AI Brain Model Revolutionizes ICU Monitoring with EEG Data
The Cleveland Clinic has announced a partnership with Piramidal, a San Francisco-based startup, to develop a pioneering artificial intelligence model designed to continuously monitor patients’ brain health within intensive care units. Unlike AI systems trained on vast datasets of text, this innovative model is built upon electroencephalogram (EEG) data, which records the brain’s electrical activity through electrodes placed on the scalp, presenting it as a series of wavy lines. Changes in these intricate patterns can signal critical neurological issues such as seizures, altered consciousness, or a decline in brain function.
Currently, medical professionals in ICUs rely on continuous EEG monitoring to detect abnormal brain activity. However, the sheer volume of data and the limited staff make real-time, individual patient monitoring a significant challenge. Typically, EEG reports are generated every 12 to 24 hours, requiring a neurologist to manually review a day’s worth of brainwave data, a process that can consume two to four hours. This method is not only time-intensive but also inherently subjective and heavily dependent on the individual clinician’s experience and expertise, as noted by Imad Najm, a neurologist and director of the Epilepsy Center at the Cleveland Clinic’s Neurological Institute.
The AI system under development by the Cleveland Clinic and Piramidal aims to revolutionize this process by interpreting continuous streams of EEG data and flagging abnormalities within seconds, enabling doctors to intervene far more rapidly. Kris Pahuja, Piramidal’s chief product officer, emphasizes that the model will serve as a constant guardian, monitoring patients in the ICU and providing real-time insights into their brain health evolution.
Piramidal was co-founded in 2023 by Pahuja and CEO Dimitris Fotis Sakellariou, with an ambitious goal: to create a “foundation model” for the brain—an AI system capable of broadly reading and interpreting neural signals across diverse individuals. Sakellariou brings 15 years of experience as a neuroengineer and AI scientist specializing in EEG research, while Pahuja previously honed his product strategy skills at Google and Spotify. The startup, backed by Y Combinator, successfully raised $6 million in seed funding last year. Their ICU brain model has been trained on a massive dataset, combining publicly available EEG information with proprietary data from the Cleveland Clinic and other partners. Sakellariou estimates the model incorporates nearly a million hours of EEG monitoring data from tens of thousands of patients, encompassing both neurologically healthy and unhealthy individuals. This enormous data volume is crucial because brain activity patterns vary significantly from person to person, necessitating extensive data to capture commonalities and features—much like how a large language model generalizes text to adapt to different writing styles.
The Cleveland Clinic and Piramidal team are currently fine-tuning the model using retrospective patient data. Over the next six to eight months, they plan to initiate testing in a tightly controlled ICU environment, utilizing live patient data from a limited number of beds and doctors. Following this initial phase, the software will be gradually rolled out across the entire ICU, with the ultimate goal of allowing the hospital system to monitor hundreds of patients simultaneously. This cautious, phased implementation is designed to minimize false positives, where the system incorrectly flags a patient, and, critically, false negatives—instances where a severe event is missed. The latter scenario, according to Najm, is “a big problem that keeps us awake at night.” While Piramidal has not disclosed specific accuracy figures, the company claims its technology has achieved “humanlike” performance when evaluated against a network of doctors and plans to publish detailed data on its accuracy in the future.
Beyond its immediate application in the ICU, Piramidal envisions using its brain foundation model for broader purposes, including epilepsy and sleep monitoring. The development also coincides with other advancements in brain foundation models, such as Synchron’s work on a model integrating data from trial participants to enhance its brain-computer interface system, and even consumer applications like EEG earbuds designed to measure emotional states. These emerging technologies, whether medical or consumer-oriented, raise profound questions about the collection, storage, and appropriate use of highly sensitive brain data. Caroline Montojo, president and CEO of the Dana Foundation, a philanthropic organization dedicated to neuroscience research, underscores the urgent need for “anticipatory ethical frameworks” to guide the responsible development and deployment of these technologies. She emphasizes the importance of incorporating diverse perspectives, including ethicists, social scientists, legal scholars, and the lived experiences of patients, at the earliest stages of technology design.