AI-based decision support for epilepsy monitoring

In this PhD we will develop an automated AI based decision support system, which automatically detects ongoing seizures and alarms the patient’s relatives. This way, both patient and relatives feel reassured knowing someone will help them when a seizure occurs in their everyday environment. Literature however showed that there are multiple challenges in this field, which currently lead to very low accuracy of these algorithms, including the wide variety of seizure types, the patient-dependency of epileptic changes, the lack of accurately annotated data, poor data quality and suboptimal multimodal combination. These challenges are however currently insufficiently addressed in the literature, whereas solving these challenges could lead to wide usage of these monitoring applications. This PhD aims to tackle the most important challenges: - Challenge 1: improved quantification of biomedical signals for seizure detection. This includes novel AI approaches for improved noise artifact removal and data quality assessment up to tensor-based data fusion - Challenge 2: improved multimodal classification. This includes intelligent model fusion to combine optimally different sensor output with expert knowledge and context-awareness - Challenge 3: improve and facilitate personalization of Seizure detection algorithms. Due to the limited amount of accurately annotated patient (seizure) data, active learning and data-efficient learning approaches are needed to obtain a personalized model within a reasonable amount of time. The end result is a fully automated and personalized user-friendly seizure warning and alarm system suitable for wearable sensors.