Comparison between automated and manual segmentation in computed tomography for body composition analysis.

Area of Science:

• Radiology and Medical Imaging
• Biomedical Engineering
• Clinical Diagnostics

Background:

• Computed tomography (CT) at the third lumbar vertebra (L3) is the gold standard for body composition (BC) assessment.
• Manual segmentation of CT scans is time-consuming and impractical for large datasets.
• Automated methods are needed to improve efficiency in BC analysis.

Purpose of the Study:

• To compare the accuracy and efficiency of an automated body composition (BC) analysis tool, Identificação Automatizada da Composição Corporal (IACC), against manual segmentation.
• To evaluate the reproducibility and agreement of automated BC measurements.
• To assess the potential of IACC for clinical and large-scale population studies.

Main Methods:

• A retrospective cross-sectional study of 126 participants using single axial CT slices at L3.
• Manual segmentation performed using 3D Slicer, validated by a radiologist.
• Automated segmentation of skeletal muscle (MUSCLE), subcutaneous adipose tissue (SAT), and visceral adipose tissue (VAT) using IACC.

Main Results:

• IACC analysis time was ~5 minutes per scan, versus 25-30 minutes for manual segmentation.
• Excellent intra-rater reproducibility (ICC > 0.99) and high inter-rater agreement (Dice coefficients > 0.93) for SAT and VAT.
• High concordance (ICC > 0.93) for SAT and VAT, and good concordance (ICC = 0.784) for skeletal muscle.
• Skeletal muscle measurements showed greater variability and disagreement compared to adipose tissue compartments.

Conclusions:

• IACC provides accurate and efficient automated segmentation of subcutaneous and visceral adipose tissue.
• The automated tool substantially reduces analysis time for body composition assessment.
• While skeletal muscle measurements require further validation, IACC shows promise for large-scale population studies and clinical applications requiring rapid, standardized analysis.