Gastroendoscopy View Synthesis: A New Real Dataset and Evaluation

2026-06-24

AI Summary

This study addresses the lack of realistic evaluation datasets for novel view synthesis in gastroscopy, which has hindered progress in applications such as field-of-view expansion and digital twins. To bridge this gap, we introduce and publicly release GastroNVS—the first real-world dataset for novel view synthesis in gastroscopy—comprising synchronously captured endoscopic images, camera poses, and 3D point clouds. Leveraging this dataset, we conduct a systematic benchmark of multiple 3D Gaussian Splatting (3DGS) approaches, revealing their respective strengths and limitations in the complex gastric environment. Our work establishes a high-quality benchmark that fills a critical void in the field and provides a foundational resource for advancing algorithm development and clinical translation.

Abstract

Novel view synthesis (NVS) is an active research topic in computer vision, owing to the success of neural radiance field (NeRF) and 3D Gaussian splatting (3DGS) methods. While NVS opens the door to potential applications in gastroendoscopy, such as extending the field of view of endoscopic images and enabling digital twins for 3D archiving and endoscopist manipulation training, the dataset is insufficient to evaluate NVS for gastroendoscopy. In this paper, we present the first real gastroscopy dataset for NVS, namely the GastroNVS dataset, which contains a set of gastroscopic images, camera poses, and a point cloud for real gastroendoscopy inspection. To assess the suitability of the GastroNVS dataset, we evaluate several 3DGS methods and discuss the challenges for future development. The dataset is available on request from our project page.

Problem

• novel view synthesis
• gastroendoscopy
• dataset
• 3D reconstruction
• medical imaging

Innovation

• novel view synthesis
• gastroendoscopy
• 3D Gaussian splatting
• real-world dataset
• camera pose