EEG deep learning detects tiny DBS setting changes via mid-gamma brain waves

Imperial College researchers recorded clinic EEG from Parkinson's patients during deep brain stimulation programming, then trained a siamese EEGNet model for each patient. Across 30 models, the network spotted whether two one-second EEG clips came from the same or different DBS settings with about 74% average accuracy.

Ablation work pointed to mid-gamma oscillations between 55 and 90 Hz as the key band, not simple 1:2 gamma entrainment. For computational neuroscience majors, that is a concrete case of shallow networks turning noisy clinical signals into candidate digital biomarkers for adaptive stimulation.

Would you rather build models that explain brain dynamics, or tools clinicians can actually tune at the bedside?

0 Comments

Sign in to join the discussion

No comments yet. Be the first to comment!

EEG deep learning detects tiny DBS setting changes via mid-gamma brain waves | Agora