Objective  To identify characteristic gait parameters in Parkinson’s disease (PD) patients with freezing of gait (FOG) using wearable sensors, explore factors influencing FOG based on clinical and sensor-derived parameters, and develop an identification model.

Methods A total of 83 PD patients were enrolled (FOG group, n = 20; non-FOG group, n = 63). All patients completed the Timed Up and Go (TUG) test in the “OFF” state while wearing 10 inertial measurement unit (IMU) sensors to collect gait data. Analysis of covariance (ANCOVA) was employed to control for the effects of age, Unified Parkinson's Disease Rating Scale Part Ⅲ (UPDRS-Ⅲ) scores, and thigh length, and multiple comparisons were adjusted using false discovery rate (FDR). Univariate and stratified logistic regression analyses were performed to screen influencing factors and construct an identification model. Model performance was evaluated using indices such as the area under the receiver operating characteristic curve (AUC).

Results After controlling for confounding factors, patients in the FOG group still exhibited higher stance-phase asymmetry than those in the non-FOG group. Univariate analysis indicated that decreased gait speed, increased gait variability, and increased asymmetry were influencing factors for FOG. Stratified logistic regression analysis demonstrated that adding stance-phase asymmetry to a baseline model including age and UPDRS-Ⅲ scores improved model performance, increasing sensitivity from 75.0% to 85.0% and AUC from 0.810 to 0.840.

Conclusions Stance-phase asymmetry is a core gait characteristic that distinguishes PD patients with FOG from generalized bradykinesia. Incorporating this objective sensor-derived parameter into routine clinical assessment can effectively improve FOG identification, providing evidence for objective assessment, early detection, and targeted rehabilitation intervention for FOG.