Summary:
Engineers at EPFL’s Integrated Neurotechnologies Lab have developed LogSync, an ultra-low-power, implantable system-on-chip designed to transform deep brain stimulation (DBS) for severe psychiatric conditions like treatment-resistant depression and PTSD. Unlike traditional open-loop devices or single-site implants, LogSync tracks oscillatory synchrony across distributed brain networks and runs on-chip machine learning to trigger stimulation only when pathological emotional or cognitive states emerge—all while operating under an ultra-efficient 200 microwatts.
Key Facts:
- Distributed Network Decoding: Rather than recording from an isolated brain locus, LogSync extracts multi-region biomarkers that quantify oscillatory power and neural phase synchrony across interacting circuits implicated in mood and anxiety disorders.
- Sub-200-Microwatt Budget: By co-designing machine-learning algorithms with custom logarithmic hardware, the chip executes real-time neural decoding and feature extraction at under 200 microwatts, meeting strict thermal and battery constraints for long-term cranial implantation.
- Human and In Vivo Validation: Tested on human intracranial datasets, the system distinguished discrete states of cognitive engagement and emotional arousal from neural oscillations alone; in vivo rat trials at the University of Minnesota confirmed precise phase-locked stimulation delivery.
Source: EPFL (École Polytechnique Fédérale de Lausanne)
Deep brain stimulation (DBS), delivering targeted electrical pulses to subcortical structures via surgically implanted electrodes, has transformed the clinical management of movement disorders like Parkinson’s disease and essential tremor. Increasingly, clinical trials are exploring its ability to alleviate severe, intractable psychiatric conditions, including major depressive disorder, obsessive-compulsive disorder (OCD), and post-traumatic stress disorder (PTSD).
However, translating DBS to psychiatric care has run into significant bioengineering hurdles. Most commercial stimulators operate in an “open-loop” configuration, broadcasting constant electrical current regardless of the patient’s immediate mental state. Even modern “closed-loop” prototypes typically measure local field potentials from just one brain region. Because complex psychiatric illnesses stem from disrupted communication across widely distributed neural circuits rather than a single damaged locus, single-channel devices often fail to capture clinically meaningful mood fluctuations.
To bridge this technological divide, an engineering team led by Mahsa Shoaran, PhD, head of the Integrated Neurotechnologies Laboratory in EPFL’s School of Engineering and Neuro-X Institute, has engineered LogSync—a specialized implantable system-on-chip that tracks network-level brain dynamics in real time.
Published in the IEEE Journal of Solid-State Circuits, the technology represents a milestone under an ongoing CHF 6.6 million initiative funded by the U.S. National Institute of Mental Health (NIMH) to engineer smart, network-responsive neural implants for severe mood and anxiety conditions.
Decoding the Language of Brain Networks
“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,” explained Dr. Shoaran.
LogSync operates by sampling neural activity across multiple electrodes, simultaneously calculating two critical network biomarkers: oscillatory power across frequency bands and phase synchrony, the precision with which distinct brain regions fire in coordinated rhythm.
These multi-site metrics feed directly into an embedded, hardware-efficient machine-learning classifier. When the algorithm detects neural signatures indicative of severe depressive withdrawal, anxiety-driven hyperarousal, or cognitive dysregulation, it instantly generates an adaptive trigger to pulse therapeutic electrical current, quieting the pathological circuit before symptoms spiral.
Engineering High Performance Under 200 Microwatts
A primary bottleneck in designing intelligent neural implants is power dissipation. Advanced machine-learning algorithms typically demand computational energy that drains miniature batteries rapidly and risks tissue heating within the brain.
The EPFL team overcame this through algorithmic-hardware co-design and logarithmic arithmetic.
“Our entire system is co-designed, meaning the algorithms and hardware are developed together with the constraints of implant in mind,” said Shoaran. “This approach helped us achieve network-level neural signal processing and decoding on a single chip, while keeping power consumption under 200 microwatts—a fraction of that typically required by many implantable neural-interface systems.”
By converting complex mathematical computations into the logarithmic domain, the chip executes demanding cross-frequency coupling and synchrony equations using simple additions and subtractions, bypassing energy-intensive hardware multipliers without compromising classification fidelity.
Validated Across Human Data and In Vivo Models
The researchers benchmarked LogSync’s decoding performance using intracranial recordings from human clinical cohorts. In one trial, the chip successfully differentiated shifting levels of cognitive engagement as subjects tackled complex tasks; in another, it classified emotional arousal states as participants viewed emotionally charged video clips, relying solely on oscillatory circuit inputs.
Subsequent in vivo validation in rodents, conducted in partnership with researchers at the University of Minnesota, established that the miniaturized chip can track continuous oscillations and execute real-time stimulation triggers locked to specific phases of the neural wave cycle.
The Road to Translation
While developed to address treatment-resistant mood disorders, LogSync’s architecture has broad implications across neuromodulation, including adaptive DBS for complex motor fluctuations in Parkinson’s, responsive neurostimulation for refractory epilepsy, and non-invasive acoustic or focused-ultrasound interfaces.
To transition these computational platforms into bedside clinical tools, Shoaran’s lab is simultaneously pioneering self-supervised foundational models, presented at ICML 2026, that learn directly from raw neural telemetry across time and channels. Commercial translation is underway through Infera Neuro, an EPFL spin-off co-founded by Shoaran to bring intelligent, network-responsive neural interfaces to patients with debilitating neurological and neuropsychiatric conditions.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- Journal paper reviewed in full.
- Additional context added by our staff.
About this neurotech and depression Research:
- Media Contact: Shira Polan
- Source: EPFL
- Image Credit: Image credited to 2026 EPFL/Lundi13
- Original Research is Open Access: IEEE Journal of Solid-State Circuits (Oct 1, 2026). “LogSync: A Low-Power Logarithmic SoC for Cross-Regional Synchrony Extraction and Adaptive Psychiatric State Classification.” Authors: Daney Alex; Amitabh Yadav; Jieun Joo; Uisub Shin; Jinhan Liu; Geoffrey W. Diehl.
- DOI: 10.1109/JSSC.2026.3728617
Abstract
LogSync: A Low-Power Logarithmic SoC for Cross-Regional Synchrony Extraction and Adaptive Psychiatric State Classification
Closed-loop deep brain stimulation (DBS) has emerged as a promising therapeutic approach for treatment-resistant psychiatric disorders, offering the potential to modulate pathological network activity in real time.
To support such interventions, we present a 32-channel brain–computer interface (BCI) system-on-chip (SoC) that integrates two key functions: 1) adaptive, biomarker-driven classification of psychiatric states and 2) phase-locked closed-loop DBS. The SoC incorporates a 32-channel low-noise analog front-end (AFE), a logarithmic arithmetic unit (LAU) for efficient and accurate feature extraction (FE), including spectral power, phase, and connectivity biomarkers, a neural additive model (NAM) classifier with backpropagation-free online adaptation, and a programmable multi-mode stimulation controller that enables flexible coordination with an external stimulator.
Fabricated in 65-nm CMOS, the chip occupies 5.46 mm2 and consumes 6.14 μ W/channel while achieving an energy efficiency of 0.197 μ J/classification. The proposed SoC was extensively validated on two intracranial electroencephalography (iEEG) datasets related to psychiatric disorders, achieving 80.9%/74.8% sensitivity and 74.2%/73.1% specificity on the multi-source interference task (MSIT) and the emotional conflict resolution (ECR) task, respectively.
We further evaluated the SoC on the public DEAP dataset for electroencephalography (EEG)-based emotion recognition, achieving 70.5%/72.5% sensitivity and 69.8%/66.3% specificity for the arousal and valence states, respectively. In addition, we performed in vitro measurements and in vivo characterization in a Long-Evans rat to validate the AFE and phase-locked stimulation controller, confirming the feasibility of the proposed SoC for real-time intervention.
Overall, the architecture enables accurate, low-power, and interpretable classification of complex neural states, supporting adaptive DBS for psychiatric neuromodulation and other BCI applications.
