Machine-learning modeL based on computed tomography body composition analysis for the estimation of resting energy...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Metabolic Research

Background:

• Assessing resting energy expenditure (REE) is crucial but challenging with current methods.
• Indirect calorimetry (IC) is the gold standard but lacks availability.
• Predictive equations and bioimpedance (BIA) show limited agreement with IC.

Purpose of the Study:

• To explore the utility of computed tomography (CT) scans for REE assessment using AI.
• To evaluate if CT-based body composition analysis can reliably estimate REE.

Main Methods:

• A pilot study involving 90 adults assessed REE using equations, IC, BIA, and skeletal CT-scan.
• An AI machine-learning model (second-order linear regression) was developed using CT data.
• The model was trained and validated using cross-validation.

Main Results:

• The CT-based AI model demonstrated excellent agreement with IC (bias 0 kcal/day).
• This CT-AI method outperformed traditional predictive equations and showed similar agreement to BIA.
• Key variables like gender and BMI were not significant for the CT-AI REE estimation.

Conclusions:

• CT-scan image analysis via AI machine learning is a promising tool for REE estimation.
• This approach could potentially revolutionize nutritional assessment guidelines.
• AI-powered CT offers a reliable alternative for REE evaluation.