A Closed-Loop Brain-Machine Interface with One-Shot Learning and Online Tuning for Patient-Specific Neurological Disorder Treatment

Abstract

Treatment of neurological disorders such as epilepsy, Parkinson's tremor, and Alzheimer's disease require energy-efficient Machine-Learning (ML) on-the-edge with one-shot learning, particularly in wearable form factor for pervasiveness. In many cases, patient-to-patient variations on neurological biomarkers are huge. Thus, patient-specific training with one-shot learning and online tuning is crucial. This paper introduces a wearable closed-loop brain-machine interface system targeting one-shot learning low-power high-accuracy seizure detection classifiers, with a special focus on a low-power online-tuning scheme to effectively track each patient's symptoms.