Neural chip decodes interactions between brain regions

EPFL researchers have developed an implantable neural part that tracks communication throughout brain regions, identifying cognitive and affectional states in genuine time.

Deep brain stimulation (DBS) is a therapeutic method that delivers electric pulses to particular areas of the brain via an implant to control neural activity. It is already extensively used for movement disorders specified as Parkinson’s disease, and researchers are increasingly exploring its possible for serious psychiatric illnesses including important depression, post-traumatic emphasis disorder, and obsessive-compulsive disorder.

However, most existing DBS systems run in what engineers call an “open loop” manner: stimulation is delivered according to predefined settings fairly than continuously adapting to the patient’s brain activity. Even among closed-loop, or adaptive systems, many depend on signals recorded from a sole brain region. This is a important limitation for psychiatric disorders, anywhere clinically applicable data may be distributed throughout networks of interacting brain regions.

“Our activity is according to the expanding build of investigation suggesting that psychiatric disorders appear additional from disruptions in communication between brain regions than from dysfunction of any sole region,” explains Mahsa Shoaran, caput of the Integrated Neurotechnologies Lab in EPFL’s School of Engineering and Neuro-X Institute.

Shoaran and her squad have developed a low-power, implantable neural system-on-chip. Dubbed LogSync, it records brain action and extracts biomarkers that depict the power of brain waves (oscillations) and synchrony between signals from distinct brain regions. A machine-learning algorithm afterward analyzes these network-level biomarkers to differentiate cognitive and affectional states, and create stimulation triggers whenever particular states are detected.

The work, published in the IEEE Journal of Solid-State Circuits, was developed as part of a CHF 6.6 myriad grant from the US National Institute for Mental health (NIMH) to create implantable systems capable of measuring and controlling oscillatory brain networks for serious psychiatric illnesses, particularly mood and anxiety disorders. It builds on the lab's former activity on brain-to-text decoding and closed-loop neuromodulation, extending these concepts to network-level biomarkers and adaptive decoding for forthcoming psychiatric therapies.

More targeted stimulation at lesser power

The researchers evaluated LogSync using individual brain recordings from former psychiatric experiments. One dataset captured changes in cognitive involvement during performing a mental task, during another captured affectional arousal as participants viewed video clips. The algorithm distinguished these cognitive and affectional states from the neural recordings alone. In additional experiments conducted in rats at the University of Minnesota, the implant tracked ongoing oscillations and generated stimulation triggers at predefined phases of brain activity.

Shoaran says that important difficulty was achieving all this functionality during operating inside the extremely constricted power prosperity of an implantable system.

“Our complete scheme is co-designed, definition the algorithms and hardware are developed together alongside the constraints of implant in mind. This method helped us accomplish network-level neural indication handling and decoding on a sole chip, during keeping power use under 200 microwatts — a fraction of that typically required by many implantable neural-interface systems,” she says. Another innovation is LogSync’s use of logarithmic computations, which simplify the complex operations needed for brain-network inspection and real-time biomarker extraction and decoding.

LogSync's first validation focused on psychiatric applications. But the capability to extract brain-network biomarkers to trigger exactly timed stimulation is broadly applicable throughout next-generation neurotechnology, including adaptive DBS for movement disorders, closed-loop non-invasive stimulation, and emerging focused-ultrasound interfaces.

In parallel, the lab is evolving new machine-learning methods that study immediately from neural recordings fairly than predefined biomarkers. Their latest ICML 2026 work, which uses self-supervised learning to grasp neural action throughout regions, channels, and time, supports broader efforts to bring intelligent neural interfaces nearer to medicinal applications, including through Infera Neuro, an EPFL spin-off co-founded by Shoaran.