BCIwiki (bciwiki.com) — BIOSerenity, a French medtech company, has developed an EEG foundation model, BIOSerenity-E1, pre-trained in a self-supervised manner on more than 4,000 hours of EEG data, and demonstrated strong performance across four clinical tasks, according to a blog post published by BIOSerenity on July 2. The work will be presented as a poster at the 8th Journées de Neurophysiologie Clinique in Grenoble, France.
BIOSerenity said that a major challenge in developing AI algorithms for clinical EEG is the scarcity of annotated data. Traditional approaches require large labeled datasets for each new task, making development time-consuming and costly. The model was evaluated on four tasks: normal/abnormal classification, Alzheimer's disease detection, pediatric sleep staging, and seizure detection. Results showed strong performance across all tasks, with high generalizability confirmed on independent external cohorts across heterogeneous populations and modalities. The normal/abnormal algorithm is already integrated into a CE-marked medical device, and the seizure detection algorithm is currently undergoing clinical evaluation. BIOSerenity described these as first concrete steps toward clinical deployment of BIOSerenity-E1.
The abstract, titled "BIOSerenity-E1: a foundation EEG model for multiple clinical applications," has been accepted for poster presentation at the conference.