Instead of being trained on the text, the system is based on the data of the electroencephalogram (EEG), which are collected via electrodes positioned on the scalp and then read by a computer in a series of wavy lines. The EEG records brain electrical activity and changes in this activity can indicate a problem. In a setting in intensive care, doctors scan the EEG data in search of convulsions, altered consciousness or a decline in brain function.
Currently, doctors rely on continuous EEG monitoring to detect an abnormal brain activity in a patient in intensive care, but cannot monitor all individual patients in real time. Instead, EEG reports are generally generated every 12 or 24 hours and therefore analyzed to determine if a patient is experiencing a neurological problem. They may be needed by two to four hours to manually review a day of brain wave data.
“This type of things takes a long time. It is subjective and is dependent on experience and experience,” says Imad Najm, neurologist and director of the Centro di Epilessia at the Neurological Institute of the Cleveland Clinic.
The system that the Cleveland Clinic and Piramidal are developing is designed to interpret continuous flows of EEG data and flag anomalies in a few seconds so that doctors can intervene before.
“Our model plays that role of constantly monitoring patients in intensive care and letting doctors know what is happening with the patient and how their brain health is evolving in real time,” says Kris Pahuja, Chief Product Officer of Piramidal.
Pahuja and the CEO Dimitris Fotis Sakellarau founded Piramidal in 2023, with the aim of building a foundation model for the brain, an artificial intelligence system capable of reading and interpreting neural signals widely between different people. Before this, Sakellariou spent 15 years as a neuroephinguer and scientist on EEG research. Pahuja worked on the product strategy on Google and Spotify. Their startup, which is supported by Y Combinator, collected $ 6 million in Seed funding last year.
The company has created its ICU brain model using EEG data sets available to the public, as well as EEG data owners of the Cleveland Clinic and other partnerships. Sakellariou says that the model incorporates almost a million hours of EEG monitoring data of “dozens of thousands” of patients, both neurologically healthy and unhealthy. The models of brain activity are extremely variable from person to person, therefore the construction of a cerebral basic model requires enormous quantities of data to capture common models and characteristics.
“The beauty of a foundation model is the same way that chatgpt can generalize the text, it can adapt to your tone, it can adapt to your way of writing: our model is able to adapt to the brain of different people,” says Sakellarou.
Currently, the Cleveland Clinic and Piramidal team uses the retrospective data of patients to optimize the model. In the following 6-8 months, they plan to test the model in an ICU environment strictly controlled with data on live patients and a limited number of beds and doctors. From there, they aim to slowly launch the software to the entire intensive therapy. In the end, the software will allow the hospital system to monitor hundreds of patients simultaneously, says Najm.
Be First to Comment