Exemplar-Based Inpainting Algorithm for Metal Artifact Areas in Brain Magnetic Resonance Images

Metal artifacts in brain magnetic resonance imaging (MRI) severely distort the magnetic field, degrading image quality. It arises from differences in magnetic susceptibility between metallic implants and surrounding tissues, distorting the magnetic field. These artifacts compromise diagnostic accuracy for conditions such as cerebrovascular diseases, tumors, and strokes, posing challenges for follow-up studies and interventional procedures. Traditional approaches to reduce artifacts, such as using low-field MRI, lead to lower signal-tonoise ratios, prompting the need for advanced software-based solutions. Addressing this challenge, we propose an exemplar-based inpainting algorithm to restore brain magnetic resonance (MR) images affected by metal artifacts. Using T1-weighted brain images from the Alzheimer’s Disease Neuroimaging Initiative, we modeled scenarios with one and three metal components. We compared our method against partial differential equation and coherence transport algorithms using structural similarity index measure (SSIM) and natural image quality evaluator (NIQE) metrics. The exemplar-based algorithm outperformed the alternatives, achieving SSIM and NIQE values of 0.980 and 4.39, respectively, demonstrating improvements of up to 4.71 % in SSIM and 14.26 % in NIQE over conventional methods. These results highlight the potential of our approach to enhance artifact reduction in MR imaging, providing a robust solution for clinical applications.