Image segmentation of liver fibrosis

Quantitation of connective tissue versus parenchymal tissue compartments is important in assessment of fibrosis (scarring) and its progression in liver disease. This paper presents a two-step algorithm for quantifying fibrosis in liver biopsies stained with Sirius red. With this staining technique, collagen and cell nuclei appear similarly stained, whereas cytoplasm appears pale. The first step of the algorithm is to separate similarly stained areas occupied by collagen and hepatocyte nuclei. Since the total area of the combined collagen and cell nuclei is usually much smaller than the remaining liver parenchyma, a non-linear intensity mapping is applied to enhance the smaller cluster in order to match the larger one in both intensity and size. The second step is to differentiate the fibrotic areas usually having irregular shapes from hepatocyte nuclei that have relatively uniform size and have circular shape. The proposed algorithm has been applied to quantify the development of progressive fibrosis and its possible regression in liver biopsy specimens from patients with parenteral-induced liver injury undergoing intestinal transplantation.