Can a single ML pipeline detect seizures in different datasets
Date:
I gave a talk about my first PhD project utilising Machine Learning for epileptic seizure detection.
The objective of this project is to compare different seizure detection methods using diverse datasets. Initially, two models were selected, representing regression and decision tree analysis. For each available EEG channel, an extensive set of over 40 temporal, frequency, correlation, and graph theory derived features were computed. The models were validated using two publicly available human datasets: the Temple University Hospital Seizure Corpus and Children’s Hospital Boston. Both models were evaluated using complete feature sets as well as reduced feature sets obtained through the Boruta feature selection method.