ODP070 A deep learning-based osteopenia and osteoporosis screening system with lateral lumbar spine X-ray in women aged 50 years or older and men aged 60 years or older

Abstract

Abstract Bone mineral density (BMD) measured by Dual-energy X-ray absorptiometry (DXA) is the gold standard to diagnose osteopenia or osteoporosis. However, this method is not always available and underutilized to screen for osteoporosis. An artificial intelligence system using plain radiography is suggested in this study to enable easy and convenient prediction. This study is conducted on Korean women and men with data derived from two university hospitals (Ajou University and Seoul National University). Retrospective data from the years 2005 to 2021 with 20,100 subjects (women 60%) were analyzed. GE Healthcare and Hologic DXA, and lumbar spine lateral X-rays were used for performance, approved by the Institutional Review Boards. Inappropriate vertebral segments (compression fracture, degenerative change, etc.) were excluded from the analysis. A total of 79,400 lumbar spine segments (L1, L2, L3, and L4) were extracted from lumbar spine BMD data are compared and analyzed. BMD is classified into three categories: normal, osteopenia, and osteoporosis by World Health Organization criteria. 34% of segments are classified with osteopenia, and 24% osteoporosis. The area under the curve of the receiver operating characteristic of training results of deep learning artificial intelligence demonstrates predictability for osteopenia of 82% and osteoporosis of 95%. A deep learning algorithm may be regarded as one of the valuable and accessible screening methods for predicting osteopenia and osteoporosis with plain X-rays in clinical settings. This study was supported by funding from the National Information Society Agency and the Ministry of Science and ICT of the Republic of Korea. Presentation: No date and time listed