Biomechanical characteristics before, during, and after freezing of gait episodes in individuals with Parkinson's disease.

Linking executive dysfunction to gait initiation deficits in Parkinson's disease with freezing of gait.

Introduction Freezing of gait (FOG) is an unpredictable and debilitating symptom of Parkinson’s disease (PD). Current approaches to treating FOG are lacking; adaptive deep brain stimulation (aDBS) holds promise in treating FOG. However, a necessary precursor to using aDBS is understanding changes in the subthalamic nucleus (STN) prior to and during FOG. The aim of this project was to develop a machine learning model to detect FOG episodes using local field potential (LFP) data from the STN in individuals with PD. Methods The LFP data were collected from five individuals with PD using the Medtronic Percept DBS platform while in the off-therapy state (off-DBS and off-medication) as they walked through virtual reality environments designed to induce FOG by manipulating levels of physical, anxiety, and cognitive load. The LFP time-series data were z-score normalized within each participant and canonical frequency band powers, rolling means, maxima, minima, and standard deviations were calculated. In addition, alpha-beta burst dynamics were calculated and included for analysis. The model was trained and subsequently tuned with a transfer learning component before testing its capacity to detect FOG episodes. Results The deep learning model achieved an average weighted F1 score of 0.67, a macro F1 score of 0.62 across eight trials from five participants and successfully detected 24/29 (80%) of FOG episodes. The predictive importance of the twelve LFP channels varied across participants and FOG triggers, underscoring the need for an individualized, adaptable approach to modelling FOG using only LFP data. Discussion Initial results indicate that FOG detection using LFP data gathered while walking is feasible. Future aDBS applications should contemplate a patient- and environmental-specific approach to optimize FOG treatment potential. Trial registration clinicaltrials.gov , Identifier (NCT05103384).

Freezing of gait is a disabling and treatment-resistant manifestation of Parkinson's disease (PD). The effectiveness of deep brain stimulation (DBS) for freezing of gait remains inconsistent across stimulation targets, frequencies, and medication states. We conducted a systematic review and meta-analysis following PRISMA guidelines to examine how DBS affects freezing of gait in patients with PD. We searched Medline, Scopus, Web of Science, and Cochrane up to September 28, 2025. For synthesis, we combined mean differences and 95% confidence intervals for the Freezing of Gait Questionnaire (FOG-Q) and the Unified Parkinson's Disease Rating Scale (UPDRS) part III across different medication and stimulation settings to calculate the final effect size. Thirty-one studies with 905 patients were included. Of these, 21 provided FOG-Q data, and all reported UPDRS-III results. DBS led to a modest decrease in FOG-Q scores (mean difference [MD] = - 2.99; 95% CI = - 5.69 to - 0.29). The biggest improvement in FOG was seen when stimulation was used while patients were off medication (Med-OFF/Stim-OFF vs. Med-OFF/Stim-ON: MD - 5.88; 95% CI - 9.28 to - 2.47). Stimulation during the medication-ON state had smaller effects (Med-ON/Stim-OFF vs. Med-ON/Stim-ON: MD - 2.65; 95% CI - 4.99 to - 0.32), and there was no significant benefit when comparing Med-ON/Stim-OFF to Med-OFF/Stim-ON (MD - 0.70; 95% CI - 3.88 to 2.48). UPDRS-III scores improved substantially in the medication-OFF state with stimulation (MD - 14.35; 95% CI - 17.39 to - 11.32). High-frequency stimulation targeting the subthalamic nucleus provided more consistent benefits, yet substantial variation persisted across studies. The results of our small cohort showed significant improvement in FOG-Q and UPDRS-III (P values = 0.034, 0.022, respectively). DBS improves freezing of gait primarily in the medication-OFF state, with greater effects observed using high-frequency stimulation and subthalamic nucleus targets. Significant heterogeneity and limited data for alternative targets warrant cautious interpretation and further controlled studies.

Dual-task walking poses substantial challenges for individuals with Parkinson's disease (PD), particularly those with freezing of gait (FOG), as dual-task interference increases fall risk and limits functional mobility. To investigate dual-task interference at both the behavioral and neural levels in PD individuals with FOG (PDFOG+) and those without FOG (PDFOG-) during dual-task walking involving a demanding manual task. Seventeen PDFOG- and 17 PDFOG+ participated in this study. Gait performance, manual-task performance, and scalp electroencephalograph (EEG) were recorded in single-task condition and dual-task walking while performing a high attentional-load manual task (an interlocking ring task). The primary behavioral outcomes were dual-task costs (DTC) of gait velocity and ring performance, representing locomotor interference and manual-task interference, respectively. The primary neural outcomes focused on theta-band EEG measures, including regional theta power and inter-regional theta connectivity, given their established role in attentional control during dual-task walking in PDFOG+. Compared with PDFOG-, PDFOG+ exhibited significantly less DTC of gait velocity and greater DTC of ring performance. At the neural level, PDFOG+ demonstrated attenuated task-dependent modulations of increased theta-band power and decreased theta-band connectivity relative to PDFOG-. PDFOG+ exhibited an atypical pattern of dual-task interference characterized by preserved gait velocity at the expense of concurrent manual-task performance, together with diminished theta-band attentional modulation. These findings suggest limited flexibility in attentional resource allocation in PDFOG+ and highlight theta-band neural dynamics as a key mechanism underlying maladaptive task-task regulation during walking. National Taiwan University Hospital Research Ethics Committee: No. NCT03298503.

Turning is a complex motor behavior that frequently triggers freezing of gait and falls in Parkinson's disease (PD), yet its neural dynamics in naturalistic settings remain unknown. Using chronic at-home intracranial recordings in four subjects with PD, we show that turning is marked by premotor cortical beta desynchronization driven by reduced burst rate. These findings identify a robust signature of ecological turning and implicate beta dynamics in adaptive motor transitions.

Increasing research efforts in recent years and new approaches in clinical studies provided significant insights into the pathophysiology of freezing of gait (FOG). However, the study of causative mechanisms for complex gait disorders typically needs refined invivo models, which are still lacking for FOG. The characteristics of FOG pose major difficulties not only for reproduction in animals with different types of gaits but also for assessment of such episodic behavior in the animal environment. In this review, we examine the currently available animal models of FOG and FOG-like phenomena for their validity and applicability and present a prospective view of modeling based on novel technologies to manipulate integrated mechanisms.

People with Parkinson's disease and freezing of gait (PD+FOG) exhibit altered gait and turning compared to nonfreezing peers with PD (PD-FOG). However, less work has examined the effects of FOG status on daily-life mobility. Therefore, we compared daily-life gait and turning across PD+FOG and PD-FOG and related gait deficits to self-reported FOG severity and health-related quality of life (QOL). We collected daily-life mobility data over 7 days from 119 people with PD (PD+FOG = 47). Inertial sensors on the feet and lower back assessed 33 gait and balance outcomes. One-way ANCOVAs assessed the effect of FOG status on mobility. Five variables reflecting gait and turning quality were significantly worse in PD+FOG compared to PD-FOG after Holm correction: smaller turn angles, smaller foot pitch at initial contact, and increased variability of double-support time, stride length, and pitch at toe-off. Of these outcomes, two (turn angle & stride length variability) were significantly correlated with freezing severity, and 4 (turn angle, stride length variability, pitch at initial contact and pitch at toe-off variability) related to QOL. Turns per hour also related to QOL. In sum, PD+FOG exhibited worse daily-life gait quality (small turn angle and foot angle at heel strike and increased gait variability) and these outcomes were related to QOL and FOG severity. Daily-life gait quantity (number of turns, strides, and gait bouts per hour) was less affected by FOG. These results provide insight into daily-life mobility of PD+FOG, and support interventions aimed at quality of gait and turning for PD+FOG.

The cognitive scaling decoupling hypothesis: Is Parkinsonian gait disruption a failure of internal physical model calibration rather than pure motor execution?

Transparent clinical decision-making remains a critical barrier to deploying deep learning in medical diagnosis. Post hoc explanation methods approximate model behaviour after training but cannot guarantee that explanations faithfully reflect the underlying reasoning. This study proposes a Self-Explaining Neural Network (SENN) for Parkinson’s Disease (PD) screening via Ground Reaction Force (GRF) gait analysis, enforcing intrinsic interpretability through learnable basis concepts and input-dependent relevance scores computed jointly with the prediction. The architecture combines a four-block residual CNN backbone with stochastic depth regularisation, a 16-concept encoder with diversity and stability constraints, and temperature-scaled probability calibration for reliable clinical operating points. Evaluated on the PhysioNet Gait in Parkinson’s Disease dataset (306 subjects, 16 GRF sensors per foot), SENN achieves a subject-level ROC-AUC of 0.916 [95% CI: 0.867–0.964], sensitivity of 0.913 [0.862–0.963], specificity of 0.671 [0.485–0.858], and Average Precision of 0.942 [0.918–0.967], reported across five independent random seeds. Comparative evaluation against four deep learning baselines—CNN-Residual, BiLSTM, CNN-LSTM, and CNN-Attention—confirms that the interpretability constraints impose no statistically significant reduction in discriminative performance, with all pairwise ROC-AUC confidence intervals overlapping. Concept-level analysis reveals that the three most discriminative concepts correspond to disrupted midfoot loading patterns, increased step-length variability, and bilateral cadence asymmetry—all established biomechanical hallmarks of parkinsonian gait—providing clinically grounded, patient-specific explanations without post hoc approximation. These findings demonstrate that rigorous intrinsic interpretability and competitive predictive accuracy are simultaneously achievable in deep gait analysis, supporting the clinical adoption of transparent diagnostic AI.

Is levodopa induced freezing of gait a paradox or an expected phenomenon?: a clinico-pathophysiological hypothesis.

Treatment adjustments in Parkinson's Disease (PD) are often based on clinical evaluations at single time points which are insufficient to adequately assess real-life motor fluctuations. Patient-written symptom diaries on the other hand are highly subjective and require well-educated and adherent patients to provide reliable results. Wearable accelerometry might provide a reliable, objective, and continuous diagnostic method to assess PD motor symptoms & fluctuations. However, large datasets of simultaneous sensor data and symptom diaries are needed for such method development and validation. We here provide a well-described, open-science dataset of simultaneous, bilateral, wrist-worn accelerometry and symptom diary data from 66 participants (41 male, 25 female) with PD. On average, participants provided data for 6.0 consecutive days resulting in a total of 393.8 days for the dataset as a whole. Symptom diaries include data on kinetic state, tremor, freezing of gait, falls, and PD-related medication intake. Further demographic information is also provided. This dataset will support the development and validation of accelerometry-based approaches to assessing motor symptoms and fluctuations in PD.

Systematic review of kinematic and kinetic parameters in Parkinson's disease, with and without freezing of gait.

Parkinson’s disease (PD) is a neurological disorder that impacts the central nervous system, leading to a progressive decline in motor functions, including symptoms like tremors and episodes of freezing. An important area of research has been the challenges posed by these freezing episodes, during which individuals suddenly halt and struggle to resume coordinated movements. Recent studies have utilized electroencephalography (EEG) to analyze PD; however, this method faces certain limitations, prompting the investigation of alternative and interdisciplinary approaches to obtain a more comprehensive understanding of EEG signals. A promising methodology involves the use of Quantile Graphs (QGs), which have demonstrated potential in differentiating patients with various medical conditions. However, they have yet to be applied to those with PD. This article aims to explore the application of QGs for quantifying differences in the brains of individuals affected by PD, thereby paving the way for the integration of this method into PD research. To achieve this goal, EEG data were collected from 18 channels across six PD patients, capturing three distinct events during the Timed Up and Go (TUG) test: normal walking, freezing of gait (FoG), and voluntary stop. The combination of the QG method with machine learning techniques was effective in distinguishing between FoG and non-FoG EEG events. The findings validated the utility of QG in analyzing complex and nonlinear signals, such as those produced by EEG recordings in individuals with PD.

Persons with Parkinson walk slowly, with short steps, reduced arm swing, and altered stride variability. Walking to a metronome or music may increase velocity, stride length, and cadence, yet few studies have directly compared the efficacy of these techniques. To determine the optimal auditory cue to improve Parkinson gait. This case-control study was conducted among persons with Parkinson disease (using medication) and healthy older adult controls at University of Florida Applied Neuromechanics laboratory in 2017. Changes in gait in participants with Parkinson disease and controls were compared with walking with no cue: walking to a regular metronome, walking to a fractal metronome, and walking to music. Condition order was randomized. Cueing frequency was set to natural cadence. Post hoc analyses restricted to participants with Parkinson disease were performed in 2026. Auditory cueing with a regular metronome, fractal metronome, or music while walking. Outcomes of interest were stride time detrended fluctuation analysis (DFA), velocity, stride length, and arm swing velocity. Spatiotemporal gait measures and DFA of stride time were compared using repeated measures multivariate analysis of variance. Analyses included 15 participants with Parkinson disease (mean [SD] age, 69 [6] years; 11 [73%] male; mean [SD] Hoehn and Yahr Parkinson disease stage. 2.3 [0.6]; mean [SD] age of onset, 63 [7] years) and 15 controls (mean [SD] age, 69 [5] years; 11 [73%] male). Stride time DFA was increased during the fractal metronome condition (α = 0.200; SE, 0.024; P < .001) vs no cue, the regular metronome (α = 0.320; SE, 0.032; P < .001), and music (α = 0.219; SE, 0.030; P < .001); worsened with the regular metronome vs no cue (α = -0.120; SE, 0.037; P = .003); and was not statistically different during music vs no cue (α = -0.019; SE, 0.031; P = .54). Music was associated with increased velocity (mean [SE] change, 0.041 [0.015] m/s; P = .01), stride length (mean [SE] change, 0.047 [0.013] m; P = .001), and arm swing velocity (mean [SE] change, 27.10 [7.33] °/s; P = .001) compared with no cue; velocity (mean [SE] change, 0.030 [0.011] m/s; P = .03), stride length (mean [SE] change, 0.034 [0.009] m; P = .002), and arm swing velocity (mean [SE] change, 34.7 [8.5] °g/s; P = .001) compared with the regular metronome; and increased stride length (mean [SE] change, 0.028 [0.009] m; P = .01) and arm swing velocity (mean [SE] change, 37.52 [7.89] °/s; P < .001) compared with the fractal metronome. Analyses restricted to participants with Parkinson disease found similar trends, with reduced levels of significance due to the smaller sample. In this case-control study, walking to the fractal metronome was associated with improved stride time fluctuations compared with the regular metronome or music. Walking to music was associated with improved velocity, stride length, and arm swing velocity compared with either metronome condition.

Background/Objectives: Freezing of gait (FoG) is a disabling motor manifestation of Parkinson's disease (PD) associated with impaired neural control of locomotion and increased gait variability. Quantitative characterization of gait kinematics may provide biomechanical insight into FoG-related instability, particularly under different dopaminergic states. Methods: This pilot study evaluated sagittal-plane knee kinematics in healthy individuals (n = 27) and patients with PD. (n = 8) under OFF and ON dopaminergic medication conditions using two-dimensional videogrammetry (Kinovea®). Knee flexion-extension trajectories were time-normalized to 0-100% of the gait cycle, and group ensemble profiles (mean ± SD) were computed. Results: Phase-specific range of motion (ROM), within-subject variability, and interlimb coordination were quantified. Interlimb coordination was assessed using Pearson's correlation coefficients (r) and cross-correlation lag analysis computed per subject and summarized statistically across groups. Compared with healthy participants, PD patients in the OFF state exhibited significantly reduced knee ROM during stance and swing (p < 0.05), accompanied by increased kinematic variability and disrupted temporal coordination. Interlimb correlation was significantly lower in PD OFF compared to healthy gait groups (p = 0.010), with larger temporal lags, indicating impaired bilateral synchronization. Following medication intake (ON state), knee excursion increased and interlimb coordination partially improved; however, correlation values and timing symmetry did not fully normalize to healthy levels. Conclusions: These findings demonstrate that sagittal-plane knee kinematics and interlimb coordination metrics derived from low-cost 2D videogrammetry are sensitive to the dopaminergic state and reveal persistent neuromotor deficits in PD. The proposed framework provides an interpretable and accessible approach for characterizing gait organization in Parkinson's disease and supports future integration with clinical assessment and longitudinal monitoring.

Presynaptic inhibition (PSI) at the spinal cord level is crucial for coordinating postural preparation with step initiation. People with freezing of gait and Parkinson's disease (PD + FOG) show loss of PSI of the soleus muscle during step initiation that is associated with abnormal anticipatory postural adjustments (APA). Here, we hypothesize that increasing PSI of the soleus muscle during step initiation via wrist vibration in PD + FOG would decrease abnormally large APA. Fifteen PD + FOG performed self-initiated steps on a force platform without electrical stimulation and with test or conditioned Hoffman reflexes (H-reflex) to measure PSI of the soleus muscle under three conditions: OFF medication, OFF medication with vibration, and ON medication without vibration. Soleus H-reflexes were recorded during quiet stance (a control task) and when the amplitude of the APA under the same leg exceeded 10%-20% of the mean baseline mediolateral displacement. Vibration consisted of 200-300 Hz applied to the wrist when the ipsilateral leg during APA (same leg where H-reflexes were evoked) was on the ground. PD + FOG showed decreased PSI during APA in OFF and ON medication, but PSI was increased during vibration (p < 0.05). Increased PSI was associated with smaller APA during vibration (p < 0.05). Smaller APAs were associated with lower subjective freezing of gait severity (p < 0.05). These preliminary results show that wrist vibration decreases abnormal APA during step initiation by increasing ipsilateral PSI levels of the soleus muscle. Because PSI is modulated by cortical and brainstem areas related to FOG and APA, proprioceptive drive during vibration may reorganize these brain circuits.

Freezing of gait (FOG), a disabling symptom in Parkinson's disease, presents a major challenge for wearable classification algorithms that struggle to distinguish freezes from voluntary stops. To address this ambiguity, we evaluated whether incorporating eye-gaze kinematics could improve classification accuracy compared to using ankle-mounted inertial measurement units (IMUs) alone. We analyzed data from 10 participants performing standardized walking tasks and compared two deep learning classifiers differing only in their inputs: an IMU-only model (bilateral ankle accelerometer and gyroscope) and an IMU + Gaze model that improved macro-averaged F1 from 0.657 (95% bootstrap CI: 0.461-0.756) to 0.757 (0.591-0.832; Δ = 0.099, bootstrap p = 0.016). Class level improvements were largest for standing (F1: 0.600 vs. 0.356; Δ = 0.244, p = 0.019, Holm-corrected p = 0.056), driven by recall increasing from 36.4% to 81.8%, and standing windows misclassified as freezing reduced from 59.1% (13/22) to 13.6% (3/22). These findings show that gaze kinematics complement ankle kinematics for disambiguating voluntary stopping from FOG and potentially strengthen automated monitoring, clinician-facing assessment, and patient-facing assistive technologies.

Freezing of gait (FoG) is a common symptom of Parkinson's disease, especially in its later stages of progression. Characterized by involuntary stopping during normal gait patterns, FoG greatly increases fall risk, reducing quality of life. Given the complex presentation and etiology of FoG, current treatments have proven ineffective in managing episodes. In recent years, machine learning algorithms have been leveraged to derive actionable clinical insights from biomedical datasets. As a manifestation of neuromechanical dysfunction, impending FoG episodes may be characterized through data collected by wearable devices and sensors. This scoping review evaluates the current landscape of machine and deep learning-derived biomarkers to enhance the personalized management of FoG. This scoping review was conducted using established methodological frameworks for scoping reviews and is reported in accordance using the PRISMA-ScR checklist. Three databases were queried, with screening yielding 60 studies. Thirty-nine papers reported on deep learning techniques, with the most common architectures being convolutional neural networks and long short-term memory models. Inertial measurement units, which can be worn on various locations, may be a promising modality for practical implementation. To generate closed-loop FoG therapies, algorithms can be integrated into real-time systems like robotic exoskeletons or adaptive deep brain stimulation. Future work in generating datasets from ambulatory devices, as well as distributed computing strategies, may lead to real-time FoG management.

Motor impairments in Parkinson's disease (PD), including gait dysfunction, postural instability, and freezing of gait, are often inadequately managed by pharmacological therapy alone, particularly in advanced stages. Vibration therapy (VT) has been proposed as a non-pharmacological adjunct to improve motor performance through sensorimotor modulation. However, existing evidence remains inconsistent, and previous systematic reviews have reported substantial heterogeneity, limiting clinical interpretability. This protocol outlines a systematic review and meta-regression analysis designed to evaluate the effects of VT on motor function in individuals with PD, with a specific focus on domain-specific outcomes and parameter-dependent response patterns. The review will include randomized and non-randomized controlled trials investigating mechanical VT in adults with PD. Motor outcomes will be categorized into four predefined domains: (1) global motor function, (2) gait quality and freezing, (3) functional mobility, and (4) dynamic balance. A comprehensive literature search will be conducted across major electronic databases. Quantitative synthesis will be performed using random-effects meta-analysis. Subgroup analyses and meta-regression will be applied to examine whether vibration characteristics, such as frequency, amplitude, modality, and total dose, explain variability in effect sizes across studies. Risk of bias will be assessed using RoB 2 and ROBINS-I, and the certainty of evidence will be evaluated using the GRADE approach. By integrating multidimensional motor outcome categorization with parameter-sensitive analyses, this review aims to clarify whether VT effects in PD are domain-specific and dependent on stimulation characteristics. The findings are expected to provide a more mechanistically interpretable synthesis of the existing evidence and to inform the design of future clinical trials evaluating VT in PD rehabilitation. PROSPERO CRD420251124906.