Skip to content
Open access

A Geometry-Guided Reconstruction Framework for Monocular Gastric Endoscopy in Medical Cyber-Physical Systems

2026 · IEEE Transactions on Industrial Cyber-Physical Systems · Vol 4, pp. 1179-1191 · 0 citations · 47 references

Abstract

Capsule endoscopy serves as a perception front end in medical cyber-physical systems, where monocular reconstruction converts gastrointestinal observations into spatial representations for scene understanding and digital-twin-oriented analysis. However, weak texture, specular reflection, low illumination and limited viewpoint overlap in gastric scenes challenge correspondence estimation and pose stabilization, reducing reconstruction fidelity. To address this issue, we develop a geometry-guided reconstruction framework for monocular gastric endoscopy. First, an adaptive reprojection-based keyframe selection strategy is established to preserve informative observations for stable sparse reconstruction. Then, a shared-aggregation local feature Transformer (SALoFTR) together with iterative multi-view trajectory refinement and bundle adjustment improves correspondence robustness and reprojection-based geometric consistency under difficult gastric imaging conditions. Finally, a KAN-NeRF representation guided by refined sparse geometry generates gastric scene rendering results with improved visual fidelity. Experimental results showed that the proposed method achieves peak signal-to-noise ratios of 28.78, 30.40 and 31.52 dB on EndoSLAM, GastroHUN and a self-built capsule-robot dataset, respectively, while maintaining competitive structural similarity and perceptual quality. These results demonstrate the effectiveness under the evaluated datasets of the proposed framework for perception-oriented monocular gastric reconstruction in medical cyber-physical systems under the evaluated datasets.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.