Intellihealth (NeuroDX) introduced MANAS-1, an EEG-based foundation model designed to interpret brain electrophysiology, featuring training on 60,000 hours of EEG from more than 25,000 patients. The model treats brain signals as a structured biological language and was launched in India for clinical and research use.
MANAS-1 analyzed functional connectivity and neural dynamics with reported capability to identify computational biomarkers, and researchers cited 95% accuracy for epilepsy markers in initial findings. The system currently uses about 400 million parameters with plans to scale toward 2 billion, and a MANAS-2 update was announced as forthcoming.
For clinicians and health systems, MANAS-1 promises scalable, noninvasive screening that could speed early detection of neurological and psychiatric conditions and support brain‑computer interface applications. By standardizing EEG interpretation at scale, it aligns with wider trends in foundation models for medical diagnostics.
EEG Foundation Model Advances
Intellihealth MANAS-1 Enables Early Brain Screening
Trend Themes
1. EEG Foundation Models - A new class of large pretrained EEG models represents brain electrophysiology as a structured language, offering generalized interpretive layers across diverse patient cohorts.
2. Scaled Neural Biomarker Discovery - Training on tens of thousands of hours of EEG is enabling high‑throughput identification of computational biomarkers with reported high accuracy for conditions such as epilepsy.
3. Noninvasive Scalable Brain Screening - Standardized, automated EEG interpretation at population scale suggests widespread screening capability for early neurological and psychiatric condition detection.
Industry Implications
1. Clinical Neurology - Clinician workflows are poised to receive standardized, data‑driven EEG reads that can shift diagnostic pathways toward earlier and more consistent identification of disorders.
2. Medical Device Manufacturing - Demand for integrated EEG hardware and cloud‑ready firmware is likely to rise as foundation models require standardized signal capture and remote scaling.
3. Brain-computer Interface and Neurotech - Large EEG models with improved neural dynamics representation could advance decoding fidelity for BCI applications and downstream assistive technologies.