Video-Endoscopic Image Registration for Cartography Reconstruction of Hollow Organs

Abstract

In examinations of hollow organs, video-endoscopic systems are used. These systems allow to analyze layers recovering the internal organ cavities (internal surfaces). In this kind of analysis, an expert must review the acquired image sequence during an endoscopic analysis, either to confirm the diagnosis issue of endoscope, or to evaluate the pathology’s evolution. For these reasons, it is difficult to find the surface of interest in a video-endoscopic image sequence (where each image shows a litt le area of examined tissue surface) due to: (1) the big number of images into the sequence, (2) the sequential nature of fi le (forcing doctor to review all sequence to find the interest region), and (3) frequently the interest region that could be physically contiguous is shown in frames far-away from each other. In this work, a methodology for automatic cartography reconstruction of the interest region is proposed. This methodology provides a panoramic view and eliminates redundant information. Furthermore, an automatic image registration algorithm for cartography construction is presented. Finally, an evaluation protocol and the obtained results are also presented.