Reliability and validity of the new VikingSlice software for computed tomography body composition analysis

Body composition assessment by computed tomography (CT) is increasingly used for diagnostic and prognostic purposes in various patient groups. This study aimed to compare the reliability and validity of a newly in-house developed segmentation software VikingSlice against a commercial software (SliceOMatic) for quantification of adipose tissue and skeletal muscle cross-sectional areas (CSA).

Fifty abdominal CT sets from chronic pancreatitis patients were analyzed (mean age 49, range 27-84 years; 38 males). Soft tissue CSAs at level of 4th lumbar vertebra were assessed by measuring standard Hounsfield unit threshold definitions with both softwares. Analysis with VikingSlice included automatic segmentation of interested region with subsequent manual corrections. Analysis with SliceOMatic included manual segmentation of each area. Same investigator measured CSAs using both programs. Inter-observer reliability of CSAs measurements with VikingSlice were assessed by comparing results from two independent investigators. Measurements were compared using the intra-class correlation coefficient (ICC), coefficient of variation (CV), Jaccard index and Bland-Altman analyses.

The inter-observer reliability of VikingSlice was excellent (CV 3.4-15.4%, ICC 0.979-0.999, Jaccard index 0.68-0.98). Validity was high (CV 1.6-10.2%, ICC 0.950-0.997) for measurements by SliceOmatic and VikingSlice. The findings were supported in the Bland-Altman plots. The reliability study had small average differences with means of soft tissue compartments in range -2.29 cm² to 1.56 cm²; average differences between both softwares were -1.28 cm² to 0.31 cm².

The in-house developed software VikingSlice was fast and showed good reliability that is comparable with commercial software in its utility to estimate adipose tissue and skeletal muscle CSAs.