Advanced Block Detection and Quantification of Fibrotic Areas in Microscopy Images of Obstructive Nephropathy

Obstructive nephropathy is not a rare disease and experts need a tool, which will provide them fast and accurate reproducible results for disease assessment. In this work we deal with the analysis of biopsy images for the detection and quantification of obstructive nephropathy. The problem is analyzed on a 3-stage approach. Block based segmentation is applied on the images. Image characterization is achieved through the classification of the informative part of the image utilizing Random Forests classifiers. The second approach deals with characterization of each block separately. Each block was classified with the above classifier and the majority vote of the blocks characterized the whole image. Additionally, a scoring system, based on the characterization of the segmentation blocks, was developed in order to describe and quantify the pathology in an image.