Hepatic Steatosis Detection

The purpose of this study was to develop and evaluate a deep learning model specifically designed to detect hepatic steatosis using chest radiographs. Using a large, retrospectively collected dataset from two institutions, the model was trained and tested on over 6,500 radiographs and demonstrated strong diagnostic performance, with AUCs of 0.83 and 0.82 on internal and external test sets, respectively. These findings highlight the potential of chest radiographs, when paired with deep learning, to serve as a noninvasive tool for identifying hepatic steatosis.