Automating measurement of renal interstitial fibrosis: Effect of colour spaces on quantification

The progression of chronic renal diseases is primarily measured by the extent of interstitial fibrosis and glomerulosclerosis in renal biopsies. The traditional method of interstitial fibrosis quantification uses visual evaluation, which presents high inter- and intra-observer variability. In this paper, we investigate automated quantification methods based on various colour spaces in renal biopsy images. The system identifies and extracts structures in the renal biopsy image based on the colour information presented by the image following a set of knowledge-based rules on the different structures in a microscopic renal biopsy image. The quantification results indicate that a hybrid method which incorporates different colour spaces to segment tissue structures achieves higher accuracy compared to thresholding or clustering methods. It reduces the error of single colour space methods by half, achieving a mean error of 6 percentage points.