Automatic Identification of Pathological Bio-markers in Histological Images

Automatic methods for visual and spatial discrimant pattern identification, at histological level

In medicine, differential diagnosis consists in the disease identification through the exclusion of possible causes that presents a similar clinical profile. In the histology domain, this process involves the analysis of pathologies that show similar distorsions in the tissue structure and functionality. The presence or abscence of visual features related to this distorsions (pathological markers) in histologic images, are used as evidence for a specific pathology diagnosis. In this project, we explore automatic methods for visual and spatial discrimant pattern identification in the histology domain. In adittion, the patterns should be interpretable. Interpretability is defined as the possibility to associate the visual/spatial pattern to concepts in the histology domain recognized by expert pathologist. The finding of discriminant and interpretable patterns, could sugest the existence of new pathological bio-markers that would allow differential diagnosis in histologic images.