Enhanced imaging colonoscopy facilitates dense motion-based 3D reconstruction.

Area of Science:

• Medical Imaging
• Computer Vision
• Gastroenterology

Background:

• Accurate 3D modeling of colon neoplasia is crucial for 3D measurements and classification using frameworks like the Paris classification.
• Conventional colonoscopy (white light) lacks texture, hindering dense 3D model creation with techniques like Structure-from-Motion.
• Enhanced imaging endoscopy modalities offer improved visualization for more reliable feature tracking and 3D reconstruction.

Purpose of the Study:

• To develop a novel approach for estimating dense 3D models of colon neoplasia.
• To leverage enhanced imaging endoscopy modalities to overcome limitations of conventional colonoscopy for 3D reconstruction.
• To improve the accuracy of 3D measurements and classification of superficial gastrointestinal lesions.

Main Methods:

• Utilized enhanced imaging endoscopy (Narrow Band Imaging, chromoendoscopy) for improved visualization.
• Employed Structure-from-Motion to generate a sparse 3D model of neoplasia.
• Applied Multi-View Stereo for densifying the sparse reconstruction.
• Transformed the dense 3D point cloud into a mesh using Poisson surface reconstruction.

Main Results:

• Achieved reliable feature tracks and improved 3D reconstruction results using enhanced endoscopy.
• Successfully generated dense 3D models of colon neoplasia.
• The resulting 3D models facilitate classification within the Paris classification system.

Conclusions:

• Enhanced imaging endoscopy modalities significantly improve 3D reconstruction of colon neoplasia.
• Dense 3D models derived from this approach aid in accurate diagnosis and classification of lesions.
• This technique offers a more robust method for analyzing neoplasia morphology in colonoscopy.