CT-Based Body Composition Measures and Systemic Disease: A Population-Level Analysis Using Artificial Intelligence...

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Computed tomography (CT)-based body composition analysis is linked to health outcomes.
• Artificial intelligence (AI) enables automated body composition measurement in large patient cohorts.
• Understanding demographic and disease-related variations in body composition is essential.

Purpose of the Study:

• To evaluate the associations between age, sex, and common systemic diseases with CT-derived body composition metrics.
• To assess the performance of automated AI tools in quantifying body composition parameters.
• To establish a foundation for normative reference ranges in AI-based body composition analysis.

Main Methods:

• Retrospective analysis of 140,606 adult abdominal CT scans.
• Application of 13 automated AI tools for quantifying liver, spleen, kidney, vertebral, muscle, and fat composition.
• Electronic health record review to identify systemic diseases like cancer, cardiovascular disease (CVD), diabetes mellitus (DM), and cirrhosis.

Main Results:

• Age, sex, and systemic diseases were significant predictors for all 13 body composition measures (p < .001).
• Age predicted all measures; sex predicted 12; cancer predicted 9; CVD predicted 11; DM predicted 13; cirrhosis predicted 12.
• Models showed variable goodness of fit (R² = 0.03-0.43).

Conclusions:

• Demographic factors (age, sex) and common systemic diseases are significant predictors of AI-derived CT body composition measures.
• These findings are critical for developing clinical reference ranges for AI-based body composition tools.
• AI tools offer a scalable approach to body composition assessment in diverse patient populations.