New AI system uses Vision Transformers and LSTM networks to automatically detect surgical instrument handovers in operating room videos for improved safety and efficiency.

Researchers developed a spatiotemporal vision framework combining Vision Transformer (ViT) backbone with LSTM networks to detect and classify surgical instrument exchanges in intraoperative videos

The multi-task model jointly predicts both handover occurrence and interaction direction, reducing error propagation compared to cascaded pipeline approaches

System addresses major challenges in surgical video analysis including occlusions, background clutter, and the temporally complex nature of instrument exchange interactions

Automatic monitoring of instrument handovers enhances procedural efficiency and patient safety in the operating room environment