EPFL LogSync SoC extracts multi-region synchrony biomarkers at under 200 μW for closed-loop psychiatric DBS

EPFL researchers fabricated a single-chip implantable system that records from multiple sites, extracts oscillation power and phase-synchrony features, and classifies states via on-chip machine learning. Human intracranial data from prior psychiatric tasks showed separation of cognitive engagement and emotional arousal; rat experiments confirmed phase-locked stimulation triggers. Power stayed below 200 μW through co-designed logarithmic arithmetic and hardware constraints. The approach directly targets the network-disruption model of mood and anxiety disorders rather than single-site open-loop DBS. Existing systems lack distributed biomarkers and exceed implant power budgets; LogSync meets both limits on one die. Integration into clinical practice still requires regulatory pathways for psychiatric indications, long-term biocompatibility data, and safeguards against network-level misclassification. NIMH funding of CHF 6.6 million supports extension to mood-disorder trials. Similar closed-loop architectures appear in recent Medtronic and Boston Scientific filings, yet none report network synchrony decoding at sub-200 μW. Deployment hinges on IRB-approved human implants and post-market surveillance of adaptive psychiatric stimulation.