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

Freezing of gait (FoG) in Parkinson’s disease (PD) is an episodic failure to initiate or sustain gait, elevating fall risk. While clinical observation remains the gold standard, inertial sensors enable the objective, event-level characterization of movements. We aimed to quantify sensor signatures across baseline, pre-FoG, FoG, and post-FoG windows and to test whether a frequency-domain FoG-ratio and time-domain acceleration (RMS) differentiate these windows during Turning and Walking. We analyzed the DeFoG dataset, recorded during off-medication periods during home-like tasks. A single lumbar accelerometer was worn during standardized protocols with video-verified FoG annotations. Medium-duration events (2–5 s) were segmented into 2-s baseline, 2-s pre-FoG, the FoG itself, and 2-s post-FoG. Features per axis (antero-posterior, AP; medio-lateral, ML; and vertical, V) were: (i) RMS acceleration and (ii) FoG-ratio. FoG-ratio peaks at FoG: strong window effects in Turning (all axes) and Walking (AP, V), with FoG > baseline, pre-, and post-windows. RMS complements this: in Turning, AP and V rise post-FoG; in Walking, AP and V are already elevated pre- and during FoG, indicating mounting instability. The AP FoG-ratio during Turning was positively correlated with the New Freezing of Gait Questionnaire score. The FoG-ratio reliably discriminated the FoG window across tasks at the event level. RMS changes captured task- and phase-specific amplitude dynamics that, together with known pre-FoG degradations, may inform future phase-aware detection pipelines. This association supports the clinical relevance of AP FoG-ratio during turning and suggests that it may reflect perceived freezing severity. • FoG-ratio peaks during episodes, distinguishing all temporal windows. • RMS and spectral shifts reveal phase-specific gait instability patterns. • Turning and walking show distinct amplitude and frequency dynamics. • Pre-FoG drifts suggest early detection potential in wearable systems.