Stratification of Liver Histologic Fibrosis using Machine Learning on MRI Radiomic Data and Clinical Features

Chronic liver diseases can lead to variable amounts of liver fibrosis, which impacts patient management and outcomes. Percutaneous liver biopsy is the clinical reference standard for assessment of liver fibrosis. However, biopsy is subject to sampling errors and poor patient acceptance. The aim of this study is to develop machine learning models to stratify the severity of biopsy-derived liver fibrosis using MR radiomic data and clinical data. Using clinical, routinely collected MRI and clinical data, our machine learning was able to stratify the severity of liver fibrosis with an AUROC of 0.71, demonstrating the feasibility of the machine learning approaches.