Objectives While vertebral fracture assessment (VFA) on dual-energy X-ray absorptiometry (DXA) images is widely used to detect osteoporotic vertebral fractures, existing deep learning approaches typically perform either fracture detection or vertebral localization, not both. We developed VERTE-X L, a deep learning model that simultaneously localizes vertebrae (T7–L5) and detects fractures on DXA VFA images, and evaluated its performance in a community-dwelling older adult cohort. Methods We used DXA VFA images from 1000 randomly sampled participants (age ≥65 years) in the Korean Urban Rural Elderly (KURE) cohort, split into training (N = 800) and test (N = 200) sets. VERTE-X L employs a DINOv2-L encoder, a localization decoder for per-vertebra coordinate prediction, and a consistent rank logits (CORAL) ordinal regression module for fracture detection. Performance was evaluated at image and vertebra levels and compared with previously reported detection-only models (VERTE-X and VERTE-X model fine-tuned on VFA data) and localization-only models (VerteNet and Multi-Context Hybrid CNN Transformer). Results On the held-out test set, VERTE-X L achieved an image-level area under the receiver operating characteristic curve (AUROC) of 0.952 (sensitivity 89.7%, specificity 94.2%, net present value [NPV] 98.2%) and vertebra-level AUROC of 0.957 (NPV 99.7%). Localization success (within 15 pixels) was 94.7% for fracture vertebrae. VERTE-X L outperformed VERTE-X in detection (AUROC 0.952 vs 0.913, P = 0.034) and outperformed VerteNet in localization (mean radial error 10.0 vs. 20.3 px, P < 0.001). Conclusions The proposed method enables simultaneous fracture detection and vertebral localization on DXA VFA images.
Objectives To re-evaluate fall incidence and its determinants in a contemporary cohort of ambulatory women with rheumatoid arthritis (RA) in a follow-up study based on our previously published cohort data. Methods We prospectively evaluated 124 ambulatory female patients aged ≥50 years with RA at a single institution in 2022–2023, monitoring falls for 1 year. Outcomes included patient characteristics, sarcopenia, disease activity, and physical function. Multivariable logistic regression and Firth-penalized logistic regression were used in sensitivity analysis to assess the interaction between sarcopenia and glucocorticoid (GC) use. Results The annual fall rate in the 2022–2023 cohort was lower than in our previous cohort (27.4% vs. 50.0%). History of falls (odds ratio [OR] 2.76, 95% confidence interval [CI] 1.02–7.45, P = 0.045) and hypnotics/anxiolytics use (OR 3.92, 95% CI 1.01–15.20, P = 0.048) were independent risk factors for falls. A significant interaction between sarcopenia and GC use was observed in the initial analysis but was no longer statistically significant in sensitivity analyses. Conclusions Compared with our previously reported cohort, the contemporary RA cohort showed a lower fall rate and a different pattern of associated risk factors. Further research is needed to explore the potential interaction between sarcopenia and glucocorticoid use. The observed interaction should be considered exploratory.
Objectives To investigate whether an artificial intelligence (AI) model could demonstrate high diagnostic accuracy in identifying dual-energy X-ray absorptiometry (DXA) confirmed osteoporosis, outperform traditional human-read metrics, and deliver substantial clinical net benefit. Methods This retrospective multicenter cohort study included 382 patients with hip fractures who had both preoperative pelvic radiographs and postoperative DXA scans. An AI model analyzed the intact contralateral proximal femur to predict both bone mineral density (BMD) values and T-scores. Diagnostic performance was evaluated against DXA-confirmed osteoporosis (T-score ≤−2.5) and compared with manual cortical thickness index (CTI) and femoral neck width (FNW) measurements using Spearman rank-order correlation analysis, receiver operating characteristic (ROC) curves, the DeLong test, and decision curve analysis (DCA). Results The AI model showed a strong correlation with DXA-derived BMD values for both the femoral neck (ρ = 0.76, R2 = 0.55, mean absolute error (MAE) = 0.06 g/cm2) and the total hip (ρ = 0.69, R2 = 0.51, MAE = 0.08 g/cm2). Correspondingly, the AI-predicted T-scores correlated well with DXA-derived T-scores for the femoral neck (ρ = 0.76, R2 = 0.55, MAE = 0.47) and the total hip (ρ = 0.69, R2 = 0.51, MAE = 0.65). The AI model demonstrated excellent discriminative performance, achieving an area under the ROC curve (AUC) of 0.88 (95% confidence interval [CI] 0.83–0.93) for the femoral neck and 0.87 (95% CI 0.82–0.92) for the total hip, both significantly higher than CTI (0.72 [95% CI 0.65–0.79]; P < 0.001) and FNW (0.53 [95% CI 0.44–0.62]; P < 0.001). DCA illustrated that the AI models provided greater clinical net benefit than manual measurements, remaining clinically advantageous up to a threshold probability exceeding 0.90. Conclusions Deep learning AI offers a superior, automated alternative to traditional manual morphometric measurements for identifying osteoporosis on routine pelvic radiographs in patients with hip fractures, potentially facilitating timely secondary fracture prevention.
