The Use of Classification Trees for Analysis of Kidney Images in Computer Assisted Microscopy

Abstract

The aim of this paper is to present an application of classification trees for quantitative analysis of kidney specimens in light and electron microscopy. Minimal change disease (MCD) cases, focal segmental glomerulosclerosis (FSGS) cases, and mesangial glomerulonephritis cases suspected of being progressive into MCD or FSGS were analysed in our study. At first, fuzzy logic was applied for processing and segmentation of structures in colour and grey scale images. Glomerular profiles, interstitial fibrosis and normal part of interstitium were segmented from colour light micrographs and then their area was estimated. The volume of glomerular mesangium, matrix and cell components were assessed in electron micrographs. These measurements were selected as predictor variables for classification trees. Results of classification presented in this paper indicated that the medical diagnosis of certain glomerulopathies should be performed in light and electron microscopy. The ratio of matrix volume to the whole mesangium determined in electron microscopy and the mean glomerular area estimated in light microscopy are the most important features for diagnosis of suspected cases as MCD or early phase of FSGS cases.