Detecting cancer regions in histopathology using multi-level features

Visual detection of cancerous areas within histopathological samples is still a problematic task, since it strongly relies on complex qualitative, thus highly subjective, analyses performed by expert histopathologists. Methods for automatic detection of cancerous regions are mostly based on tissue representation built on low level textural features, whose semantic level is far from the visual characteristics that histopathologists consider during the analysis. In this work we propose an algorithm for the detection of cancerous areas within lung and bladder adenocarcinoma samples, based on a multi-level representation of histopathological structures, in order to bridge the ”semantic gap” between histopathological considerations and machine representation.