Since the development of capsule endoscopcy technology, substantial progresswere made in converting passive capsule endoscopes to robotic active capsuleendoscopes which can be controlled by the doctor. However, robotic capsuleendoscopy still has some challenges. In particular, the use of such devices togenerate a precise and globally consistent three-dimensional (3D) map of theentire inner organ remains an unsolved problem. Such global 3D maps of innerorgans would help doctors to detect the location and size of diseased areasmore accurately, precisely, and intuitively, thus permitting more accurate andintuitive diagnoses. The proposed 3D reconstruction system is built in amodular fashion including preprocessing, frame stitching, and shading-based 3Dreconstruction modules. We propose an efficient scheme to automatically selectthe key frames out of the huge quantity of raw endoscopic images. Together witha bundle fusion approach that aligns all the selected key frames jointly in aglobally consistent way, a significant improvement of the mosaic and 3D mapaccuracy was reached. To the best of our knowledge, this framework is the firstcomplete pipeline for an endoscopic capsule robot based 3D map reconstructioncontaining all of the necessary steps for a reliable and accurate endoscopic 3Dmap. For the qualitative evaluations, a real pig stomach is employed. Moreover,for the first time in literature, a detailed and comprehensive quantitativeanalysis of each proposed pipeline modules is performed using a non-rigidesophagus gastro duodenoscopy simulator, four different endoscopic cameras, amagnetically activated soft capsule robot (MASCE), a sub-millimeter preciseoptical motion tracker and a fine-scale 3D optical scanner.
展开▼
