Beyond BMI: Deep Learning Segmentation-Driven CT Reveals Body Composition Changes after Metabolic and Bariatric...

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Bariatric surgery outcomes

Background:

• Body mass index (BMI) is standard for metabolic and bariatric surgery (MBS) outcomes but doesn't differentiate tissue compartments or visceral adiposity (VAT).
• Visceral adiposity (VAT) is a key determinant of cardiometabolic risk.

Purpose of the Study:

• Evaluate the relationship between BMI and VAT.
• Characterize compartment-specific tissue remodeling after MBS using AI-enabled CT segmentation.

Main Methods:

• Retrospective analysis of prospectively collected abdominal CT scans.
• Utilized Comp2Comp deep-learning pipeline for automated segmentation of visceral adipose tissue, subcutaneous adipose tissue, and skeletal muscle.
• Analyzed population (n=435) and longitudinal MBS cohorts (n=39) for BMI-VAT associations and temporal changes.

Main Results:

• BMI and VAT showed moderate correlation in the general population but not in patients with BMI ≥35 kg/m².
• Postoperatively, VAT reduction was more strongly associated with time than BMI decline.
• Skeletal muscle exhibited a distinct recovery trajectory.

Conclusions:

• BMI incompletely reflects postoperative tissue remodeling and VAT reduction after MBS.
• AI-enabled CT volumetric analysis offers compartment-specific assessment beyond BMI.
• Prospective validation is needed to confirm VAT-derived metrics for predicting cardiometabolic outcomes.