Anatomical Landmark-Guided Deep Reinforcement Learning for Autonomous Gastric Navigation

TL;DR

This paper introduces a transferable deep reinforcement learning framework guided by anatomical landmarks for autonomous gastric navigation in wireless capsule endoscopy, achieving high coverage and efficiency.

Contribution

The proposed AL-DRL method uses landmark-guided policies and a sim-to-real pipeline to improve gastric navigation across patient models.

Findings

Achieves over 97% mucosal coverage in simulations within 50 seconds.

Demonstrates 87% coverage and 53% reduced procedure time in ex-vivo experiments.

Outperforms vanilla RL agents like PPO, SAC, and DQN in coverage and speed.

Abstract

Wireless capsule endoscopy (WCE) enables painless visualization of the gastrointestinal tract, but its diagnostic potential is limited by incomplete mucosal coverage and poor transferability of existing navigation methods across patient anatomies. We propose a transferable, anatomical landmarkguided deep reinforcement learning (AL-DRL) framework for autonomous gastric navigation. Leveraging a lightweight edgecontour-depth fusion module, our policy operates on stable, lowdimensional landmark coordinates rather than high-dimensional video streams, effectively bridging the sim-to-real gap. In simulations across eight patient-derived models, the method achieves over 97% coverage within 50 seconds, significantly outperforming vanilla PPO, SAC, and DQN agents. A two-stage sim-to-real pipeline with an adaptive dynamic programming controller actively mitigates physical disturbances. Ex-vivo experiments demonstrate a mean coverage of 87% and a 53% reduction in procedure time compared with expert manual control.