sqFibrosis: A Robust Liver Fibrosis Scoring System for Telepathology

Yang Yu, Jiahao Wang, Hanry Yu

Abstract: Histopathological scoring is the gold standard for assessing liver fibrosis on biopsy images. However, accurate computational assessment of liver fibrosis suffers from data variations of the sampling error caused by staining and imaging heterogeneity in variable clinical lab settings, particularly in the context of computer-assisted diagnostics for telepathology applications. Experienced pathologists address these challenges by leveraging their expertise to reference samples, and selecting relevant spatial features while filtering out noise. Drawing from this domain knowledge, we present sqFibrosis (spatially-sensitive and quantitative fibrosis), a robust and automated scoring index with pathological-relevant features incorporated to provide quantitative analysis for staging liver fibrosis using conventional collagen-stained images and standard light microscopes. Specifically, sqFibrosis consolidates 63 architectural and morphological features from various collagen fibers (portal, fibrillar and septal collagen), resulting in a data-efficient single index for liver fibrosis evaluation. This index also demonstrates the robustness to sampling errors introduced from staining and imaging variations. Furthermore, it is well-suited for liver fibrosis assessment within telepathology settings.