Automatic extraction of semantic metadata from optical biopsy images

Abstract

Anglès: Confocal Laser Microendoscropy is a novel medical imaging technique to obtain non-super cial optical sections of organic tissue without physically disrupting it. When applied to human colon endoscopy examination, the explorer can visually detect in vivo anomalies in real time. However, he or she may lack training in interpreting the resulting images. This project aims to help by speeding up the learning curve, automatically analysing them. A set of numerical descriptors is algorithmically extracted, and relevant tissue objects are highlighted. Finally, a classi cation based on their degree of illness is proposed and tested.