AI-Driven Navigation and Reconstruction for Minimally Invasive Endoscopic Procedures

Minimally invasive endoscopic surgery is rapidly evolving with the integration of advanced imaging, computer vision, and robotics. Compared to conventional open surgical approaches, flexible and robotic endoscopy provides reduced patient trauma, faster recovery, and improved accessibility to complex anatomical regions. However, navigation within narrow, deformable, and visually challenging luminal environments remains a significant barrier. Issues such as texture scarcity, specular reflections, tissue deformation, and continuous camera motion make tasks like depth estimation, scene reconstruction, localization, and real-time tracking particularly difficult.



Recent advances in computer vision—including monocular and stereo depth learning, SLAM, NeRF-based scene representation, 3D Gaussian Splatting, and large foundation models—offer new pathways to real-time perception in these environments. At the same time, the emergence of flexible and robotic endoscopes provides new kinematic and sensing frameworks that can support sensor fusion, autonomous assistance, and surgical decision-making.



This Research Topic invites contributions at the intersection of computer vision, endoscopic imaging, and robotic surgical systems. Relevant themes include, but are not limited to:

- Monocular, stereo, and multi-view depth estimation in endoscopy

- Endoscopic SLAM, 3D reconstruction, and navigation

- Vision-based robotic control, autonomy, and shared control

- Simulation, synthetic data generation, and domain adaptation

- Vision-language models and foundation models in endoscopic perception

- Clinical evaluation, safety validation, and translation strategies



The goal is to highlight methodologies that push toward robust, real-time, and clinically deployable visual intelligence for next-generation robotic and image-guided endoscopic surgery.