Cleveland Clinic and Piramidal Collaborate on Large - scale AI Model for ICU Brain Health Monitoring

The Cleveland Clinic has embarked on a strategic partnership with Piramidal, a San Francisco - based startup, to develop a large - scale artificial intelligence (AI) model. This model is specifically designed to monitor the brain health of patients in intensive care units (ICUs).

System Basis: EEG - Driven Data

Unlike traditional AI models trained on text, this system is grounded in electroencephalogram (EEG) data. EEG data is collected through electrodes placed on the scalp. A computer then reads this data, presenting it as a series of wavy lines. EEG serves to record the brain's electrical activity, with any alterations in this activity potentially indicating an underlying problem. In an ICU, doctors meticulously scan EEG data to identify signs of seizures, changes in consciousness, or a decline in brain function.

Current Monitoring Challenges

Presently, doctors rely on continuous EEG monitoring to detect abnormal brain activity in ICU patients. However, real - time monitoring of every individual patient is not feasible. Typically, EEG reports are generated every 12 or 24 hours, after which they are analyzed to determine if a patient is experiencing a neurological issue. Manually reviewing a day's worth of brainwave data can take between two to four hours.

According to Imad Najm, a neurologist and the director of the Epilepsy Center at the Cleveland Clinic’s Neurological Institute, “This process is not only time - consuming but also subjective, being highly dependent on experience and expertise.”

The New AI System's Functionality

The AI system jointly developed by the Cleveland Clinic and Piramidal is engineered to interpret continuous streams of EEG data. It can flag abnormalities within seconds, enabling doctors to intervene promptly.

As Kris Pahuja, Piramidal’s chief product officer, states, “Our model assumes the role of continuously monitoring ICU patients, providing doctors with real - time insights into a patient's condition and the evolution of their brain health.”

Piramidal's Background and Model Development

Piramidal was founded in 2023 by Kris Pahuja and CEO Dimitris Fotis Sakellariou. Their objective was to construct a foundation model for the brain - an AI system capable of comprehensively reading and interpreting neural signals across different individuals. Prior to this, Sakellariou spent 15 years as a neuroengineer and AI scientist conducting EEG research, while Pahuja worked on product strategy at Google and Spotify. Backed by Y Combinator, the startup raised $6 million in seed funding last year.

The company built its ICU brain model using a combination of publicly available EEG datasets, along with proprietary EEG data from the Cleveland Clinic and other partnerships. Sakellariou reveals that the model incorporates nearly a million hours of EEG monitoring data from “dozens of thousands” of patients, both those with healthy and unhealthy neurological conditions. Given the high variability of brain activity patterns among individuals, a vast amount of data is required to capture common patterns and features for building a brain foundation model.

Sakellariou further elaborates, “Similar to how ChatGPT can generalize text, adapt to tone, and writing style, our model can adapt to the brains of different people, which is the beauty of a foundation model.”

Model Testing and Roll - out Plan

At present, the Cleveland Clinic and Piramidal team are utilizing retrospective patient data to fine - tune the model. Over the next six to eight months, they intend to test the model in a strictly controlled ICU environment, using live patient data with a limited number of beds and doctors. Subsequently, they plan to gradually roll out the software to the entire ICU. Ultimately, as Najm indicates, the software will enable the hospital system to monitor hundreds of patients simultaneously.

The slow roll - out strategy is aimed at reducing the occurrence of false positives and false negatives. False positives occur when the system misidentifies patients without a severe event, while false negatives involve the system failing to detect a patient with a genuine issue. Najm emphasizes that the latter scenario is “a significant concern that keeps us vigilant.”

Model Accuracy and Future Plans

Piramidal has not disclosed the model’s current accuracy but has stated that it has benchmarked its technology against a network of doctors, achieving “human - like” performance. The company intends to publish data on the model’s accuracy at a later date.

Broader Applications and Ethical Considerations

While Piramidal’s immediate focus is on applying its brain foundation model in the ICU, Sakellariou and Pahuja also envision its use for epilepsy and sleep monitoring. Additionally, brain - computer interface company Synchron is developing a brain foundation model, incorporating data from trial participants to enhance system accuracy and generalizability for more users. There are also potential consumer applications, such as using EEG earbuds to measure emotional states.

These medical and consumer applications raise crucial questions regarding the use, storage, and appropriate timing of brain data utilization.

Caroline Montojo, president and CEO of the Dana Foundation, a private philanthropic organization dedicated to neuroscience research, remarks, “Advancements such as these underscore the necessity for anticipatory ethical frameworks to support the responsible development and use of these technologies. It is of utmost importance to incorporate diverse perspectives from multiple disciplines, including ethicists, social scientists, and legal scholars, as well as patient experiences, at the early stages of technology design.”