Real-Time Kinematic Adaptive Deep Brain Stimulation Safely Reduces Gait Impairment and Freezing of Gait in Parkinson's Disease

Gait impairment (GI) and freezing of gait (FOG) affect 80% of patients with advanced Parkinsons disease. Continuous deep brain stimulation (cDBS) provides limited adaptability to address the episodic nature of FOG due to fixed parameters. Neural biomarkers for adaptive DBS are limited by signal artifacts and poor FOG classification. Wearable inertial measurement units (IMUs) offer a promising alternative by directly measuring signatures of GI&FOG. We developed Kinematic adaptive DBS (KaDBS), the first intelligent system to dynamically modulate stimulation in response to real-time gait metrics. KaDBS integrates bilateral shank-mounted IMUs with an investigational neurostimulator through a wireless architecture enabling step-detection, arrhythmicity calculation, and probabilistic FOG classification. Two control algorithms were implemented: an arrhythmicity model based on stride variability, and a P(FOG) classifier implementing tri-state control based on stepwise freezing probabilities. In the largest KaDBS cohort to date (n=8), we compared OFF, cDBS, KaDBS, and intermittent DBS during harnessed stepping and free walking. KaDBS was safe and well tolerated with no serious adverse events; symptom-free reports were 87.5% and 71.4% for arrhythmicity and P(FOG) models respectively, compared to 50.0% for cDBS. All symptoms were mild, transient, and resolved without intervention. KaDBS significantly reduced percent time freezing versus OFF during stepping-in-place (35.8%, P= 4.80 x 10-3) and free walking (33.4%, p = 9.00 x 10-). Therapeutic effects concentrated in baseline freezers: two participants with 100% time freezing during OFF achieved complete resolution with KaDBS, while non-freezers maintained stable gait. These findings establish KaDBS as a safe, effective approach to personalized neuromodulation for PD.