Retained bullets in the head on computed tomography - Get the most out of iterative metal artifact reduction

Metal artifacts from retained bullets impair the image quality on computed tomography (CT) and may compromise the detection of critical lesions or the bullet path. To reduce metallic artifacts from medical implants on CT, special algorithms have been developed, e.g., iterative metal artifact reduction (iMAR). The aim of this prospective study was to evaluate the application of iMAR in cases of retained bullets.

In this study, nine different types of projectiles (n = 9) were selected to evaluate the effect of iMAR. The study consisted of two settings. In the first setting, each projectile was fixed on a thin thread and placed in the middle of a water-filled container to demonstrate the effect of iMAR in a homogenous medium. In the second setting, each projectile was placed in the severed head of a pig cadaver as a substitute for human tissue to evaluate iMAR in cases of retained bullets. The raw data from CT scans of both settings were reconstructed with and without iMAR (standard filtered back-projection). The reconstructions with iMAR were calculated using eight different presets provided by the software, namely, neuro-coils, dental fillings, spine implants, shoulder implants, pacemaker, thoracic-coils, hip implants and extremity implants. For each setting, nine reconstructions (n = 9; iMAR: n = 8; without iMAR: n = 1) for each projectile were subjectively evaluated for the image quality and extent of streak artifacts by ten independent and blinded raters (residents: n = 5; radiologists: n = 5). The reconstructions of the second setting were evaluated in a soft tissue window and bone window. A 5-point Likert scale was used for the evaluation of image quality based on the extent of streaks as follows: 1 = severe; 2 = considerable; 3 = moderate; 4 = minimal; and 5 = not apparent. Kendall's W was used for assessing agreement among the ten raters. The Wilcoxon test was used to reveal whether there was a difference in the subjective evaluations between residents and radiologists. Nonparametric Friedman and post hoc tests were used to analyze the Likert scores. The mean difference was considered significant at the 0.05 level.

The agreement among the raters was reasonably high for all projectiles in both settings. In the phantom setting, the iMAR presets neuro-coils and dental fillings yielded the best results. In the pig's head setting regarding the soft tissue window, the presets neuro-coils and extremity implants were preferred. Regarding the bone window, the presets extremity implants and hip implants had the best results. Statistical significance (p < 0.01) between reconstructions without iMAR and neuro-coils or extremity implants was calculated. The iMAR preset spine implants had poor values comparable to reconstructions without iMAR.

The applied iMAR presets revealed different effects on the image quality. Selecting an inappropriate preset (e.g., spine implants) may result in unsatisfactory artifact reduction. The results of this study indicate that the neuro-coils preset is the most appropriate preset for soft tissue, and the preset extremity implants is favorable for bones in cases of retained bullets.