Opportunistic Osteoporosis Screening in Breast Cancer Using AI-Derived Vertebral BMD from Routine CT: Validation...

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Breast cancer survivors have an increased risk of bone loss due to treatments.
• Current bone health assessments are underutilized in this population.
• Opportunistic bone density measurement from routine CT scans offers a potential solution.

Purpose of the Study:

• To validate artificial intelligence-derived vertebral bone mineral density (AI-vBMD) from non-contrast CT scans for osteoporosis screening in breast cancer patients.
• To assess the diagnostic value of AI-vBMD compared to clinical variables alone.

Main Methods:

• Retrospective study of 332 breast cancer patients.
• AI-vBMD extracted from non-contrast thoracoabdominal CT scans.
• Quantitative CT (QCT) used as the reference standard for validation.
• Statistical analysis included correlation, Bland-Altman analysis, weighted kappa, logistic regression, and decision-curve analysis.

Main Results:

• AI-vBMD showed strong correlation (r=0.98) and excellent agreement (weighted κ=0.90) with QCT-vBMD.
• AI-vBMD alone demonstrated high diagnostic discrimination for osteoporosis (AUC=0.986).
• Integrating AI-vBMD into clinical models significantly improved diagnostic performance (AUC=0.988 vs. 0.879).

Conclusions:

• AI-vBMD from routine CT is a reliable imaging biomarker for opportunistic osteoporosis screening in breast cancer survivors.
• AI-vBMD provides significant incremental diagnostic value beyond clinical information.
• This approach can help address the underutilization of bone health assessments in this patient group.