Cleveland Clinic has partnered with AI startup Piramidal to develop a foundation model trained on EEG data that can monitor ICU patients’ brain health in real time, with the potential to transform neurological care while raising critical ethical questions. (Source: Image by RR)

Long-Term Applications Include Epilepsy, Sleep Monitoring and Consumer Devices

The Cleveland Clinic has announced a partnership with San Francisco–based startup Piramidal to develop a large-scale AI model for monitoring brain health in intensive care units. Unlike traditional models trained on text, this system is trained on EEG data—electrical activity captured by electrodes on the scalp—that can indicate seizures, changes in consciousness, or early neurological decline. Today, continuous EEG monitoring requires hours of manual review, often leaving doctors with delayed insights. The new system is designed to analyze brain activity in real time and flag abnormalities within seconds, enabling doctors to respond far more quickly in critical situations.

Founded in 2023, Piramidal is building what it calls a “foundation model for the brain,” designed to interpret neural signals across diverse patients. The startup, backed by Y Combinator with $6 million in seed funding, is led by CEO Dimitris Fotis Sakellariou, a neuroengineer and AI scientist, and Chief Product Officer Kris Pahuja, formerly of Google and Spotify. Their model, as noted in wired.com, is trained on nearly a million hours of EEG recordings from tens of thousands of patients, both healthy and neurologically impaired. This massive dataset provides the scale needed for generalization, as brainwave patterns vary dramatically between individuals. By leveraging reinforcement learning and advanced modeling, Piramidal aims to make brain monitoring more accurate and scalable.

Currently, the model is being fine-tuned on retrospective data from the Cleveland Clinic and other sources, but live ICU testing is expected within six to eight months. The trial will begin with a small number of patients and physicians before gradually expanding across the hospital system. Ultimately, the software could allow hundreds of ICU patients to be monitored simultaneously, reducing delays caused by manual reviews and improving patient safety. Physicians note that a major challenge is avoiding false negatives—cases where the system fails to detect a real neurological event—which could have devastating consequences. Accuracy data has not yet been published, but Piramidal claims the system already achieves “humanlike” performance when compared to expert neurologists.

Beyond the ICU, Piramidal envisions applications in epilepsy care, sleep monitoring, and eventually consumer-grade brain–computer interfaces. The effort parallels broader work in brain foundation models by companies like Synchron, which integrates EEG data into neuroprosthetic development. While the promise is clear, the project raises significant ethical questions about privacy, consent, and the future use of brain data. Neuroscience experts stress the importance of building ethical frameworks early, involving not only scientists and engineers but also ethicists, legal scholars, and patient advocates. As hospitals prepare for real-world testing, the project reflects both the transformative potential and the risks of applying foundation-model AI to human brain health.

read more at wired.com