India unveils ‘MANAS 1’: AI trained on 60,000 hours of brainwaves aims to detect disorders early | India News
New Delhi: Artificial intelligence may soon help doctors “read” the brain before diseases become visible. An Indian team unveiled MANAS 1, a brain language foundation model built on 60,000 hours of brainwave recordings from more than 25,000 patients, with the aim of enabling earlier detection of neurological and psychiatric diseases.Developed by Intellihealth (NeuroDx), led by neurologist Dr Puneet Agarwal, former professor at the All India Institute of Medical Sciences, and his team, the model was launched during an AI summit and released as open source on Hugging Face. The project received computational support under the Indian AI Mission of the Ministry of Electronics and Information Technology.
Unlike conventional AI systems, MANAS 1 is trained to interpret EEG signals – the electrical activity generated by the brain. Built with 400 million parameters, it is described by its developers as a basic platform on which disease-specific AI tools can be developed.Dr Aggarwal told TOI that MANAS 1 is designed to “understand the basic language of the brain”. He describes it as a foundational model — similar in concept to ChatGPT — that learns from large-scale EEG data to interpret brain signals that traditional tests like MRI can’t fully decode. According to him, the model creates a platform on which AI tools for epilepsy, dementia and other disorders can later be developed, while also helping researchers explore aspects of brain function that are poorly understood.The rationale for public health is early access. India faces a shortage of neurologists and psychiatrists, especially outside the big cities. Brain disorders are often diagnosed late, increasing disability and long-term costs. Developers say the tools built on MANAS 1 can help doctors at Ayushman Arogya Mandirs, community health centers and district hospitals perform early screening and make timely referrals. Any disease-specific AI models derived from the platform will require regulatory approval prior to clinical deployment.When validated at scale, such systems could help reduce the gap between symptom onset and diagnosis—an important factor in conditions such as epilepsy and dementia.A next-generation version, MANAS 2, is expected in the coming weeksAs artificial intelligence advances in neuroscience, MANAS 1 signals efforts to interpret the brain’s electrical language from language analysis on screens—with implications for research, diagnosis, and access to care.
