Vision and inertial-based image mapping for capsule endoscopy

Capsule endoscopy is a non-invasive procedure for gastrointestinal diagnosis. It does not require sedation and it is comfortable and well tolerated by patient. However, the problem with such procedure is that a huge number of images is collected, which require time to investigate and diagnose; furthermore, the capsule movement is not controlled leading, in some cases, to inaccurate diagnosis. In this context, a mapping of the lumen is required to guarantee a higher reliability of the inspection, enabling the medical doctor to evaluate all the parts of the lumen for a better diagnosis. In this paper, we propose a method for mapping images from a capsule-based endoscope: the technique uses visual and inertial-based data fusion to obtain a 3D map of the lumen from 2D capsule images, also paving the way for the implementation of a path planning and autonomous locomotion and inspection.