Effectiveness of Artificial Intelligence-Based Metal Artifact Reduction Algorithms in Enhancing Diagnostic Accuracy of Cone-Beam Computed Tomography: A Systematic Review

This article aims to systematically assess the effectiveness of artificial intelligence (AI)-based metal artifact reduction (MAR) algorithms in enhancing image quality and diagnostic reliability of dental cone-beam computed tomography (CBCT) scans affected by metal-induced artifacts. Seven studies published between 2015 and 2025 applying AI-driven MAR techniques to CBCT scans with metallic restorations, implants, or orthodontic appliances were reviewed. Study designs included experimental, retrospective, and phantom-based investigations. Deep learning architectures such as U-Net variants, dual-encoder attention networks, and diffusion models were evaluated. Owing to methodological heterogeneity, findings were synthesized narratively with an emphasis on artifact suppression and quantitative image quality outcomes. All included studies reported meaningful reductions in metal-related artifacts using AI-based MAR algorithms. Improvements were consistently observed across key image quality metrics, including artifact index, peak signal-to-noise ratio, structural similarity index, and contrast-to-noise ratio. Importantly, these algorithms preserved essential anatomic structures and enhanced the interpretability of reconstructed images.