A novel system design for monocular laser assisted capsule endoscopy for 3D small bowel modeling

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Geometry

Background:

• Small bowel 3D modeling is complex due to its tortuous and homogeneous structure.
• Existing wireless capsule endoscopy (WCE) systems provide only 2D images, limiting precise 3D reconstruction.
• Traditional 3D modeling methods struggle with monocular WCE data.

Purpose of the Study:

• To develop a novel capsule endoscopy prototype for 3D reconstruction of the small bowel from monocular images.
• To enable precise geometric analysis and modeling of the intestinal surface.
• To overcome the limitations of existing WCE systems for 3D reconstruction.

Main Methods:

• A novel capsule endoscopy prototype projects laser points onto the intestinal surface.
• Laser points are segmented using HSV color space transformation and area-based mask generation.
• Consecutive monocular images with laser points are used to generate a dense 3D point cloud for reconstruction.

Main Results:

• The proposed system achieves geometrically consistent 3D bowel model reconstruction with an RMSE of approximately ±0.3 cm.
• The 3D point cloud reconstruction achieved an RMSE of 2.27 mm in Python and 3.13 mm in CloudCompare.
• The laser-assisted monocular approach successfully reconstructed models, outperforming SIFT-SfM and ORB-SfM on phantom data.

Conclusions:

• The novel laser-assisted monocular capsule endoscopy system provides a reliable method for 3D small bowel reconstruction.
• This framework is suitable for challenging anatomical environments with limited visual features.
• The system offers a practical, low-hardware solution paving the way for advanced clinical 3D modeling.