Abstract
Organ dose calculation software offers a convenient way to estimate organ and effective doses in computed tomography (CT) for radiation risk assessment. However, even under identical scan conditions the estimates differ among software packages, and users must understand these differences when applying the results. We quantified the magnitude and causes of inter-software differences and evaluated typical doses for the Japanese population. Three packages-NCICT, WAZA-ARI v2, and VirtualDose CT-were compared under identical conditions (120 kV, 1 s rotation, pitch 1, no tube current modulation) for head, chest, and abdomen-pelvis scans. Tube current was matched to the national median dose indices (Japan DRLs 2025), and the body size to the Japanese adult average (male 170 cm/70 kg; female 160 cm/50 kg). Excluding muscle, organ doses differed among software by up to 28% (male) and 47% (female) relative to WAZA-ARI v2. Muscle was a separate outlier: the muscle dose from the size-adjustable phantom of NCICT was about 90% lower than that from the other two packages. Scanner-dependent differences appeared only in WAZA-ARI v2 (up to 19%), and a ± 2 cm change in scan position markedly altered doses to organs at the scan boundary. Effective doses for the Japanese population were approximately 2.2 mSv (head), 9.6 mSv (chest), and 13.0 mSv (abdomen-pelvis), all within the ranges reported by UNSCEAR 2020/2021. Because all packages have been examined against measurements previously, these differences arise from phantom anatomy and size and from calculation algorithms rather than from inaccuracy; users should interpret results with awareness of this variability.