Hepatic proton-density fat fraction (PDFF) has emerged as an imaging biomarker for liver fat content in non-alcoholic fatty liver disease. To account for the spatial heterogeneity of liver fat, one approach is to manually place a region of interest (ROI) and estimate the PDFF in each hepatic Couinaud segment separately. We trained a convolutional neural network to automatically locate each liver segment and calculate segmental PDFF values. We show that segmental PDFF measurement based on our automated approach closely matches those based on manual placement by a trained image analyst, demonstrating the feasibility of utilizing CNNs to automatically extract clinically valuable quantitative information from source images.
This abstract and the presentation materials are available to members only; a login is required.