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

Abstract

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.

Keywords: Arteriovenous malformation; Artifact; Embolization, therapeutic; Image processing, computer-assisted; Tomography, X-ray computed.