Predicting Drug-Efficiency and Cancer Evolution Using an Automated Histological Classifification Based on Whole Slide Images
Yating Pan, Xinyi Liao, Xiang Ruan, Yating Deng, Fan Du, Guanzhen Yu, Xiaojun Wu · Research Square · 2022
Abstract Liver disease is a general term for all diseases that occur in the liver, including cirrhosis, tumors, and drug-induced injury. Liver pathology reveals the pathological appearance of various common liver diseases. Automated classification of liver pathological features based on whole slide images can help observe disease processes and assess drug efficacy, and reduce scientific fraud. Using pathological slides of rat liver induced by the chemical carcinogen TAA, we developed a deep-learning framework to detect abnormal lesions in rat liver from whole histopathological slides. Our framework provides an objective and reproducible method to observe liver lesions and also calculates the proportion of lesions in each liver slide. The framework can also clearly delineate the edges of the liver, revealing whether there are uneven, jagged or wavy surfaces. This tool not only discloses the multi-level pathological evolution of cholangiocarcinoma, but also presents a visual image of the protective effect administered by H2. Artificial intelligence-based algorithms are highly promising methods for studying cancer evolution, assessing drug efficiency, and reducing scientific misconduct.