Evaluation of virtual monochromatic imaging combined with metal artifact reduction in low-dose dual-energy computed tomography: a phantom study
Year: 2026
Type: article
Abstract: Background: Metal implants can cause substantial artifacts on computed tomography (CT), and the optimal virtual monochromatic (VM) energy for metal artifact reduction (MAR) remains uncertain. This study assessed metal-artifact reduction and overall image quality on low-dose dual-energy computed tomography (DECT) VM images reconstructed with Smart MAR. Methods: A multi-energy CT phantom (GPH-75B; Kyoto Kagaku Co., Ltd., Kyoto, Japan) was scanned in DECT mode with a 12-mm titanium rod positioned at the center, right, and superior portions. VM images from 40 to 140 keV were reconstructed with MAR and without MAR (no-MAR). Image quality was assessed according to artifact and structure visibility scores, CT values, standard deviations (SDs), titanium-rod dimensions, structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and root mean square error (RMSE). MAR and no-MAR measurements were compared via two-sided paired t-tests. Results: Across the three titanium-rod positions, MAR generally reduced artifacts, improved titanium-rod edge visibility, and reduced geometric distortion compared with no-MAR. At every rod position and energy level, MAR yielded higher SSIM and PSNR and lower RMSE compared to no-MAR (all P values <0.05). At the central position, the transverse-to-longitudinal diameter ratio was 0.94-1.03 with MAR and 0.21-0.81 with no-MAR (all P values <0.001), remaining close to 1.0 at energies ≥60 keV with MAR. For MAR images, adjacent-structure clarity was highest at 60 keV. In both groups, SSIM and PSNR increased with increasing energy, whereas RMSE and image noise decreased. Conclusions: MAR consistently improved image quality across titanium-rod positions. In terms of artifact reduction, implant depiction, and adjacent-structure visibility, 60 to 80-keV VM images with MAR provided the most favorable overall performance under the tested low-dose conditions.
Language: English
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7223196835