Cleveland Clinic A large AI model is being developed in collaboration with San Francisco-based startup Piramidal, which will be used to monitor brain health of patients in intensive care units.
Instead of being trained on text, the system is based on electroencephalography (EEG) data, collected through electrodes placed on the scalp and read through a computer in a series of wavy lines. EEG records electrical activity in the brain, and changes in that activity may indicate a problem. In an ICU environment, doctors scan EEG data for evidence of epilepsy seizures, altered consciousness, or decreased brain function.
Currently, doctors rely on continuous EEG monitoring to detect abnormal brain activity in ICU patients, but they cannot monitor each patient in real time. Instead, an EEG report is usually generated every 12 or 24 hours and then analyzed to determine if the patient is experiencing neurological problems. It can take two to four hours to manually view the value of brainwave data every day.
“This type of thing is time consuming. It’s subjective, it’s dependent on experience and expertise,” said Imad Najm, a neurologist and director of the Epilepsy Center at the Cleveland Clinic Neurology Institute.
The Cleveland Clinic and Piramidal are developing systems designed to interpret continuous flows of EEG data and marking abnormalities in seconds so doctors can intervene faster.
“Our model plays a role in continuously monitoring patients in the ICU and let doctors know what is happening to patients and how their brain health can develop in real time,” said Kris Pahuja, chief product officer at Piramidal.
Pahuja and CEO Dimitris Fotis Sakellariou founded Piramidal in 2023 with the goal of building the basic model for the brain, a system that can be widely spread across different human AI systems that can read and interpret neural signals. Prior to this, Sakellariou worked as a neuroengineer and AI scientist for 15 years in EEG research. Pahuja has developed product strategies at Google and Spotify. Their startup, backed by Y Combinator, raised $6 million in seed funding last year.
The company built its ICU brain model using publicly available EEG datasets and proprietary EEG data from the Cleveland Clinic and other partnerships. Sakellariou said the model contains nearly one million hours of EEG monitoring data, obtaining data from “hundreds of thousands of neurologically healthy and unhealthy patients.” Brain activity patterns vary greatly between people, so building basic brain models requires a lot of data to capture common patterns and functions.
“The beauty of the basic model is the same as Chatgpt can generalize text, it can adapt to your tone, it can adapt to the way you write – our model can adapt to the brains of different people,” said Sakellariou.
Currently, the Cleveland Clinic and the piramidal team are using retrospective patient data to fine-tune the model. Over the next six to eight months, they plan to test the model in an ICU environment with data on living patients as well as a limited number of beds and doctors. From there, their goal is to slowly roll out the software to the entire ICU. Ultimately, the software will allow hospital systems to monitor hundreds of patients at a time, Najm said.
