Behavior Decoding Delineates Seizure Microfeatures and Associated Sudden Death Risks in Mouse Models of Epilepsy

Area of Science:

• Neuroscience
• Artificial Intelligence
• Behavioral Science

Background:

• Epileptic seizures involve distinct motor and behavioral changes that are crucial for understanding seizure types and outcomes.
• Manual analysis of seizure behaviors is subjective, time-consuming, and fails to capture complex dynamics.
• There is a need for objective, data-driven methods to analyze behavioral repertoires in epilepsy research.

Purpose of the Study:

• To investigate the utility of artificial intelligence (AI)-aided tools for decoding complex seizure behaviors.
• To determine if AI-identified behavioral features can delineate seizure outcomes in diverse mouse models.
• To explore the potential of AI for large-scale behavioral analysis in epilepsy research.

Main Methods:

• Utilized DeepLabCut (DLC) and Behavioral Segmentation of Open Field in DLC (B-SOiD) AI tools for automated behavior analysis.
• Applied these tools to induced seizures in 32 inbred mouse strains and an Angelman syndrome mouse model.
• Analyzed 63 identified behavior groups, transition dynamics, and action kinematics.

Main Results:

• AI tools identified 63 interpretable behavior groups, revealing significant differences in behavior usage and complexity across seizure states.
• Behavioral patterns successfully delineated seizure states, progression over time, and were influenced by sex, genetic background, and mutations.
• Analysis of behavior transitions and kinematics, such as hindlimb motions, predicted the risk of sudden unexpected death in epilepsy (SUDEP).

Conclusions:

• AI-driven behavioral microfeatures offer a scalable approach for preclinical mechanistic studies and antiseizure medication screening.
• Video-based seizure behavior decoding holds significant translational potential for clinical settings, including remote patient monitoring.
• This AI methodology enables detailed analysis of seizure kinematics and dynamics, advancing epilepsy research and clinical applications.