CergenX’s Post

First baby steps in what I hope will be a very fertile research area of "#AI4Science" for Neonatal EEG. 🧪 🔬 🥼 This work with John M. O'Toole will be presented today at #AES2024 in LA. What did we do? - We built the first neonatal seizure detection AI model that performs at human expert level - We used the model to annotate EEG recordings from babies at an unprecedented scale - With this dataset, we were able to discover a novel insight enabling us to predict progression of seizures in newborns. - We found that babies with localised seizures had significantly lower seizure burden overall that those with generalized seizures, using just the first hour after seizure onset. Why is this important? - Because Neonatal EEG is highly specialised, there's still a lot we don't know. - Uncovering this difference improves our understanding of the different trajectories of newborn brain injury - Ultimately, this understanding could lead to more targeted treatments Why do we need AI? - This insight required large scale data annotated channel-by-channel - This is extraordinarily time consuming for experts whose time is very tight - As a result, these studies simply don't get done ... until now. What's next? - This is just one small example of what's possible - We have some additional promising results currently under review for upcoming conferences - We have big plans to look more deeply here, in collaboration with our friends at INFANT Research Centre - If this kind of research interests you and you want to take it to the next level, please reach out.