Body composition analysis, utilizing Computed Tomography (CT) scans, has become increasingly relevant in assessing patient health, particularly concerning conditions like sarcopenia and visceral obesity. These conditions, identified through CT-based measurements, are crucial indicators for treatment outcomes and survival rates across diverse patient demographics. Numerous software solutions are employed in this domain, each possessing unique features. However, the consistency and comparability of measurements derived from these different software programs remained unclear. This study addresses this gap by providing a comparative analysis of several prominent software tools used for body composition assessment in CT scans.
This comparative study rigorously evaluated the agreement between four distinct software programs: FatSeg, OsiriX, ImageJ, and sliceOmatic. Fifty abdominal CT scans, randomly selected from different patients, were independently analyzed by two trained observers. The core measurements included cross-sectional muscle area (CSMA), encompassing rectus abdominis, oblique and transverse abdominal muscles, paraspinal muscles, and the psoas muscle. Additionally, visceral adipose tissue area (VAT) and subcutaneous adipose tissue area (SAT) were quantified. Segmentation was performed using standardized Hounsfield unit ranges for accurate region of interest computation. To determine the level of agreement across software, observers, and within observers, intra-class correlation coefficients (ICCs) and Bland-Altman analyses were employed. Cohen’s κ was utilized to assess the consistency in diagnosing sarcopenia and visceral obesity. Furthermore, the Jaccard similarity coefficient was calculated to quantify the similarity and diversity of the measurements obtained from each software.
The findings revealed a high degree of comparability among the software programs. Both Bland-Altman analyses and ICC values demonstrated that CSMA, VAT, and SAT measurements were consistently similar across FatSeg, OsiriX, ImageJ, and sliceOmatic (ICC ranging from 0.979 to 1.000, P < 0.001). Crucially, all software programs showed excellent agreement in distinguishing between the presence or absence of sarcopenia (κ = 0.88-0.96 for observer one, and κ = 1.00 for all comparisons by observer two) and visceral obesity (all κ = 1.00). Intra-observer agreement was exceptionally high (ICC 0.999-1.000, P < 0.001), indicating consistent measurements by the same observer using different software. Similarly, inter-observer agreement was also excellent (ICC 0.998-0.999, P < 0.001), demonstrating consistency between different observers using the same software. These robust agreements were further supported by excellent Jaccard similarity coefficients (mean ≥ 0.964) across all software comparisons, confirming the substantial overlap in measurements.
In conclusion, this Comparative Software study definitively demonstrates that FatSeg, OsiriX, ImageJ, and sliceOmatic exhibit excellent agreement for measuring CSMA, VAT, and SAT in abdominal CT scans. The study also confirms excellent inter-observer and intra-observer reliability across these platforms. Therefore, the results from studies employing these diverse software programs for body composition analysis can be reliably compared and synthesized. This finding is significant for researchers and clinicians who rely on CT-based body composition measurements, assuring that the choice of software does not significantly impact the comparability and validity of their findings.
