Dental cone beam computed tomography metal artifact reduction based on arbitrary plane metal segmentation

The presence of metallic implants results in severe streak artifacts in dental cone beam computed tomography (CBCT) images. The effectiveness of projection-based metal artifact reduction (MAR) methods depends on precise metal segmentation. Existing metal segmentation approaches struggle with the complex spatial distribution of metals in the oral cavity. This study proposes a dual-threshold metal segmentation method using the convex hull algorithm on an anatomically optimized plane.

We employed multi-plane reconstruction to determine the parameters of chosen slicing directions. The metal segmentation strategy was integrated into an enhanced normalized MAR (NMAR) framework-Arbitrary Plane Segmentation MAR (APSMAR). In addition to qualitative observation, the subjective evaluation was conducted using a 5-point scale by senior dentists. Quantitative performance was assessed by calculating the standard deviation of CBCT values in peri-implant regions.

The proposed method reduced the standard deviation of CBCT values by an average of 66.5%, outperforming reference methods, which achieved a maximum reduction of 59.5%. In subjective evaluations, our method received an average score above 4.3, which is higher than reference methods with scores below 4.1. Effectiveness was indicated in preserving anatomical structures, especially for complex cases.

In this preliminary study, APSMAR significantly outperforms conventional projection-based MAR by leveraging directional optimization for metal segmentation.