Detecting spongiosis in stained histopathological specimen using multispectral imaging and machine learning

Abstract

Pathologists spend nearly 80% of their time analysing pathological tissue samples. In addition, the diagnosis is subject to inter/intra-observer variability. Thus to increase productivity and repeatability, a new field known as Computational Pathology has emerged which combines the field of pathology with computer vision, pattern recognition and machine learning. This research develops a new computational pathology framework specifically to aid with detecting a condition known as spongiosis caused by Newcastle Disease Virus infection in poultry. It combines the use of multispectral imaging with feature extraction and classification to detect areas of spongiosis in tissue of infected poultry. The success of this framework is the first step towards a completely automated diagnosis tool for histopathology.