Activity-Aware Deep Brain Stimulation Targets Parkinson's Gait Problems Separately From Tremor

Summary

Researchers at EPFL and Lausanne University Hospital, led by Eduardo Moraud, found that the subthalamic nucleus encodes which specific daily activity a person with Parkinson's disease is doing (sitting, standing, walking, avoiding obstacles) through detectable shifts in its electrical signals, and that this encoding survives even after L-DOPA and standard stimulation are applied. Using this signal, the team built a system that switches between separate machine-learning decoders depending on medication timing and stimulation level, since a single decoder trained across all conditions could not reliably tell activities apart. In a four-patient feasibility trial, stimulation that adapted in real time to detected activity reduced freezing-of-gait episodes and improved walking quality without worsening rigidity or tremor control, including during unsupervised real-world testing outside the lab.