TodayToday's edition
EPFL's LogSync brain implant chip, about 4 by 2 mm, tracks activity across regions to flag cognitive and emotional states, but no patient implant study is reported yet.
2 min readAI-assisted
Three LogSync chips, each measuring approximately 4x2 mm, in the EPFL release photo.
Image: 2026 EPFL / Lundi13 (CC BY-SA 4.0)
Engineers at EPFL in Switzerland have built LogSync, a brain implant chip that identifies cognitive and emotional states from brain recordings while drawing under 200 microwatts, the university said on Oct. 23. The chip measures roughly 4 by 2 millimeters.
The team is led by Mahsa Shoaran at EPFL's Integrated Neurotechnologies Lab. The work appeared in the IEEE Journal of Solid-State Circuits in September 2026 and was developed under a CHF 6.6 million grant from the US National Institute of Mental Health.
LogSync extracts wave strength and cross-region synchrony, classifies the state and triggers stimulation; the under-200 µW power figure is researcher-reported, per EPFL.
Illustration (AI-generated)
Deep brain stimulation delivers electrical pulses to targeted brain areas. Most systems run open loop, meaning the stimulation follows preset settings rather than the patient's current brain activity. Many adaptive systems also listen to only one brain region. LogSync records activity and extracts two kinds of biomarker: the strength of brain waves, and the synchrony between signals from different regions. A machine-learning algorithm then uses those network-level biomarkers to tell cognitive and emotional states apart and to generate stimulation triggers when a state is detected.
An implant has a tiny energy budget, so every operation counts. Shoaran said the team co-designed the algorithms and hardware, and that this let network-level processing and decoding fit on one chip at under 200 microwatts, a fraction of what many implantable neural-interface systems need, in her words. The chip also uses logarithmic computations to simplify the operations behind brain-network analysis. The release gives the figure as an upper bound, not a measured total.
The researchers evaluated LogSync on human brain recordings from earlier psychiatric experiments. One dataset captured changes in cognitive engagement during a mental task and another captured emotional arousal while participants watched video clips. The algorithm separated these states from the recordings alone. In rats at the University of Minnesota, the implant tracked ongoing oscillations and produced stimulation triggers at set phases of brain activity.
No new implant study in patients is described, and no clinical timeline is given. The initial validation targeted psychiatric uses, and EPFL says the approach could extend to adaptive stimulation for movement disorders and other neurotechnology. For chip designers, the result is a worked example of how algorithm and circuit co-design can push classification into a power budget measured in microwatts.
