Neural chip decodes interactions between brain regions

Deep brain stimulation uses a small implant to send electrical pulses into specific parts of the brain. Doctors already use it for movement problems such as Parkinson’s disease, and researchers are testing it for serious mental health conditions including major depression, PTSD, and obsessive-compulsive disorder.

Most current systems deliver stimulation on a fixed schedule instead of reacting to the brain’s own signals. Even systems that adapt often look at activity in only one brain area. That approach has limits for psychiatric conditions, which appear linked more to problems in how different brain regions talk to each other than to any single area going wrong.

“Our work is based on the growing body of research suggesting that psychiatric disorders arise more from disruptions in communication between brain regions than from dysfunction of any single region,”

says Mahsa Shoaran, who leads the Integrated Neurotechnologies Lab at EPFL.

Her team built a small, low-power chip called LogSync. It records brain signals, measures the strength of brain waves and how well signals from different areas stay in sync, then uses a machine-learning program to recognize certain thinking or emotional states and send stimulation only when those states appear.

The work, published in the IEEE Journal of Solid-State Circuits, was supported by a CHF 6.6 million grant from the US National Institute of Mental Health.

In tests on earlier human brain recordings, the chip could tell apart focused mental effort from emotional responses to videos. In rat experiments, it tracked ongoing brain waves and triggered stimulation at chosen moments. By designing the software and hardware together and using simplified logarithmic calculations, the team kept power use under 200 microwatts—much lower than typical brain implants.

While first aimed at psychiatric uses, the same ability to watch network activity and time stimulation could help with movement disorders, non-invasive brain treatments, and newer ultrasound methods. The lab is also developing learning methods that work directly from raw brain signals, including through an EPFL spin-off company Infera Neuro