An original research from Radiology evaluated the application of deep-learning for abdomen segmentation on 2D CT sections, in order to quantify body composition. Results showed that the automated algorithm met or even exceeded the accuracy of manual segmentation performed by expert radiologists.