A Comparative Study of iMAR Preset Parameters for Metal Artifact Suppression in CT Images of Cervical Cancer Brachytherapy

Objective This study investigated the effect of eight preset parameters of the iterative Metal Artifact Reduction (iMAR) algorithm on metal artifact severity (MAS) and low-contrast resolution (LCR) in cervical-cancer brachytherapy CT images and identified the optimal setting.Methods CT scans from 22 patients were retrospectively reconstructed with and without iMAR; MAS and LCR were assessed subjectively and quantitatively, and multi-threshold segmentations of the applicator’s central cross-section in the left–right and superior–inferior directions were measured. The responsiveness of quantitative metrics to subjective ratings was analyzed, and each parameter set was pairwise-compared with the original images.Results Inter-reader agreement for MAS and LCR was 0.85 and 0.84, respectively, and correlations between quantitative metrics and ratings exceeded 0.85. All eight presets significantly reduced MAS relative to the original images (p < 0.01) but affected LCR to varying degrees. The SHI preset performed best, yielding the lowest MAS score (1.45 ± 0.60), the smallest NPSr (0.20 × 107), and a low RMSE (29.95 ± 7.32), followed by the PM and TC presets. Although SPI achieved the weakest artifact reduction, it preserved LCR most effectively. Except for the NC and DF presets, changes in applicator size at thresholds ≥ 1000 HU were within 0.2 mm, though some presets introduced new artifacts.Conclusions NPSr and RMSE reliably quantify MAS and LCR for stainless-steel applicators; the SHI preset is the optimal iMAR setting, whereas SPI may be preferable when rectum and bladder visibility must be prioritized.