DARIL: When Imitation Learning outperforms Reinforcement Learning in Surgical Action Planning
This challenges assumptions about RL superiority in sequential decision-making for surgical AI, with implications for real-time assistance in robotic surgery.
The paper compared imitation learning (IL) versus reinforcement learning (RL) for surgical action planning on the CholecT50 dataset, finding that their IL baseline (DARIL) outperformed all RL variants, with DARIL achieving 34.6% mAP for action recognition and maintaining 29.2% mAP at 10-second horizons while RL dropped to as low as 3.1% mAP.
Surgical action planning requires predicting future instrument-verb-target triplets for real-time assistance. While teleoperated robotic surgery provides natural expert demonstrations for imitation learning (IL), reinforcement learning (RL) could potentially discover superior strategies through self-exploration. We present the first comprehensive comparison of IL versus RL for surgical action planning on CholecT50. Our Dual-task Autoregressive Imitation Learning (DARIL) baseline achieves 34.6% action triplet recognition mAP and 33.6% next frame prediction mAP with smooth planning degradation to 29.2% at 10-second horizons. We evaluated three RL variants: world model-based RL, direct video RL, and inverse RL enhancement. Surprisingly, all RL approaches underperformed DARIL--world model RL dropped to 3.1% mAP at 10s while direct video RL achieved only 15.9%. Our analysis reveals that distribution matching on expert-annotated test sets systematically favors IL over potentially valid RL policies that differ from training demonstrations. This challenges assumptions about RL superiority in sequential decision making and provides crucial insights for surgical AI development.