Usefulness of metal artifact reduction on CT angiography after massive coil embolization in peripheral AVM

Purpose To evaluate the image quality of three reconstruction methods—filtered back projection (FBP), adaptive statistical iterative reconstruction (AR50), and deep learning–based reconstruction (DL-M)—processed with the Smart Metal Artifact Reduction (SMAR) technique for post-embolization assessment of peripheral arteriovenous malformations (AVMs). Materials and methods In this prospective single-center study, 30 patients who underwent coil embolization for AVM were included. Post-embolization CT angiography was performed using dual-energy CT. Virtual monoenergetic images at 50 and 70 keV were reconstructed using FBP, AR50, and DL-M. All were processed with SMAR, and DL-M without SMAR served as the baseline. Artifact severity was objectively assessed using the standard deviation (SD) around the AVM, artifact index (AI), and contrast-to-noise ratio (CNR). Two readers (a resident and a staff radiologist) subjectively graded artifact severity, vessel visualization, and new artifacts using 4-point scales. Results At both energy levels, average SD and AI were significantly lower in SMAR-processed images than in baseline DL-M (all p < 0.001). Subjective scores for artifact reduction and visualization of adjacent vessels were also significantly improved (p < 0.001). There were no significant differences among the three SMAR-processed methods. New artifacts appeared in three cases but had minimal effect on interpretability. Conclusions SMAR-processed FBP, AR50, and DL-M reconstructions significantly reduced metal artifacts and improved visualization after AVM coil embolization, supporting their value for post-treatment evaluation and clinical decision-making.