Segmentation of visceral and subcutaneous adipose tissue in abdominal CT-datasets with and without contrast medium: Influence of iterative reconstruction on 2D- and 3D-segmentation

Abstract

Purpose

Segmentation of visceral and subcutaneous adipose tissue in computed tomography (CT) datasets has shown much promise in research and for medical applications, e.g. for risk stratification and guiding therapies. This study evaluates the influence of iterative reconstruction (IR) and filtered back projection (FBP) techniques on 2D- and 3D-segmentation of adipose tissue in CT images with and without contrast medium.

Methods

We retrospectively analyzed 31 patients to compare adipose tissue density and quantity between IR and FBP across different compartments and contrast phases. Segmentation was performed using a fixed threshold (-190 to -30 HU).

Results

Significant differences were observed in 2D-segmentation, particularly for visceral adipose tissue in non-enhanced scans (-0.54 ± 1.4 HU; p = 0.04) and subcutaneous adipose tissue in venous scans (-0.48 ± 1.2 HU; p = 0.03). In 3D-segmentation, subcutaneous adipose tissue density in venous scans was also lower with IR compared to FBP (-0.67 ± 1.2 HU; p = 0.004).

Conclusion

Adipose tissue segmentation between IR and FBP revealed minimal and only occasionally yields statistically significant differences in density and quantity across adipose tissue compartments and contrast phases. The observed differences were very small, casting doubt on their clinical relevance at the level of individual patients. However, even subtle systematic variations may warrant consideration in population-based studies or longitudinal research where methodological consistency is critical.