Image-domain metal artifact reduction in low-energy VMI using high-energy regional prior

Abstract

Background

Metal artifacts degrade the clinical utility of virtual monochromatic images (VMIs), particularly in low energy levels. Nevertheless, low-energy VMIs have essential clinical applications, such as reducing the volume of iodinated contrast material administered, salvaging poorly attenuated contrast-enhanced CT studies, and analyzing arterial vasculature during the venous phase. Conventional metal artifact reduction algorithms may introduce new artifacts and obscure soft tissue details.

Purpose

The aim of this study is to develop a practical image-domain solution for significantly reducing the metal artifacts in low-energy VMIs while preserving the clarity of soft tissues and metal boundaries.

Methods

A mapping model was developed to establish a relationship between optimal VMI and the material basis images (MBIs) in artifact-free regions. This model was subsequently used to correct artifact-affected regions in MBIs. Finally, artifact-reduced low-energy VMIs were synthesized from the updated MBIs. The approach, referred to as regional model-based metal artifact reduction (rMAR), utilized the mapping model to effectively reduce metal artifacts. To validate the efficacy of the proposed method, both phantom and patient data acquired from Philips scanner were used. The scanner's built-in metal artifact reduction for orthopedic implants, known as OMAR, was employed. Comprehensive comparisons were conducted among four image processing strategies: VMI alone, VMI combined with OMAR (VMI + OMAR), VMI combined with the proposed rMAR (VMI + rMAR), and a combination of all three methods (VMI + OMAR + rMAR). Evaluations were performed using visual assessment, line profile analysis, and measurement of the ∆CT number.

Results

High-energy VMIs exhibit significantly fewer metal artifacts compared to those at low energy levels, as demonstrated in both phantom and patient results. Although conventional metal artifact reduction algorithms can mitigate the existing artifacts, they often introduce new ones. In contrast, the proposed rMAR method effectively reduces artifacts in low-energy VMIs, achieving improved image quality without introducing new artifacts. In specific cases, such as postoperative VMIs of hip prosthesis implants, the combined VMI + OMAR + rMAR approach demonstrates superior metal artifact reduction compared to either OMAR or rMAR alone. Quantitative line profile analysis indicated that the proposed rMAR method produced images with artifact levels more closely resembling the ground truth than those processed with OMAR. The ΔCT number was significantly lower in the images processed with rMAR than in those processed with OMAR.

Conclusion

The proposed rMAR method effectively achieves metal artifact reduction, particularly in low-energy VMIs, while preserving the clarity of soft tissues and metal boundaries. Consequently, the diagnostic value of low-energy VMIs containing metal implants is enhanced.