Renal corpuscle morphometry with increased reliability and high level of automation

The paper presents an original computerized histometric technique for renal biopsies that outlines the role of quantitative information in the morphologic analysis of renal corpuscle (RC). We investigated 10 cases diagnosed as rapidly progressive glomerulonephritis (RPGN) by randomly selecting 10 RCs from each case.Our technique is based on two algorithms (implemented in the Zeiss KS400) that process digital color images for extracting the morphometric features at two levels: (1) of a single RC and (2) of a whole case. The algorithm operating at level (1) includes instruments especially designed for ensuring the accuracy in the identification of the regions occupied by the RC components; it evaluates the areas of the RC components and their percentages relative to the total area of the RC. The novelty of the instruments lies in the usage of intermediate recoloring for multifold thresholding segmentation and in the mutual validation of the regions provided by red-green-blue (RGB) and hue-lightness-saturation (HLS) segmentation. The algorithm operating at level (2) averages the data obtained from all the RCs considered in a case. The proposed technique offers a high level of automation; it is reliable and repeatable. The basic principles of the presented technique can also be adapted to the histometric approaches of other pathology domains. Results illustrate the utilization of the technique at both levels.The extraction of the RC morphometric features allows for establishing a significant correlation between the quantitative and qualitative information. This may contribute to a better understanding of the RPGN evolution and, implicitly, to the refinement of the diagnosis.