Meeting Banner
Abstract #3395

Deep Learning Based Automatic Pipeline for Quantitative Assessment of Thigh Muscle Fatty Infiltration in Post-traumatic Osteoarthritis

Sibaji Gaj1, Brendan L. Eck1, Dongxing Xie1, Richard Lartey1, Charlotte Lo1, Mingrui Yang1, Kunio Nakamura1, Carl S. Winalski2, Kurt Spindler3, and Xiaojuan Li1
1Biomedical Engineering, Cleveland Clinic, Cleveland, OH, United States, 2Radiology, Cleveland Clinic, Cleveland, OH, United States, 3Orthopaedics, Cleveland Clinic, Cleveland, OH, United States


Quantitative assessment of thigh muscle morphology and fatty infiltration (FF%) in post-traumatic osteoarthritis is limited. In this work, we developed a deep learning based accurate segmentation method for muscles, bone and adipose tissue from thigh MRI and used these segmentation for automated quantification of FF and cross sectional area(CSA) of these tissues. 16 patients at 10 years after ACL reconstruction were studied. The proposed method showed significant improvement in segmentation metrics (Dice, Average surface distance (ASD)) and CSA compared with popular U-Net based deep learning models. For CSA and FF% quantification, automated methods had similar measurements compared with manual segmentation.

This abstract and the presentation materials are available to members only; a login is required.

Join Here