EPFL chip could make brain stimulation for depression smarter

Deep‑brain stimulation (DBS) has helped people with Parkinson’s disease and tremor. Clinical trials are now testing it for depression, OCD and PTSD, but translating it to psychiatric care has run into engineering hurdles. Most devices send a constant pulse, no matter what the patient feels. Even newer closed‑loop prototypes usually read one brain region, missing the network problems that cause mood disorders.

Network‑level decoding replaces single‑site sensing

EPFL’s LogSync chip looks at many brain sites at once. It measures two key network markers – the power of brain waves in different frequency bands and how well separate regions fire together in phase. These signals feed into a small on‑chip machine‑learning classifier that can spot a depressive or anxious pattern and then send a trigger to the stimulator.

Keeping a tiny battery inside the skull is hard. LogSync uses algorithm‑hardware co‑design and logarithmic math to turn heavy calculations into simple adds and subtracts. The result is a 65‑nanometer chip that uses less than 200 microwatts, and only 6.14 µW per channel, a fraction of what many implants need.

The chip was tested on real human data from patients with implanted electrodes. It could tell when people were concentrating or watching emotional movies, using only brain signals. In rats, the chip timed stimulation to match a brain‑wave phase, showing it works in live animals. EPFL’s spin‑off Infera Neuro is pushing the technology toward market, hoping it could help treat Parkinson’s, epilepsy and other conditions that need smart brain stimulation.