2026 American Academy of Neurology Abstract Website

Objective:

To develop and validate a deep-learning-based behavioral quantification approach using the open-source LabGym AI platform to identify and characterize motor seizures in the Scn1b null mouse model of Dravet syndrome.

Background:

Animal behaviors and movements have traditionally been analyzed manually. Recently, an open-source AI platform, LabGym, was developed to provide an integrated platform for the holistic assessment of animal behavior. We applied LabGym to identify seizures in a well-characterized mouse model of developmental and epileptic encephalopathy. By allowing for a flexible recording background and training on pre-recorded video data, this approach increases the throughput of seizure analysis in mouse models of epilepsy.

Design/Methods:

Neonatal Scn1b null mice were previously recorded exhibiting characteristic motor seizure activity. A LabGym Detector model, which tracks individual animals, was trained to accurately detect mice during the recording. A LabGym Categorizer, which classifies behaviors, was trained to identify standing, running, sniffing, turning, and seizure behaviors.

Results:

Preliminary LabGym classifiers achieved a 94% precision in detecting seizure-like behaviors in Scn1b null mice with a 95% accuracy across all 5 behavior classifiers. The model reliably identified overt motor seizures.

Conclusions:

This work demonstrates LabGym as a novel method for quantifying complex behaviors, including motor seizures, in Scn1b null mice. Further refinement of the model, including incorporation of subtle seizure types, will enable the development of a robust, scalable tool for mouse epilepsy phenotyping.