Automated whole slide image analysis for a translational quantification of liver fibrosis

Abstract

Abstract Current literature highlights the need for precise histological quantitative assessment of fibrosis which cannot be achieved by conventional scoring systems, inherent to their discontinuous values and reader-dependent variability. Here we used an automated image analysis software to measure fibrosis deposition in two relevant preclinical models of liver fibrosis, and established correlation with other quantitative fibrosis descriptors. Longitudinal quantification of liver fibrosis was carried out during progression of post-necrotic (CCl4-induced) and metabolic (HF-CDAA feeding) models of chronic liver disease in mice. Whole slide images of picrosirius red-stained liver sections were analyzed using a fully automated, unsupervised software. Fibrosis was characterized by a significant increase of collagen proportionate area (CPA) at weeks 3 (CCl4) and 8 (HF-CDAA) with a progressive increase up to week 18 and 24, respectively. CPA was compared to collagen content assessed biochemically by hydroxyproline assay (HYP) and by standard histological staging systems. CPA showed a high correlation with HYP content for CCl4 (r = 0.8268) and HF-CDAA (r = 0.6799) models. High correlations were also found with Ishak score or its modified version (r = 0.9705) for CCl4 and HF-CDAA (r = 0.9062) as well as with NASH CRN for HF-CDAA (r = 0.7937). Such correlations support the use of automated digital analysis as a reliable tool to evaluate the dynamics of liver fibrosis and efficacy of antifibrotic drug candidates in preclinical models.