CT metal artifact reduction algorithms: Toward a framework for objective performance assessment.

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• Commercial metal artifact reduction (MAR) algorithms for CT lack standardized performance assessment.
• This deficiency poses challenges for regulators, consumers, and industry.
• Objective evaluation is needed to understand MAR's impact on image quality and diagnostic tasks.

Purpose of the Study:

• To develop and test a phantom-based framework for assessing MAR performance.
• To evaluate how MAR affects model observer performance in low-contrast detectability (LCD) tasks.
• To establish a reproducible method for comparing different MAR algorithms.

Main Methods:

• A numerical head phantom with metal implants and a rotatable inset was designed.
• Simulated projection data were generated and processed using two sinogram inpainting MAR variants.
• A channelized Hotelling observer (CHO) assessed LCD task performance using the area under the ROC curve (AUC).

Main Results:

• The framework successfully differentiated the impact of two MAR algorithm variants on LCD task performance.
• It also distinguished performance differences between MAR-applied and non-applied scans.
• Simulation testing validated the framework's ability to resolve performance variations.

Conclusions:

• A phantom-based framework for objective MAR assessment was established and tested.
• MAR does not universally improve image quality for all diagnostic tasks; parameter testing is vital.
• This framework is a foundational step toward comprehensive MAR evaluation methods.