Learning Cross-Joint Attention for Generalizable Video-Based Seizure Detection

TL;DR

This paper introduces a joint-centric attention model using body joint dynamics and transformer-based techniques to improve generalization in video-based seizure detection across different subjects.

Contribution

It proposes a novel cross-joint attention framework that emphasizes body dynamics and suppresses background bias for better cross-subject seizure detection.

Findings

Outperforms state-of-the-art methods on unseen subjects

Effectively models spatial and temporal body part interactions

Enhances generalization in seizure detection

Abstract

Automated seizure detection from long-term clinical videos can substantially reduce manual review time and enable real-time monitoring. However, existing video-based methods often struggle to generalize to unseen subjects due to background bias and reliance on subject-specific appearance cues. We propose a joint-centric attention model that focuses exclusively on body dynamics to improve cross-subject generalization. For each video segment, body joints are detected and joint-centered clips are extracted, suppressing background context. These joint-centered clips are tokenized using a Video Vision Transformer (ViViT), and cross-joint attention is learned to model spatial and temporal interactions between body parts, capturing coordinated movement patterns characteristic of seizure semiology. Extensive cross-subject experiments show that the proposed method consistently outperforms state-of-the-art CNN-, graph-, and transformer-based approaches on unseen subjects.