Spine age estimation using deep learning in lateral spine radiographs and DXA VFA to predict incident fracture and
Sang Wouk Cho1,2, Namki Hong3,4,5,6, Kyoung Min Kim7
1Institute for Innovation in Digital Healthcare (IIDH), Yonsei University Health System, Seoul, Korea.
Abstract:
Spine age estimated from lateral spine radiographs and DXA VFAs could be associated with fracture and mortality risk. In the VERTE-X cohort (n = 10,341, derivation set) and KURE cohort (n = 3517; external test set), spine age discriminated prevalent vertebral fractures and osteoporosis better than chronological age. Predicted age difference was associated with overall (adjusted HR [aHR] 1.22 per 1 SD increment, p < 0.001), vertebral, non-vertebral incident fractures, and mortality (aHR 1.31, p = 0.001) during a median 6.6 years follow-up in KURE, independent of chronological age and covariates. Spine age to estimate FRAX hip fracture probabilities, instead of chronological age, improved the discriminatory performance for incident hip fracture (AUROC 0.83 vs. 0.78, p = 0.027). Shorter height, lower femoral neck BMD, diabetes, vertebral fractures, and surgical prosthesis were associated with higher predicted age difference, explaining 40% of variance. Spine age estimated from lateral spine radiographs and DXA VFA enhanced fracture risk assessment and mortality prediction over chronological age.
