Evaluation of Deep Learning-Based Quantitative Computed Tomography for Opportunistic Osteoporosis Screening

Abstract

Abstract Background To evaluate diagnostic efficacy of deep learning (DL)-based automated bone mineral density (BMD) measurement for opportunistic screening of osteoporosis with routine computed tomography (CT) scans. Methods A DL-based automated quantitative computed tomography (DL-QCT) solution was evaluated with 92 routine clinical CT scans from 65 patients who underwent either chest (N:29), lumbar spine (N:34), or abdominal CT (N:29) scan. The automated BMD measurements (DL-BMD) on L1 and L2 vertebral bodies from DL-QCT were validated with manual BMD (m-BMD) measurement from conventional asynchronous QCT using Pearson’s correlation and intraclass correlation. Receiver operating characteristic curve (ROC) analysis identified the diagnostic ability of DL-BMD for low BMD and osteoporosis, determined by dual-energy x-ray absorptiometry (DXA) and m-BMD. Results Excellent concordance were seen between m-BMD and DL-BMD in total CT scans (r = 0.960/0.980). The ROC-derived AUC of DL-BMD compared to that of central DXA for the low-BMD and osteoporosis patients was 0.840 and 0.784 respectively. The sensitivity, specificity, and accuracy of DL-BMD compared to central DXA for low BMD were 73.1%, 68.0%, and 71.7%, respectively, and those for osteoporosis were 78.9%, 83.6%, and 82.6%. The AUC of DL-BMD compared to the m-BMD for low BMD and osteoporosis diagnosis were 0.982 and 0.934, respectively. The sensitivity, specificity, and accuracy of DL-BMD compared to m-BMD for low BMD were 94.8%, 94.1%, and 94.6%, and those for osteoporosis were 73.3%, 91.9%, and 85.9%, respectively. Conclusions DL-BMD exhibited excellent agreement with m-BMD on L1 and L2 vertebrae in the various routine clinical CT scans and had comparable diagnostic performance for detecting the low-BMD and osteoporosis on conventional QCT.