Enhanced Muscle and Fat Segmentation for CT-Based Body Composition Analysis: A Comparative Study

TL;DR

This study compares an internally developed CT body composition segmentation tool to the public TotalSegmentator using 900 SAROS CT series. It employs Dice similarity for subcutaneous fat and muscle, Cohen's Kappa for visceral fat, and $R^2$ correlations with ground-truth annotations, finding the Internal tool achieves higher Dice scores ($83.8$ vs $80.8$ for subcutaneous fat; $87.6$ vs $83.2$ for muscle) with $p<0.01$, and a visceral-fat agreement of $\kappa=0.856$. The tool also shows strong muscle metrics ($R^2=0.99$ for volume, $R^2=0.93$ for attenuation) and subcutaneous fat volume ($R^2=0.99$), with moderate fat attenuation ($R^2=0.45$). Together, these results indicate the Internal tool improves segmentation accuracy for body composition from CT and holds promise for enhanced risk assessment in cardiovascular events and fractures.

Abstract

Purpose: Body composition measurements from routine abdominal CT can yield personalized risk assessments for asymptomatic and diseased patients. In particular, attenuation and volume measures of muscle and fat are associated with important clinical outcomes, such as cardiovascular events, fractures, and death. This study evaluates the reliability of an Internal tool for the segmentation of muscle and fat (subcutaneous and visceral) as compared to the well-established public TotalSegmentator tool. Methods: We assessed the tools across 900 CT series from the publicly available SAROS dataset, focusing on muscle, subcutaneous fat, and visceral fat. The Dice score was employed to assess accuracy in subcutaneous fat and muscle segmentation. Due to the lack of ground truth segmentations for visceral fat, Cohen's Kappa was utilized to assess segmentation agreement between the tools. Results: Our Internal tool achieved a 3% higher Dice (83.8 vs. 80.8) for subcutaneous fat and a 5% improvement (87.6 vs. 83.2) for muscle segmentation respectively. A Wilcoxon signed-rank test revealed that our results were statistically different with p<0.01. For visceral fat, the Cohen's kappa score of 0.856 indicated near-perfect agreement between the two tools. Our internal tool also showed very strong correlations for muscle volume (R^2=0.99), muscle attenuation (R^2=0.93), and subcutaneous fat volume (R^2=0.99) with a moderate correlation for subcutaneous fat attenuation (R^2=0.45). Conclusion: Our findings indicated that our Internal tool outperformed TotalSegmentator in measuring subcutaneous fat and muscle. The high Cohen's Kappa score for visceral fat suggests a reliable level of agreement between the two tools. These results demonstrate the potential of our tool in advancing the accuracy of body composition analysis.

Figures (6)

• Figure 1: Example axial, coronal and sagittal slice of from SAROS dataset. In the axial slice, the muscle (yellow), subcutaneous fat (red) and abdominal cavity (green) are shown. The grey regions in the coronal/sagittal views indicate no segmentation masks available in that area, while the streaks in between them contain segmentations.
• Figure 2: Violin plot of TotalSegmentator (green) vs. our internal tool (blue) for the segmentation of (a) subcutaneous fat and (b) muscle.
• Figure 3: $R^2$ correlation plots of the automatic segmentation results compared against ground truth annotations. Top Row: TotalSegmentator (TS). Bottom Row: Internal (Int) tool. L-to-R: Muscle Volume, Muscle Attenuation, Fat Volume, Fat Attenuation.
• Figure 4: Bland-Altman plots of the volume measurements between the automatic segmentations compared against manual annotations. L-to-R: TotalSegmentator Muscle Volume, Internal Muscle Volume, TotalSegmentator Subcutaneous Fat, Internal Subcutaneous Fat.
• Figure 5: Example segmentation of . Top-to-Bottom: axial, coronal, sagittal views. L-to-R: CT image, manual annotation (ground truth), TotalSegmentator segmentation, Internal tool segmentation. Red: Subcutaneous Fat, Yellow: Muscle, Green: Internal Abdominal Cavity (ground truth only) / Visceral Fat, Grey: No ground truth labels. Blue arrows shows over segmentation of subcutaneous fat by TotalSegmentator where it was correctly segmented as muscle by our Internal tool.