A MACHINE LEARNING-BASED APPROCH FOR QUANTITATIVE AND AUTOMATED NON-ALCOHOLIC FATTY LIVER DISEASE (NAFLD)/NON-ALCOHOLIC STEATOHEPATITIS (NASH) ASSESSMENT USING PATHOLOGICAL STAINED SLIDES

Abstract

Liver is the second largest organ in human body, which has many essential functions in metabolism. Fully automated assessments were adapted to develop an improved system for diagnosing another liver disease: non-alcoholic fatty liver disease (NAFLD)/non-alcoholic steatohepatitis (NASH), based on Hematoxylin and Eosin (H&E) and Sirius Red (SR) stained samples and machine learning. This machine learning based assessment can automatically and quantitatively classify different stages of NAFLD/NASH using only conventional stained liver sample on mice model, which has the potential to be applied in clinical use to release doctors’ burden from spending a lot of time to classify the stages of NAFLD/NASH by analyzing thousands of medical images manually.