Biohacking
EPFL chip reads multi-region brain activity to trigger depression stimulation in real time
By Life and Health Today Staff, . Life and Health Today.
A chip consuming less power than a small LED could one day decide, in real time, when a depressed brain needs a pulse of electricity and when it does not. Engineers at EPFL's Integrated Neurotechnologies Laboratory, led by Mahsa Shoaran, have built a device called LogSync that monitors coordinated firing patterns across multiple brain regions simultaneously and triggers stimulation only when it detects neural signatures associated with severe depression, anxiety, or cognitive dysregulation. The work was published October 1, 2026, in the IEEE Journal of Solid-State Circuits.
Why current brain stimulators fall short
Deep brain stimulation, or DBS, works by delivering targeted electrical pulses through electrodes implanted deep in the brain. It has been used for years to manage movement disorders like Parkinson's disease. Researchers have been exploring whether it can do the same for severe psychiatric conditions, including major depressive disorder and post-traumatic stress disorder, but the technology has lagged behind the ambition.
Most commercial DBS devices are "open-loop": they deliver a constant electrical current regardless of what the patient's brain is doing at any given moment. Newer "closed-loop" prototypes can adjust based on brain signals, but they typically read from only one brain region. Psychiatric conditions are thought to arise from disrupted communication across widely distributed neural circuits rather than from a single damaged spot, which means a single-channel device often misses the picture. "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," Shoaran said, as quoted by Neuroscience News.
What LogSync does differently
LogSync samples activity across 32 channels at once, calculating two network-level measurements: oscillatory power, the strength of rhythmic electrical activity across different frequency bands, and phase synchrony, the precision with which separate brain regions fire in coordinated rhythm. An on-chip machine-learning classifier processes those measurements and generates a stimulation trigger when it detects a pathological pattern.
The engineering challenge was power. Machine-learning algorithms are computationally demanding, and an implant that drains its battery quickly or heats brain tissue is not viable. The EPFL team addressed this through what they call algorithmic-hardware co-design, building the algorithm and the physical chip together from the start. By converting complex calculations into the logarithmic domain, the chip performs demanding synchrony equations using simple additions and subtractions rather than energy-intensive multiplication. The published paper reports the chip consumes 6.14 microwatts per channel and runs the entire system on under 200 microwatts total.
"Our entire system is co-designed, meaning the algorithms and hardware are developed together with the constraints of implant in mind," Shoaran said, as quoted by Neuroscience News. "This approach helped us achieve network-level neural signal processing and decoding on a single chip, while keeping power consumption under 200 microwatts."
What the testing actually showed
LogSync has not been implanted in a human patient. The validation so far has two parts. First, the chip was tested against existing intracranial recordings from human clinical cohorts: in one set, it distinguished different levels of cognitive engagement as subjects worked through complex tasks; in another, it classified emotional arousal states as participants watched emotionally charged video clips, relying on neural oscillations alone. Second, in live rat experiments conducted with researchers at the University of Minnesota, the chip tracked continuous oscillations and delivered stimulation triggers locked to specific phases of the neural wave cycle.
Those are meaningful engineering benchmarks, but they are not the same as showing the device reduces depression symptoms in a human patient. The study was designed to demonstrate that the chip can decode brain states accurately at low power, not to measure clinical outcomes. That distinction matters for anyone reading this as a near-term treatment prospect.
What comes next
The research sits within a broader initiative funded by the US National Institute of Mental Health, budgeted at CHF 6.6 million, to develop network-responsive neural implants for severe mood and anxiety conditions. Commercial translation is being pursued through Infera Neuro, an EPFL spin-off co-founded by Shoaran. The team is also developing self-supervised learning models, presented at the International Conference on Machine Learning in 2026, that could allow future chips to learn from raw neural data across time without labeled training examples.
The open question is whether a device that accurately classifies brain states in laboratory conditions will translate into meaningful symptom relief when implanted in patients with treatment-resistant depression. That requires human trials, which have not yet been reported. Anyone weighing a DBS procedure for a psychiatric condition should discuss the current state of clinical evidence with a specialist, because what is available today differs substantially from what this chip is designed to do.
Source: https://neurosciencenews.com/dbs-depression-neurotech-32338/
Life & Health Today provides information for educational purposes only, not medical advice, diagnosis, or treatment. Always consult a qualified healthcare professional.