Novel experimental and software methods for image reconstruction and localization in capsule endoscopy

Area of Science:

• Medical Imaging
• Gastroenterology
• Robotics

Background:

• Capsule endoscopy (CE) is a vital tool for minimally invasive gastrointestinal examination.
• Current CE technology faces limitations, particularly in precise lesion localization.
• Accurate localization is crucial for effective diagnosis and treatment planning.

Purpose of the Study:

• To develop and evaluate a 3D reconstruction and localization system for CE.
• To utilize visual data from 2D CE images for creating a 3D model.
• To improve the accuracy of lesion localization during capsule endoscopy procedures.

Main Methods:

• A robotic arm precisely navigated a capsule endoscope through a bowel model.
• A motion estimation algorithm processed 2D CE images to reconstruct a 3D model.
• The ORB-SLAM technique was employed for simultaneous 3D reconstruction and CE localization.

Main Results:

• The algorithm successfully reconstructed the 3D geometry of the bowel model, including fine details.
• The estimated travel distance of the capsule showed a mean absolute error of 4.1 ± 3.9 cm.
• ORB-SLAM accurately mapped the capsule's trajectory within the reconstructed 3D model.

Conclusions:

• The developed reconstruction methods provide high-quality 3D models of the bowel.
• Accurate capsule localization was achieved solely using information from CE video and images.
• This approach enhances the diagnostic capabilities of capsule endoscopy by improving localization.