Cleveland Clinic partners with Piramidal to monitor brain health in real time

The Cleveland Clinic is partnering with San Francisco-based startup Piramidal to develop an AI foundation model that monitors patients’ brain health in intensive care units using electroencephalogram (EEG) data. The system aims to interpret continuous streams of brain wave data and flag abnormalities in seconds, potentially transforming how doctors detect neurological issues in critically ill patients.

Why this matters: Current EEG monitoring in ICUs requires manual review that can take two to four hours for a day’s worth of data, with reports generated only every 12 to 24 hours—delays that could prove critical for patients experiencing seizures or declining brain function.

How it works: The AI model processes EEG data collected via electrodes placed on patients’ scalps, analyzing the brain’s electrical activity patterns to detect problems in real time.

• Unlike text-based AI models like ChatGPT, this system is specifically trained on brain wave data to identify abnormal neural activity.
• The model incorporates nearly a million hours of EEG monitoring data from “dozens of thousands” of patients, both neurologically healthy and unhealthy.
• “Our model plays that role of constantly monitoring patients in the ICU and letting the doctors know what’s happening with the patient and how their brain health is evolving in real time,” says Piramidal’s chief product officer Kris Pahuja.

The company behind it: Piramidal was founded in 2023 by CEO Dimitris Fotis Sakellariou, a neuroengineer with 15 years of EEG research experience, and Kris Pahuja, who previously worked on product strategy at Google and Spotify.

• The Y Combinator-backed startup raised $6 million in seed funding last year.
• The company built its model using publicly available EEG datasets plus proprietary data from Cleveland Clinic and other partnerships.

What makes brain AI challenging: Brain activity patterns vary dramatically between individuals, requiring massive datasets to capture common patterns and features.

• “The beauty of a foundation model is the same way ChatGPT can generalize text, it can adapt to your tone, it can adapt to your way of writing—our model is able to adapt to the brains of different people,” Sakellariou explains.

Current limitations: The team is focused on reducing false positives and false negatives, particularly the latter scenario where the system fails to catch someone experiencing a severe neurological event.

• “This is a big problem that keeps us awake at night,” says Imad Najm, a neurologist and director of the Epilepsy Center at Cleveland Clinic’s Neurological Institute.
• While Piramidal claims “humanlike” performance against a network of doctors, the company has not disclosed specific accuracy metrics.

Timeline and rollout: The team plans to test the model with live patient data in a controlled ICU environment within six to eight months.

• They’ll start with a limited number of beds and doctors before gradually expanding to monitor hundreds of patients simultaneously.
• The company is currently using retrospective patient data to fine-tune the model.

Beyond the ICU: Piramidal envisions expanding its brain foundation model to epilepsy monitoring, sleep studies, and potentially consumer applications like EEG earbuds for measuring emotional states.

• Brain-computer interface company Synchron is also developing a similar brain foundation model using data from trial participants.

Ethical considerations: The development raises important questions about brain data usage, storage, and privacy.

• “Advancements like this one highlight the need for anticipatory ethical frameworks that support responsible development and use of these technologies,” says Caroline Montojo, president and CEO of the Dana Foundation, a private philanthropic organization dedicated to neuroscience research.
• She emphasizes the need for diverse perspectives from ethicists, social scientists, legal scholars, and patients in early technology design stages.