phase in which both feet land simultaneously and the center of gravity shifts from the hind foot to the forefoot), reflecting an abnormal weakening of sensorimotor integration function in FOG patients during the center of gravity shift. This abnormality independently constitutes a neurological marker of freezing risk. Furthermore, 130Hz high-frequency STN stimulation can specifically restore the lowbeta power of SPL during the double support phase and simultaneously achieve a significant reduction in the double support time, verifying the functional significance of this feature from the perspective of intervention effect. Feature (3) is characterized by a reverse regulation pattern in which theta band coherence is significantly enhanced during the support phase and significantly weakened during the swing phase.

[0026] The three types of abnormal neural features mentioned above can only be revealed through phase-specific analysis using the mean of the gait cycle as a baseline. Conventional analysis using a single full gait cycle baseline would be masked by the averaging effect, which also indirectly verifies the technical necessity of the dual-baseline parallel architecture of this invention. In a specific embodiment, the freeze-related abnormal electrophysiological feature identification module can also be configured to extract freeze risk-related abnormal gait parameters. These abnormal gait parameters characterize the significant differences in stride length, stride speed, step length, and stride velocity between the non-freeze trial nFOGT and the frozen trial FOGT, as well as the significant prolongation in double support time, thereby constituting behavioral indicators that corroborate the abnormal neural features.

[0027] Considering the significant stimulation artifacts in STN signals during electrical stimulation, in some implementations, the system may further include a levodopa state comparison module, which is configured to record the theta coherence of S1-STN in the levodopa "off" and "on" states, respectively. The "weakened theta coherence in the support phase and enhanced theta coherence in the swing phase" in the ON state is used as the threshold reference for the freeze-related abnormal electrophysiological feature identification module to determine theta coherence inverse regulation abnormality, thereby providing an objective and quantifiable basis for the identification of abnormal neural features.

[0028] In one embodiment, the gait phase-specific stimulation parameter feedback modulation module may include:

[0029] The low-frequency modulation unit is configured to output 60 Hz low-frequency STN stimulation parameters in the support phase and the swing phase. The 60 Hz low-frequency stimulation parameters are set to increase the lowbeta power of the S1 brain region in the support phase and decrease the lowbeta power of the S1 brain region in the swing phase, so that the cross-gait phase modulation of the lowbeta power of the S1 brain region is negatively correlated with the corresponding gait cycle component parameters.

[0030] The high-frequency modulation unit is configured to output a 130 Hz high-frequency STN stimulation parameter in the dual support phase, the 130 Hz high-frequency stimulation parameter being set to specifically enhance the lowbeta power of the SPL brain region in the dual support phase; the 130 Hz high-frequency stimulation is triggered by a decrease in the lowbeta power of the SPL brain region in the dual support phase.

[0031] The switching control unit is configured to switch between the low-frequency adjustment unit and the high-frequency adjustment unit based on the current gait phase output in real time by the gait cycle division module and the abnormal neural feature recognition results output by the freeze-related abnormal electrophysiological feature recognition module in the corresponding phase.

[0032] The technical rationale for the above-mentioned 130 Hz high-frequency stimulation parameter setting lies in the sensory-motor integration function of SPL during the center of gravity transfer process in the double support phase. The 130 Hz high-frequency stimulation restores the lowbeta power of SPL in this phase by regulating the cortical-basal ganglia circuit, thereby improving the neural coordination mechanism of center of gravity transfer and shortening the double support time.

[0033] Furthermore, the switching control unit can be configured to execute the following triggering logic: when the current gait phase is the support phase or the swing phase, and an abnormal phase-dependent fluctuation of lowbeta power in the S1 brain region and / or an abnormal inverse regulation of S1-STN theta coherence are detected, the low-frequency modulation unit is triggered to output a 60 Hz stimulation parameter; when the current gait phase is the double support phase, and an abnormality of decreased lowbeta power in the SPL brain region is detected, the high-frequency modulation unit is triggered to output a 130 Hz stimulation parameter; the switching of the stimulation parameter is performed at the transition between adjacent gait phases.

[0034] In one specific implementation, the pulse width of both the 60 Hz low-frequency stimulation and the 130 Hz high-frequency stimulation can be fixed at 60 μs, and the stimulation intensity can be individually calibrated according to the current value corresponding to the subject's optimal gait improvement.

[0035] In one implementation, the data synchronization acquisition module may include a gait data acquisition submodule and an electrophysiological data acquisition submodule, wherein the gait data acquisition submodule and the electrophysiological data acquisition submodule achieve time synchronization through synchronization pulses.

[0036] Specifically, the gait data acquisition submodule may include an optical motion capture system and multiple marker sensors. These marker sensors can be deployed on the following parts of the subject's body: feet, heels, calves, thighs, and waist. Foot markers located on the lateral side of the heels and fifth metatarsals of both feet can be used to calibrate key gait events such as LHS, RTO, RHS, LTO, and LHS2. Those skilled in the art can also use inertial measurement units (IMUs), plantar pressure sensors, or other kinematic acquisition devices capable of accurately capturing heel strike and toe lift-off events instead of the aforementioned optical motion capture system.

[0037] In one specific implementation, the cortical electrode connected to the electrophysiological data acquisition submodule can be a multi-contact strip electrode, implanted along the cortical surface and sequentially covering the anterior motor cortex (PMC), primary motor cortex M1, primary somatosensory cortex S1, and superior parietal lobule SPL region from front to back, and records containing at least two adjacent contacts in each cortical region are included in the analysis; the connected bilateral STN electrodes are multi-contact DBS electrodes.

[0038] In a preferred embodiment, the contact diameter of the cortical strip electrode is 4 mm, the exposed diameter is 2.5 mm, and the contact spacing is 10 mm. Those skilled in the art can implement this invention using other multi-contact cortical electrodes and STN DBS electrode configurations according to actual clinical needs.

[0039] In a second aspect, the present invention provides a method for frozen gait neural feature recognition and stimulation parameter regulation based on gait cycle analysis, which is executed using the system described in the first aspect and includes the following steps:

[0040] S100, simultaneously acquire the subject’s three-dimensional gait kinematics data and LFP signals of bilateral STN and target cortical brain regions, with the two types of data aligned in the time dimension;

[0041] S200, taking the initial heel strike of one side to the second heel strike of the same side as a complete gait cycle, divides each gait cycle into a support phase, a swing phase, and a double support phase, and divides the gait tests collected during the drug-freeze state into frozen tests FOGT and unfrozen tests nFOGT.

[0042] S300 uses the LFP characteristics in the standing state and the mean of the gait cycle as two types of baseline references to perform a standardized analysis of the cortical power and cortical-STN coherence in the complete gait cycle and its various phases.

[0043] S400, based on the standardized analysis results obtained from S300, compares the non-frozen walking gait cycle in the frozen trial FOGT with the walking gait cycle in the non-frozen trial nFOGT within the same subject, and identifies the abnormal neural features related to the freezing risk based on the differences in gait and electrophysiological features between the two types of trials.

[0044] S500 generates corresponding STN deep brain stimulation parameters based on the combined judgment results of the current gait phase and the abnormal neural features identified by S400.

[0045] S600 switches the generated stimulus parameters at the transition between adjacent gait phases.

[0046] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the method described in the second aspect.

[0047] It should be noted that the various implementation schemes and any combinations thereof in the first aspect are equally applicable to the methods of the second aspect and the computer-readable storage media of the third aspect.

[0048] The beneficial effects of the present invention include at least the following aspects:

[0049] 1. This invention uses the gait cycle as the analysis unit, achieving refined extraction of gait features and gait phase-specificity of neural feature extraction. By examining the dynamic evolution of lowbeta power and theta coherence in the support, swing, and double support phases, it effectively avoids the masking of key phase information by conventional full-cycle averaging analysis. For example, this invention found that the lowbeta power in the S1 brain region exhibits an abnormal phase-dependent fluctuation pattern in FOGT, characterized by a significant decrease in the support phase and an increase in the swing phase, which is masked by full-cycle averaging analysis. Simultaneously, it discovered a reverse regulation pattern in FOGT: significantly enhanced S1 STN theta coherence in the support phase and significantly weakened S1 STN theta coherence in the swing phase. These phase-specific findings provide new electrophysiological insights into the mechanism of gait freezing. Furthermore, this invention also found a specific decrease in SPL lowbeta power in the double support phase of FOGT patients, suggesting an abnormality in the sensorimotor integration function of the superior parietal lobule during the center-of-gravity shift phase. This abnormality can be specifically corrected by 130 Hz high-frequency STN stimulation, accompanied by synchronous improvement in the double support time, providing an electrophysiological basis for gait phase-specific high-frequency stimulation protocols.

[0050] 2. This invention creatively employs different baseline references (the mean power spectrum of the standing state and the mean gait cycle, respectively) for full gait period analysis and gait phase-specific analysis. This allows the same LFP data segment to simultaneously output overall energy information for "walking vs. stationary" and local dynamic information for "phase vs. phase," effectively avoiding the technical defect of mutual masking between the two types of information under a single baseline strategy. This design is the technical prerequisite for achieving the key discovery that "full gait period analysis did not find M1 STN coherence differences, but phase-specific analysis revealed the inverse regulation of S1 STN theta coherence."

[0051] 3. This invention proposes an extended therapeutic window for intervention during stable non-freezing episodes, aiming to correct abnormal gait rhythms to prevent them from progressing to freezing, overcoming the technical bottleneck of the existing 1.5-second instantaneous critical state time window before freezing occurs being too short.

[0052] 4. Compared to the current technology that mainly relies on physician experience for programming through a trial-and-error approach, and on gait assessment based on limited clinical scales such as the Unified Parkinson's Disease Rating Scale and the Gait Freeze Questionnaire, this invention shifts the optimization of stimulation parameters from experience-driven to neural feature-driven, reducing the human and time costs of clinical programming. Specifically, this invention found that 60 Hz low-frequency stimulation can specifically restore the lowbeta power regulation across gait phases in the S1 brain region and is significantly negatively correlated with gait cycle parameters (stance phase ρ=-0.821, P=0.034; swing phase ρ=-0.893, P=0.012), while 130 Hz high-frequency stimulation specifically enhances the power of SPL in the dual stance phase, thus providing objective and quantitative parameter recommendations for clinical programming. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in this invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0054] Figure 1 This is a flowchart of the overall analysis process of the system of the present invention, showing the complete data processing path from the acquisition of raw LFP and gait data to dual baseline normalization analysis and final correlation integration.

[0055] Figure 2 This is a schematic diagram of the postoperative reconstruction and positioning of the DBS electrode and the cortical strip electrode in this invention. The left figure shows the reconstruction of the right cerebral hemisphere cortex, including three anatomical markers: M1 (primary motor cortex, marked with orange dots), S1 (primary somatosensory cortex, marked with blue dots), and SPL (superior parietal lobule, marked with green dots). The right figure shows the three-dimensional reconstruction of the nuclei surrounding the bilateral STN and the DBS electrode implantation trajectory.

[0056] Figure 3 This diagram illustrates the precise division of the key gait cycle events LHS, RTO, RHS, LTO, and LHS2 in this invention, as well as the vertical displacement signals of the lower limb markers corresponding to the four times DS1, SS, DS2, and SW.

[0057] Figure 4 This is a representative schematic diagram of the local field potentials of the cortex and the STN of the subthalamic nucleus, which are synchronously acquired by the 64-channel EEG acquisition system during the walking process of this invention.

[0058] Figure 5 This is a schematic diagram comparing the time-frequency plot of the full gait cycle and the time-frequency plot of the gait phase under the dual baseline standardization strategy of the present invention. The left figure uses the standing LFP as the baseline, and the right figure uses the mean of the gait cycle as the baseline.

[0059] Figure 6 Box plots comparing abnormal gait parameters between frozen gait trials (FOGT) and unfrozen gait trials (nFOGT), including eight parameters: stride length, stride speed, stride length, stride velocity, support time, swing time, double support time, and single support time.

[0060] Figure 7 Scatter plots showing the correlation between the standardized lowbeta power spectral density of the primary motor cortex throughout the gait cycle and stride length, stride speed, stride length, and stride velocity, including FOGT, nFOGT grouped regression, and merged sample regression.

[0061] Figure 8 This diagram shows the abnormal power of the primary somatosensory cortex during the gait cycle and its correlation with gait indices. Figure 8 A represents the quantization results of the time-frequency map and violin plot of the primary somatosensory cortex. Figure 8 B is a scatter regression plot of the normalized lowbeta power of the oscillation / support phase of the primary somatosensory cortex versus the oscillation / support time.

[0062] Figure 9 This is a theta coherence map of STN abnormalities in the primary somatosensory cortex during the gait cycle. Figure 9 A is the time-frequency diagram of coherence. Figure 9 B represents the violin plot quantification result of the theta coherence between the support phase and the oscillating phase.

[0063] Figure 10 Radar charts showing the improvement effects of DBS STN at different frequencies on gait parameters and freezing indexes, comparing multidimensional gait parameters and freezing indexes under four conditions: FOGT, nFOGT, HFS, and LFS.

[0064] Figure 11This is a diagram showing the phase-specific modulation effect of DBS STN at different frequencies on abnormal electrophysiological features, including time-frequency diagrams of gait phases in the primary somatosensory cortex and superior parietal lobule under three conditions: FOGT, HFS, and LFS, as well as scatter regression diagrams showing the correlation between the corresponding lowbeta power changes and gait phase parameters.

[0065] Figure 12 This diagram illustrates the phase-specific modulation effect of levodopa on the STN circuit of the primary somatosensory cortex in gait. Figure 12 A represents the time-frequency diagram of wavelet coherence between FOGT and the STN wavelet in the primary somatosensory cortex under ON drug-induced states. Figure 12 B represents the violin plot quantification result of the theta coherence between the support phase and the oscillating phase. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments, and they should not be construed as limiting the present invention.

[0067] This invention uses gait cycles as the basic analytical unit. Through standardized analysis with dual baseline references, it stably identifies abnormal neural features related to freezing risk during the stable walking phase without freezing. Based on the joint determination results of gait phase and abnormal neural features, it performs feedback regulation of STN deep brain stimulation parameters, thereby overcoming the technical bottlenecks in the prior art, such as the short time window of freezing electrophysiological markers, the lack of freezing risk behavioral indicators during the stable walking phase, and the reliance on clinical experience for trial-and-error adjustment of DBS-STN parameters.

[0068] Therefore, this invention provides a frozen gait neural feature recognition and stimulation parameter feedback control system based on gait cycle analysis, comprising four core modules:

[0069] 1. The data synchronization acquisition module is configured to simultaneously acquire three-dimensional gait kinematic data and local field potential (LFP) signals of the bilateral subthalamic subnucleus (STN) and target cortical brain regions during the subject's walking process, and to strictly align the gait kinematic data with the LFP signals in the time dimension, providing a data basis for the accurate correspondence between gait phase and neural activity in the future.

[0070] 2. The gait cycle segmentation module is configured to take the initial heel strike of one side to the second heel strike of the same side as a complete gait cycle, and to divide each gait cycle into a support phase, a swing phase, and a double support phase within the support phase, so that the gait phase becomes the unified spatiotemporal coordinate system for all subsequent electrophysiological feature extraction and stimulation parameter generation in this invention.

[0071] 3. The module for identifying abnormal electrophysiological features related to freezing is configured to use LFP features in the standing state and the mean gait cycle as two baseline references. It performs standardized analysis on cortical power and cortical-STN coherence within the complete gait cycle and its various phases, and identifies abnormal neural features related to freezing risk during the non-freezing stable walking phase based on this standardized analysis. This dual-baseline reference design allows the same LFP data to simultaneously output overall energy information (walking vs. stationary) and local dynamic information (phase vs. phase), effectively avoiding the technical defect of mutual masking between the two types of information under a single baseline strategy.

[0072] 4. The gait phase-specific stimulation parameter feedback control module is configured to receive the current gait phase output by the gait cycle division module and the abnormal neural feature recognition result output by the freeze-related abnormal electrophysiological feature recognition module. Based on the joint judgment result of the two, it generates and switches STN deep brain electrical stimulation parameters of different frequency bands, thereby forming a complete closed loop of "recognition-decision-feedback", so that the optimization of stimulation parameters shifts from experience-driven to neural feature-driven.

[0073] The four modules described above form a vertically integrated data processing link through four key technical nodes: simultaneous acquisition of dual-source data, unified gait cycle segmentation, standardized dual-baseline analysis, and joint phase-feature determination. This invention provides various optional implementation schemes regarding gait cycle segmentation granularity, specific algorithm implementation for dual-baseline analysis, specific types of abnormal neural features, specific frequency bands and triggering rules for feedback control, and hardware implementation configuration. However, the scope of this invention is not limited to these specific implementation schemes; those skilled in the art can select or combine them according to specific clinical scenarios and equipment conditions.

[0074] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0075] 1. Subject selection

[0076] Subjects were selected based on the following inclusion criteria: (1) patients with primary PD diagnosed according to the UK Brain Bank criteria; (2) aged 55–80 years; (3) clinical FOG successfully induced by at least one experienced neurologist specializing in movement disorders; and (4) FOG of moderate severity with a Gait Freeze Questionnaire score >10. Exclusion criteria: (1) inability to walk independently without medication; (2) severe cognitive impairment preventing cooperation; and (3) significant tremor (any one of the UPDRS 3.15–3.18 ≥3).

[0077] This embodiment initially evaluated 16 patients with FOG (Forward-of-Groogection). Eight patients were excluded due to inability to identify gait cycles (n=2), lack of nFOG trials (n=2), and lack of FOG trials (n=4), leaving 8 FOG patients (4 males and 4 females). These 8 patients sequentially completed three conditions: OFF, HFS (130Hz), and LFS (60Hz) (n=8 each), with 5 patients further completing the ON condition (n=5). The number of gait cycles identified under each condition was 327, 217, 264, and 140, respectively.

[0078] 2. Surgical procedure

[0079] See Figure 2 Under local anesthesia, a Leksell stereotactic head frame (ElektaInstrument AB) was installed along the anterior-posterior commissure line. Preoperative T2-weighted MRI was fused with intraoperative head CT to determine the bilateral STN coordinates. Bilateral four-contact DBS electrodes (model L301C, Pins Medical) were implanted under stereotactic guidance. At the same burr hole, an eight-contact strip electrode (Sinovation, Beijing, China; contact diameter 4mm, exposure diameter 2.5mm, contact spacing 10mm) was implanted into the right cortex along the superior sagittal sinus, covering the M1, S1, and SPL regions sequentially from anterior to posterior. Both bilateral STN electrodes and the right cortical electrode were led out via external leads for easy recording of electrophysiological signals during postoperative walking. Postoperatively, head CT was fused with preoperative T1-weighted MRI. The LeadDBS toolkit and Freesurfer software were used to reconstruct and locate both types of electrodes. No significant displacement of the bilateral DBS-STN electrodes was observed in all patients included in the study.

[0080] 3. Gait data acquisition

[0081] Gait testing began 3–5 days after electrode implantation. Three-dimensional gait data were acquired using the Codamotion optical motion capture system (CODAMotion, Charnwood Dynamics Ltd.) at a sampling rate of 100 Hz or 200 Hz. Patients wore 22 marker sensors throughout their body (2 on the feet, 2 on the heels, 8 on the calves, 8 on the thighs, and 2 on the waist). Four foot markers located on the heels and the lateral aspect of the fifth metatarsal bone were used for subsequent gait key event calibration.

[0082] The experiment consisted of three phases: Phase 1 was the OFF state, where patients discontinued medication for at least 12 hours and stopped DBS stimulation for at least 1 hour; Phase 2 involved STN-DBS in the unmedicated state, with two parameters: high-frequency stimulation (HFS, 130 Hz) and low-frequency stimulation (LFS, 60 Hz). The stimulation pulse width was fixed at 60 μs, and the stimulation intensity was set to the current value that would achieve optimal gait improvement for the patient. The two states were performed in a pseudo-random order, with a washout period of at least 30 minutes between each state; Phase 3 was the ON state with medication. Before each TUGT test, patients focused on a cross marker fixed 2 meters in front of them and remained standing for 3 minutes to obtain stable baseline electrophysiological data. All patients completed at least four timed "stand-up-walk" tests (TUGT) in each of the four states, walking back and forth in a 5-meter open corridor at a natural pace.

[0083] 4. Gait data analysis and gait periodization

[0084] All gait trial data were exported offline, and freeze events were assessed and determined by two independent experts through offline video evaluation. Based on whether a freeze event occurred, gait trials in the OFF state were divided into frozen trials (FOGT) and non-frozen trials (nFOGT), and the severity of freeze was quantified by three indicators: number of freezes, freeze duration, and freeze time percentage.

[0085] After excluding turning and freezing periods, for data during straight-line walking, combined with the Codamotion system's visualization interface and synchronously acquired 3D motion trajectories of marker points, experienced analysts manually annotated key events of the gait cycle (see [link]). Figure 3 The marking is based on: heel markers ( Figure 3 The local lowest point of the vertical displacement trajectory of the heel marker (heel) at the moment of contact corresponds to the heel strike event (LHS, RHS), and the forefoot marker ( Figure 3 The vertical displacement trajectory of the fifth metatarsal marker (toe) at the moment of takeoff corresponds to the toe takeoff event (LTO, RTO). All annotation results were independently verified by a second analyst.

[0086] A gait cycle is defined as the entire process from the first heel strike (LHS) to the second heel strike (LHS2), which involves five key events in sequence: LHS, right toe lift-off (RTO), right heel strike-off (RHS), left toe lift-off (LTO), and LHS2. Based on this, it is divided into the support phase (LHS→LTO) and the swing phase (SW, LTO→LHS2). The support phase is further subdivided into the first double support phase DS1 (LHS→RTO), the single support phase SS (RTO→RHS), and the second double support phase DS2 (RHS→LTO). DS1 and DS2 are combined into the double support phase. Three types of parameters are calculated for each gait cycle: gait space parameters (stride length, stride speed, stride velocity), gait time parameters (step frequency, stride duration, stride duration), and gait cycle component parameters (double support time, single support time, support time, swing time, each value being the proportion of the corresponding phase duration to the total gait cycle duration). Indicators related to stride length and step frequency are all calculated using the mean of two-sided data to reduce the influence of one-sided differences.

[0087] 5. Electrophysiological data acquisition

[0088] See Figure 4 During gait testing, electrophysiological data were acquired simultaneously with the patient's gait, with the two systems synchronized via 5V TTL pulses. All electrophysiological data were recorded using a 64-channel EEG acquisition system (JE-212 amplifier, Nihon Kohden, Japan; filtering range 0.08–660 Hz, sampling frequency 2000 Hz) to record local field potentials in the bilateral STN and right cortex. Considering spatial resolution, only records containing at least two adjacent contacts in each cortical region were included, and the analysis was limited to the M1, S1, and SPL regions.

[0089] 6. Electrophysiological data analysis and dual baseline normalization strategy

[0090] Offline LFP data were first preprocessed using SPM (Wellcome Centre for Human Neuroimaging, UCL, UK), which performed the following steps in sequence: (1) deleting irrelevant channels; (2) bandpass filtering (1~45Hz); (3) notch filtering to remove 50Hz power frequency interference and 60Hz / 130Hz stimulus-related artifacts and their harmonics; (4) downsampling to 250Hz; (5) re-referenced using the bipolar reference method with adjacent electrodes; and (6) manually inspecting and removing data segments with obvious artifacts or drift.

[0091] The preprocessed LFP data underwent time-spectrum analysis using Morlet wavelets to obtain power spectrum features. After time synchronization with key gait events (LHS, RTO, RHS, LTO, and LHS2), the data was divided into complete gait cycles and individual time phases. To achieve feature averaging across gait cycles, linear time deformation was used to align the time points of key events in each cycle, followed by time resampling using cubic spline interpolation. The frequency bands of interest were divided into theta (4–8 Hz), alpha (8–13 Hz), lowbeta (13–20 Hz), and highbeta (20–30 Hz).

[0092] To perform z-score normalization analysis, this invention employs a dual-baseline strategy (see [link to relevant documentation]). Figure 5 Baseline A (left figure) – Mean power spectrum of standing state: Specifically used for full gait cycle analysis, the mean power spectrum of standing state LFP is used as the baseline to reveal the overall energy change during walking relative to the stationary standing state; specifically, the standing state LFP data is divided into equal segments according to the duration of the corresponding gait cycle, and the mean of the power spectrum of each segment is used as the baseline A to ensure the comparability of the baseline calculation and gait cycle analysis on the time scale;

[0093] Baseline B (right figure) – Gait cycle mean: Specifically used for gait phase-specific analysis, using the gait cycle mean as the baseline to reveal the relative dynamic changes between the four phases DS1, SS, DS2, and SW within the same gait cycle.

[0094] This dual-baseline strategy enables the same LFP data segment to simultaneously output overall energy information for "walking vs. stationary" and local dynamic information for "phase vs. phase," which is the technical premise for the discovery of the key result in this invention: "no difference in M1-STN coherence was found in full gait period analysis, while phase-specific analysis revealed the inverse regulation of S1-STN theta coherence."

[0095] To investigate the changes in cortical-STN connectivity, wavelet coherence was further employed in MATLAB 2022b. Continuous wavelet transform (CWT) was performed on the paired right cortical electrodes and the same-side STN electrodes. After smoothing, coherence values ​​in the range of 0 to 1 were obtained. The subsequent processing procedure was the same as that for power spectrum analysis.

[0096] 7. Extraction of Frozen-Related Abnormal Gait Periodic Features

[0097] Eight patients underwent a total of 123 TUG trials, with a total duration of 6268.60 s, identifying 948 gait cycles (total duration 1115.99 s), including 44 frozen trials and 79 unfrozen trials. In subsequent electrophysiological analysis, due to artifacts from individual electrodes, data from the S1 brain region of patient 6 and the SPL brain region of patient 7 were not included in the analysis; however, data from the remaining patients were fully included. Therefore, the electrophysiological feature analysis involving the S1 brain region was based on data from seven patients, the electrophysiological feature analysis involving the SPL brain region was based on data from seven patients, and the gait parameter analysis was still based on data from all eight patients.

[0098] Studies have found that during normal walking, FOGT already exhibits worse gait performance compared to nFOGT (see [reference]). Figure 6 Stride length (T=3, P=0.036), stride speed (T=0, P=0.012), stride length (T=2, P=0.025), and stride velocity (T=0, P=0.012) were all significantly reduced. Phase-specific analysis showed that the double-support time of FOGT was significantly prolonged (T=0, P=0.012), while support time and swing time showed increasing and decreasing trends, respectively (P=0.069). These results indicate that FOGT exhibits identifiable abnormal gait parameters during the non-frozen stable walking phase, providing experimental evidence for the gait cycle segmentation module to output frozen risk-related abnormal gait characteristics.

[0099] 8. Identification of abnormal electrophysiological features related to freezing

[0100] (1) Abnormal features of multiple cortical regions

[0101] During walking, the M1 lowbeta power of nFOGT showed an increasing trend compared to FOGT (t=2.255, P=0.059). A linear mixture model (LMM) was further used to analyze the association between M1 lowbeta power and gait parameters. Group interaction effect analysis of stride length parameters showed no significant difference between FOGT and nFOGT (stride length: βinter[SE]=8.120[4.251]; stride length: βinter[SE]=16.239[8.501], P=0.057); pooled sample analysis showed a significant positive correlation between M1 lowbeta power and stride length parameters (stride length: β[SE]=14.643[2.636]; stride length: β[SE]=29.287[5.272], P<0.001), suggesting a positive association between M1 lowbeta power and stride length in both groups. In contrast, a significant group interaction effect was found in the gait parameter (gait speed: βinter[SE]=23.739[7.627]; stride speed: βinter[SE]=23.528[7.668], P=0.002), indicating that the effect of M1 lowbeta power on gait speed differs significantly between FOGT and nFOGT, and the positive correlation slope between M1 lowbeta power and gait speed in FOGT is significantly higher than that in nFOGT (see [link to relevant documentation]). Figure 7 ).

[0102] Gait phase specificity analysis (see Figure 8The study found that the lowbeta power in the S1 brain region of FOGT exhibited an abnormal phase-dependent fluctuation pattern: compared with nFOGT, the lowbeta power was significantly reduced during the support phase (T=1, P=0.028) and increased during the swing phase (T=3, P=0.063), with a significant group interaction effect only observed in the swing phase lowbeta power (βinter[SE]=-4.343[2.189], P=0.048). Further analysis showed that only in FOGT was the swing phase lowbeta power negatively correlated with swing time (βFOGT[SE]=-4.776[1.713], P=0.006), and the pooled sample analysis also showed a negative correlation between the support phase lowbeta power and support time (β[SE]=-3.154[1.501], P=0.037). In addition, compared with nFOGT, the lowbeta power of SPL in FOGT was significantly weakened during the dual support phase (T=2, P=0.043). The reduced lowbeta power of the SPL suggests a persistent abnormality in the sensorimotor integration function of the superior parietal lobule during the critical phase of center of gravity shift in FOG patients. Combined with the results of subsequent stimulation intervention experiments showing that 130Hz high-frequency STN stimulation specifically enhanced the lowbeta power of the SPL during the double-support phase (P=0.043) and simultaneously shortened the double-support time (P=0.028), this indicates that the reduced lowbeta power of the SPL during the double-support phase is a functional electrophysiological abnormality that can be targeted for intervention and has clinical significance as a trigger point for high-frequency stimulation.

[0103] (2) Cortical-STN anomalous coupling characteristics

[0104] Full gait periodicity analysis showed no significant difference in M1-STN coherence between FOGT and nFOGT, while gait phase specificity analysis (see...) Figure 9 The results show that the theta coherence of S1-STN exhibits an inverse regulation pattern between the support and oscillation phases: the theta coherence of FOGT is significantly higher than that of nFOGT in the support phase, and significantly lower than that of nFOGT in the oscillation phase (T=0, P=0.018). This inverse regulation anomaly is only revealed in phase-specific analysis using the mean of the gait cycle as the baseline, further validating the technical necessity of the dual-baseline normalization strategy.

[0105] Based on the above results, the abnormal neural features identified by the freeze-related abnormal electrophysiological feature identification module of the present invention include: phase-dependent fluctuation abnormalities of lowbeta power in the S1 brain region, which decreases in the support phase and increases in the swing phase; abnormalities of lowbeta power in the SPL brain region, which decreases in both support phases; and inverse regulation abnormalities of theta coherence of S1-STN, which increases in the support phase and decreases in the swing phase. These three features together constitute the set of electrophysiological features for freeze risk identification.

[0106] 9. Optimization scheme for gait phase-specific stimulation parameters

[0107] (1) The regulatory effect of different frequencies of DBS-STN on gait parameters and electrophysiological characteristics

[0108] STN-DBS at different frequencies effectively improved frozen gait (see...) Figure 10 Stimulation conditions had significant main effects on the number of freezes (χ²(2)=8.857, P=0.012), freeze duration (χ²(2)=11.143, P=0.004), and freeze time percentage (χ²(2)=8.857, P=0.012). Compared with FOGT, low-frequency stimulation (60Hz) significantly improved all freeze indicators (P<0.01), while high-frequency stimulation (130Hz) only significantly shortened freeze duration (P=0.047). Regarding gait parameters, both high-frequency and low-frequency stimulation significantly increased stride length (P=0.045; P=0.008) and shortened double support time (P=0.028; P=0.041); low-frequency stimulation also further improved gait speed (P=0.010), stride length (P=0.008), stride velocity (P=0.010), and significantly shortened support time and prolonged swing time (P=0.043).

[0109] Phase-specific electrophysiological analysis (see Figure 11 The study found that stimulation had a significant main effect on lowbeta power (stance phase S1: F(2,12)=5.378, P=0.021; oscillation phase S1: χ²(2)=7.143, P=0.028; dual-stance phase SPL: χ²(2)=6.00, P=0.049). Post-hoc tests showed that low-frequency stimulation (60Hz) significantly increased the lowbeta power of S1 in the stance phase and decreased its lowbeta power in the oscillation phase (P=0.021), while high-frequency stimulation (130Hz) specifically enhanced the lowbeta power of SPL in the dual-stance phase (P=0.043). Correlation analysis showed that the lowbeta power of S1 induced by low-frequency stimulation alone was significantly negatively correlated with gait period parameters (stance phase: ρ=-0.821, P=0.034; oscillation phase: ρ=-0.893, P=0.012). This result corroborates the abnormal neural characteristics of reduced lowbeta power in the dual-support phase SPL, indicating that 130Hz high-frequency stimulation exerts a gait improvement effect by targeting and restoring the sensorimotor integration function of the SPL, thus verifying the technical rationality of using the reduction of lowbeta power in the dual-support phase of SPL as the trigger point for high-frequency stimulation.

[0110] Based on the above findings, the parameter correspondence scheme established by the gait phase-specific stimulus parameter optimization module of this invention is as follows:

[0111] Gait phase Stimulation frequency Pulse width Targeted brain regions and regulatory effects Support phase 60Hz (low frequency) 60μs Increase S1 lowbeta power and shorten support time Oscillating phase 60Hz (low frequency) 60μs Reduce S1 lowbeta power and extend oscillation time Double support phase 130Hz (high frequency) 60μs Enhanced SPL lowbeta power, shortened dual-support time

[0112] Gait phase stimulation frequency, pulse width, target brain region and regulatory effect: Support phase 60Hz (low frequency) 60μs increases S1 lowbeta power and shortens support time. Swing phase 60Hz (low frequency) 60μs decreases S1 lowbeta power and prolongs swing time. Double support phase 130Hz (high frequency) 60μs enhances SPL lowbeta power and shortens double support time.

[0113] (2) Gait phase-specific adjustment effect of cortical-STN coupling characteristics

[0114] Considering the significant stimulation artifacts in the STN during stimulation, levodopa administration was used as an alternative verification method for cortical-STN coupling regulation (see [link]). Figure 12 At the behavioral level, the ON state after medication reduced all freezing indicators and improved gait spatial parameters (P<0.05) and gait cycle parameters (T=0, P=0.043). Electrophysiological analysis showed that, compared with FOGT, S1-STN in the ON state after medication exhibited weaker theta coherence in the support phase and stronger coherence in the sway phase (support phase: t=4.033, P=0.027; sway phase: t=-3.688, P=0.035), that is, the pattern of "weakened theta coherence in the support phase and enhanced theta coherence in the sway phase" is exactly the opposite of the inverse regulatory abnormality observed in FOGT. These results suggest that restoring the normal phase fluctuations of theta coherence of S1-STN within the gait cycle can serve as an electrophysiological target for gait freezing intervention and provides an objective calibration basis for the abnormal electrophysiological characteristic threshold of the levodopa state comparison module of this invention (claim 5).

[0115] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0116] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0117] This invention has the following core innovative features:

[0118] First, this invention, for the first time, uses the gait cycle as the basic analytical unit for extracting frozen gait electrophysiological features, and further analyzes the complete gait cycle into four phases: the first double-support phase (DS1), the single-support phase (SS), the second double-support phase (DS2), and the swing phase (SW). This achieves refined extraction of gait features and gait phase-specificity for neural feature extraction. Compared to existing methods that use the average value of the entire gait cycle as the analytical granularity, this invention can effectively capture dynamic neural activity abnormalities masked by averaging in key phases. Specifically, this manifests as the abnormal low-beta power phase-dependent fluctuation pattern in the S1 brain region, which is not apparent in the average analysis of the entire gait cycle but is clearly presented in the gait phase-specific analysis, and the opposite regulatory pattern of the support and swing phases of S1-STN theta coherence.

[0119] Secondly, this invention creatively proposes a dual-baseline normalization strategy that employs different baseline references for both full gait cycle analysis and gait phase-specific analysis: the mean LFP power spectrum of the standing state is used as baseline A to reveal the overall energy change during walking relative to the stationary standing state; the mean gait cycle is used as baseline B to reveal the relative dynamic changes between different phases within the same gait cycle. The two baselines operate in parallel, enabling the same LFP data segment to simultaneously output information in two dimensions without masking each other, overcoming the technical deficiency of mutual interference between overall changes and phase-specific changes under a single baseline strategy. This strategy is the key technical prerequisite for this invention to simultaneously analyze overall energy information and phase dynamic information from the same raw signal segment.

[0120] Third, this invention identifies for the first time two combined biomarkers that can stably predict freezing risk during the non-freezing stable walking phase: (1) a phase-dependent fluctuation abnormality of lowbeta power in the S1 brain region, manifested as a significant decrease in the support phase power and an increase in the swing phase power in FOGT, and this fluctuation pattern is significantly correlated with the corresponding gait cycle parameters; (2) an inverse regulation abnormality of theta coherence of S1-STN, manifested as a significant increase in the support phase coherence and a significant decrease in the swing phase coherence in FOGT, which is exactly the opposite of the normal phase fluctuation pattern in the ON state under medication. The above biomarkers are all extracted from the stable walking phase during the interictal period of freezing attacks, extending the effective time window for freezing risk identification and intervention from the instantaneous critical state of about 1.5 seconds before freezing in the prior art to the entire non-freezing stable walking phase.

[0121] Fourth, this invention establishes a gait phase-specific STN DBS parameter regulation scheme, which switches the stimulation frequency in real time according to the current gait phase: 60Hz low-frequency stimulation is output during the support and swing phases to regulate the cross-phase dynamics of lowbeta power in the S1 brain region, and 130Hz high-frequency stimulation is output during the dual support phases to specifically enhance lowbeta power in the SPL brain region. This scheme uses objective electrophysiological characteristics as the decision-making basis and the significant negative correlation between the cross-phase regulation of S1 lowbeta power induced by low-frequency stimulation and gait cycle parameters (support phase ρ=-0.821, P=0.034; swing phase ρ=-0.893, P=0.012) as quantitative support. This shifts the optimization of STN DBS parameters from trial-and-error adjustments relying on clinical experience to objective quantitative recommendations driven by neural characteristics, providing an executable parameter template for individualized closed-loop neural modulation.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A frozen gait neural feature recognition and stimulation parameter feedback control system based on gait cycle analysis, characterized in that, include: The data synchronization acquisition module is configured to acquire three-dimensional gait kinematic data and local field potential (LFP) signals of bilateral subthalamic subnuclear (STN) and target cortical brain regions, which are synchronously recorded during the subject's walking process. The gait kinematic data and the LFP signals are aligned in the time dimension. The gait cycle segmentation module is configured to take the initial heel strike of one side to the second heel strike of the same side as a complete gait cycle, and to divide each gait cycle into a support phase, a swing phase, and a double support phase within the support phase; The module for identifying abnormal electrophysiological features related to freezing is configured to use the LFP features in the standing state and the mean of the gait cycle as two types of baseline references to perform a standardized analysis of the cortical power and cortical-STN coherence in the complete gait cycle and its various phases, and to identify abnormal neural features related to freezing risk in the non-freezing stable walking phase based on the standardized analysis. The gait phase-specific stimulation parameter feedback control module is configured to receive the current gait phase output by the gait cycle division module and the abnormal neural feature recognition result output by the freeze-related abnormal electrophysiological feature recognition module, and generate and switch STN deep brain electrical stimulation parameters of different frequency bands based on the joint judgment result of the two.

2. The system according to claim 1, characterized in that, The gait cycle segmentation module is configured to take the first left heel strike (LHS) to the next left heel strike (LHS2) as a complete gait cycle. Based on five key events—LHS, right toe lift-off (RTO), right heel lift-off (RHS), left toe lift-off (LTO), and LHS2—the support phase is further subdivided into a first double support phase (DS1), a single support phase (SS), and a second double support phase (DS2). DS1 and DS2 are combined into the double support phase.

3. The system according to claim 1, characterized in that, The gait cycle division module is configured to divide the gait trials collected during the drug off-phase state into frozen trials and non-frozen trials, wherein the frozen trials are trials in which at least one freezing episode occurs, and the non-frozen trials are trials in which no freezing episode occurs. The freeze-related abnormal electrophysiological feature identification module is also configured to compare the non-freezing walking gait cycle in the freeze trial with the walking gait cycle in the non-freezing trial within the same subject, and identify the freeze risk-related abnormal neural features based on the differences in gait features and electrophysiological features between the two types of trials.

4. The system according to claim 1, characterized in that, The freeze-related abnormal electrophysiological feature identification module includes a parallel full gait cycle analysis unit and a gait phase-specific analysis unit. The full gait cycle analysis unit uses the mean LFP power spectrum of the subject in the standing state as a baseline to perform z-score normalization on the cortical power and cortical-STN coherence within the complete gait cycle. The gait phase-specific analysis unit uses the mean gait cycle as a baseline to perform z-score normalization on the cortical power and cortical-STN coherence within the support phase, swing phase, and double support phase.

5. The system according to claim 1, characterized in that, The freeze-related abnormal electrophysiological feature identification module is configured to divide the analysis frequency band into four frequency bands: theta, alpha, lowbeta, and highbeta, and use the lowbeta power and theta coherence as the target frequency band for freeze-related risk-related abnormal neural features.

6. The system according to claim 5, characterized in that, The abnormal neural features associated with the risk of freezing include at least one or more of the following features: (1) a phase-dependent fluctuation abnormality in the lowbeta power of the primary somatosensory cortex; (2) an abnormality in the decrease of the lowbeta power of the superior parietal lobule during the double support phase. And (3) abnormal inverse regulation of primary somatosensory cortex-STNtheta coherence.

7. The system according to claim 5, characterized in that, It also includes a levodopa state comparison module, configured to record theta coherence of the primary somatosensory cortex-STN in the "off" and "on" states of levodopa, respectively, and use the weakened theta coherence of the support phase and the enhanced theta coherence of the swing phase in the "on" state as the threshold reference calibration for the freeze-related abnormal electrophysiological feature recognition module to determine theta coherence inverse regulation abnormality.

8. The system according to any one of claims 1-7, characterized in that, The gait phase-specific stimulation parameter feedback control module includes: The low-frequency adjustment unit is configured to output 60 Hz low-frequency STN stimulation parameters in the support phase and the swing phase. The 60 Hz low-frequency stimulation parameters are set to increase the lowbeta power of the primary somatosensory cortex in the support phase and decrease the lowbeta power of the primary somatosensory cortex in the swing phase, so that the cross-gait phase adjustment of the lowbeta power of the primary somatosensory cortex is negatively correlated with the corresponding gait cycle component parameters. A high-frequency adjustment unit is configured to output a 130 Hz high-frequency STN stimulation parameter in the dual-support phase, the 130 Hz high-frequency stimulation parameter being set to specifically enhance the lowbeta power of the superior lobule in the dual-support phase; The switching control unit is configured to switch between the low-frequency adjustment unit and the high-frequency adjustment unit based on the current gait phase and the presence or absence of abnormal neural features identified in the corresponding phase.

9. The system according to claim 8, characterized in that, The switching control unit is configured to execute the following triggering logic: when the current gait phase is the support phase or the swing phase, and an abnormal phase-dependent fluctuation of the lowbeta power of the primary somatosensory cortex and / or an abnormal inverse regulation of the theta coherence of the primary somatosensory cortex-STN are detected, the low-frequency modulation unit is triggered to output a 60 Hz stimulation parameter; when the current gait phase is the double support phase, and an abnormality of a decrease in the lowbeta power of the superior lobule during the double support phase is detected, the high-frequency modulation unit is triggered to output a 130 Hz stimulation parameter; the switching of the stimulation parameter is performed at the transition between adjacent gait phases.

10. The system according to claim 1, characterized in that, The data synchronization acquisition module includes a gait data acquisition submodule and an electrophysiological data acquisition submodule. The gait data acquisition submodule and the electrophysiological data acquisition submodule achieve time synchronization through synchronization pulses.

Images