Enhanced opportunistic CT screening for osteoporosis using Machine learning derived volumetric vertebral and complementary body composition information.

Volumetric vertebral features significantly improved lumbar spine BMD prediction compared to single-slice lumbar attenuation alone.

• Lumbar spine correlation coefficient was 0.92 for the volumetric vertebral feature model versus 0.56 for single-slice L1 attenuation (P < 0.001)
• The study included 383 adults with a mean age of 59.8 years, 50.1% women, undergoing routine health check-ups with same-day abdomen CT and DXA
• A two-stage 3DnnU-Net was developed using 475 CT scans from clinical and public datasets to segment individual thoracolumbar vertebrae

Volumetric vertebral features significantly improved osteoporosis classification compared to lumbar vertebral attenuation alone.

• AUROC was 0.95 for the volumetric model versus 0.87 for single-slice L1 attenuation (P = 0.004)
• The comparison was made against conventional linear regression using single-slice lumbar (L1) attenuation
• Models were built to estimate DXA-derived lumbar spine, femoral neck, and total hip BMD

Adding body composition metrics to volumetric vertebral features significantly increased sensitivity for osteoporosis classification while maintaining high specificity.

• Sensitivity increased from 76% to 86% when body composition metrics were added (P = 0.046)
• High specificity of 95% was maintained after adding body composition features
• Muscle and fat were segmented using a predeveloped 3D U-Net called DeepCatch
• Body composition features also further enhanced hip BMD predictions

Incorporating clinical variables (age, sex, BMI) provided no additional benefit beyond volumetric vertebral and body composition features.

• Three model types were compared: vertebral features alone, combined vertebral and body composition features, and these features plus clinical data (age, sex, body mass index)
• Adding clinical variables to the combined vertebral and body composition model did not further improve BMD prediction or osteoporosis classification
• This was assessed across lumbar spine, femoral neck, and total hip BMD predictions

The study used a retrospective design with same-day CT and DXA measurements as the reference standard for BMD.

• 383 adults undergoing routine health check-ups were included
• All participants had same-day abdomen CT scans and dual-energy X-ray absorptiometry (DXA)
• Mean participant age was 59.8 years with 50.1% women
• The segmentation model was trained on 475 CT scans from both clinical and public datasets