Influence of iterative reconstruction and dose levels on metallic artifact reduction: A phantom study within four CT systems

Purpose: To compare metallic artifact reduction (MAR) algorithms proposed by four vendors according to the delivered dose and iterative level using a phantom study. Methods: Four CT systems (Revolution GSI (R), Ingenuity Elite (R), Somatom Edge (R), and Aquilion Prime (R)) equipped with MAR algorithms (Smart MAR (R), O-MAR (R), iMAR (R), and SEMAR (R)) were compared. Acquisitions were performed with CIRS model 062M Phantom containing a titanium rod core insert using 120 kV and two dose levels (3 and 7 mGy). Images were reconstructed with and without MAR algorithms using standard "soft tissue'' kernel for filtered back projection (FBP) and intermediary iterative level. Artifact propagation was assessed by counting the number of pixels containing a HU outside a defined threshold interval (> 100HU and < -80 HU). Artifact correction was evaluated by computing the differences between images with and without MAR. Results: Accuracy of NCT values increased significantly using MAR algorithms, IR, and high dose levels (P < 0.001). Image noise reduced -31 +/- 15 (SD) % (range: -50%;-14%) with Smart-MAR (R), -28 +/- 3 (SD) % (range -31%; -25%) with O-MAR (R), -32 +/- 7 (SD) % (range: -40%; -24%) with iMARCN (R), -52 +/- 8 (SD) % (range: -60%; -42%) with iMARTH (R) and -29 +/- 6 (SD) % (range: -37%; -23%) with SEMAR (R). The number of pixels outside the threshold interval was also reduced using MAR algorithms. Each MAR algorithm corrected in distinct patterns, with satisfactory artifact correction for all MAR algorithms. Conclusion: This study demonstrates that artifact correction using MAR algorithms differs according to the main manufacturers, although corrections are satisfactory for all systems. Corrections also improved by using IR and increasing the dose level. (c) 2019 Societe francaise de radiologie. Published by Elsevier Masson SAS. All rights reserved.