Artificial Intelligence-enabled Chest X-ray Classifies Osteoporosis and Identifies Mortality Risk.

Area of Science:

• Radiology
• Artificial Intelligence
• Osteoporosis Research

Background:

• Osteoporosis diagnosis typically relies on Dual-energy X-ray Absorptiometry (DXA), which is not universally accessible.
• Chest X-rays (CXRs) are widely available and may contain features indicative of osteoporosis.
• Early detection of osteoporosis is crucial for timely intervention and risk management.

Purpose of the Study:

• To develop and validate a deep learning model (DLM) for identifying osteoporosis from CXR features.
• To investigate the prognostic implications of AI-detected osteoporosis, specifically its association with all-cause mortality.
• To evaluate the potential of an AI-enabled CXR strategy as an early screening tool for osteoporosis.

Main Methods:

• A dataset of 48,353 CXRs with corresponding DXA T-scores was curated.
• A DLM was trained on 35,633 CXRs to identify osteoporosis (CXR-OP) and validated on 12,720 CXRs.
• Model performance was assessed using the area under the receiver operating characteristic curve (AUC); mortality risks were analyzed using survival analysis and Cox proportional hazards models.

Main Results:

• The DLM achieved high accuracy, with AUCs of 0.930 (internal validation) and 0.892 (external validation).
• Individuals identified as CXR-OP demonstrated a significantly higher risk of all-cause mortality (HR 2.59 internal, HR 1.67 internal without DXA).
• Similar mortality risks were observed in the external validation set, confirming the prognostic value.

Conclusions:

• A DLM utilizing CXR features can accurately detect osteoporosis.
• AI-enabled CXR analysis offers significant prognostic value, identifying individuals at increased risk of mortality.
• This AI-driven CXR strategy shows promise as a scalable and accessible screening tool for early osteoporosis detection.