Comprehensive Summary
In this paper, Artificial Intelligence (AI) was used to identify different behavioral features during epileptic seizures. Using the mouse model of Angelman syndrome, 32 mice were exposed to flurothyl to induce seizures once daily for 8 consecutive days. The AI model was trained on three different data sets and then applied to the test group of 32 mice. Videos of seizures were recorded at 480p and 30fps. DeepLabCut (DLC) and Behavioral Segmentation of Open Field in DLC (B-SOiD) were used to identify 63 interpretable behavior groups. The behavioral groups can be used to identify seizure states, progression, and mortality. The system was shown top separate GS from preictal behavior. Future research should explore translation potential to other model organisms as well as epilepsy patients.
Outcomes and Implications
Analysis of behavior and motor aspects of epileptic seizures are important for delineating seizure outcomes. Research in the identification and categorization of different behaviors of epileptic seizures can provide a data-driven way of evaluating the risks of sudden unexpected death in epilepsy (SUDEP). Identification of these features can assist in identifying epileptogenic zones and seizure types, while also being useful for additional preclinical studies.