Effect of blending ratio on automated body composition analysis in portal venous phase dual-energy CT: agreement with single-energy nonenhanced CT.

PURPOSE: To evaluate the effect of blending ratio on automated body composition analysis (BCA) from contrast-enhanced dual-energy CT (CE-DECT), using single-energy nonenhanced CT (NECT) as the reference, and to develop correction equations to reduce blending-related bias. METHODS: This retrospective study included 737 patients who underwent both NECT and portal venous phase CE-DECT (90/Sn150 kVp). Automated BCA was performed on NECT, 90-kVp source, Sn150-kVp source, and blended images with low-kVp fractions of 81%, 66%, and 54%. Skeletal muscle density (SMD) and skeletal muscle area (SMA) were analyzed as key metrics. CE-DECT-derived measurements were compared with NECT using adjusted analyses and quadratic regression to estimate NECT-equivalent low-kVp fractions, and correction equations were developed with an 80:20 patient-level training/test split. RESULTS: BCA measurements varied by blending ratio. SMD ranged from 37.7 ± 7.7 HU on Sn150-kVp source images to 46.1 ± 8.8 HU on 90-kVp source images, compared with 40.5 ± 7.8 HU on NECT. SMA was higher on all CE-DECT image sets than on NECT. These differences persisted after adjustment. The optimal low-kVp fraction differed between SMD and SMA, measuring 30.0% and 78.1%, respectively. In the held-out test set, comprehensive correction reduced mean bias from 2.4 HU to 0.03 HU for SMD and from 5.0 cm2 to - 0.1 cm2 for SMA. CONCLUSION: Blending ratio influences automated BCA measurements from portal venous phase CE-DECT, with metric-specific effects on SMD and SMA. Protocol-specific correction equations may reduce bias relative to NECT, but require external validation before broader application.