Quantification of abdominal fat from computed tomography using deep learning and its association with electronic health records in an academic biobank

• Published in: Journal of the American Medical Informatics Association : JAMIA 2021-06, Vol.28 (6), p.1178-1187
• Main Authors: MacLean, Matthew T, Jehangir, Qasim, Vujkovic, Marijana, Ko, Yi-An, Litt, Harold, Borthakur, Arijitt, Sagreiya, Hersh, Rosen, Mark, Mankoff, David A, Schnall, Mitchell D, Shou, Haochang, Chirinos, Julio, Damrauer, Scott M, Torigian, Drew A, Carr, Rotonya, Rader, Daniel J, Witschey, Walter R
• Format: Article
• Language: English
• Subjects: Abdominal Fat; Biological Specimen Banks; Deep Learning; Electronic Health Records; Humans; Research and Applications; Tomography, X-Ray Computed
• ISSN: 1527-974X; 1067-5027; 1527-974X
• DOI: 10.1093/jamia/ocaa342

Abstract Objective The objective was to develop a fully automated algorithm for abdominal fat segmentation and to deploy this method at scale in an academic biobank. Materials and Methods We built a fully automated image curation and labeling technique using deep learning and distributive computing to identify subcutaneous and visceral abdominal fat compartments from 52,844 computed tomography scans in 13,502 patients in the Penn Medicine Biobank (PMBB). A classification network identified the inferior and superior borders of the abdomen, and a segmentation network differentiated visceral and subcutaneous fat. Following technical evaluation of our method, we conducted studies to validate known relationships with visceral and subcutaneous fat. Results When compared with 100 manually annotated cases, the classification network was on average within one 5-mm slice for both the superior (0.4 ± 1.1 slice) and inferior (0.4 ± 0.6 slice) borders. The segmentation network also demonstrated excellent performance with intraclass correlation coefficients of 1.00 (P