A robust index for metal artifact quantification in computed tomography.

Area of Science:

• Medical Imaging Physics
• Radiological Sciences
• Image Analysis

Background:

• Quantitative assessment of metal artifacts in computed tomography (CT) is crucial for evaluating artifact reduction algorithms.
• Existing metrics often lack robustness against varying scan conditions and repeated quality assurance scans.

Purpose of the Study:

• To introduce a new, robust metric for quantifying metal artifacts in CT images.
• To compare the proposed metric against commonly used artifact quantification methods.

Main Methods:

• A novel artifact metric based on the Gumbel distribution's location parameter, normalized for noise independence.
• Comparison with artifact-index, contrast-to-noise ratio, and Gumbel-evaluation method using phantom and clinical CT images.
• Evaluation of metric robustness against varying noise levels and region-of-interest (ROI) selection variations.

Main Results:

• The proposed metric demonstrated superior noise independence and reproducibility with minor ROI variations compared to other methods.
• The new metric showed a coefficient-of-variation of 5.7% (phantoms) and 2.5% (patients) with ROI changes, significantly better than the next best.
• Contrast-to-noise ratio was found inadequate due to poor robustness and lack of correlation with artifact strength.

Conclusions:

• A new metal artifact metric offers enhanced robustness under changing CT scan conditions.
• The proposed metric is less sensitive to user-dependent ROI selection, simplifying implementation for medical imaging system evaluation.