Single-Image-Based 3D Reconstruction of Endoscopic Images

Area of Science:

• Medical Imaging
• Computer Vision
• Gastroenterology

Background:

• Wireless capsule endoscopy (WCE) aids gastrointestinal examination but faces challenges in 3D reconstruction due to device limitations.
• Accurate 3D models from WCE images can enhance pathology detection and visualization for gastroenterologists.

Purpose of the Study:

• To develop and evaluate a single-image-based 3D reconstruction method for WCE.
• To assess the accuracy and robustness of the 3D reconstruction technique on artificial colon models and WCE images.

Main Methods:

• Employed a shape from shading (SFS) algorithm for single-image 3D reconstruction.
• Utilized an endoscope simulating WCE to capture images of an artificial colon and primitive objects.
• Performed comprehensive camera calibration and quantitative error analysis against ground truth data.

Main Results:

• The SFS method accurately reconstructed the 3D geometry of primitive objects and an artificial colon.
• The approach demonstrated robustness in reconstructing colon geometry from noisy images with an unknown imaging pipeline.
• Preliminary results show the method's applicability for 3D reconstruction from WCE images.

Conclusions:

• Single-image 3D reconstruction using SFS is a viable technique for WCE applications.
• This method offers a promising solution for improving diagnostic capabilities in gastrointestinal examinations.
• Further research can optimize the technique for clinical WCE data.