Clinical evaluation of a new adaptive iterative metal artifact reduction method in whole-body low-dose CT skeletal survey examinations

Objective: To evaluate a new adaptive iterative metal artifact reduction algorithm (AiMAR) in whole-body low-dose CT (WBLDCT) skeletal survey examinations. Methods: Projection data were retrospectively obtained from 25 clinical WBLDCT skeletal survey patients, each with two types of metal implants. Images were reconstructed with bone and soft tissue kernels using four settings—original and AiMAR with strengths of 2, 4, and 5. All images were anonymized and randomized for a reader study, where three musculoskeletal radiologists independently determined the overall ranking of all series based on diagnostic quality, and local scoring of metal artifact and anatomy visualization for each implant. Quantitative image noise analysis was performed in areas close to the implants. Intraclass correlation coefficients (ICC) and Krippendorff’s alpha were computed for inter-rater reliability. Results: AiMAR 4 was ranked the highest for 64.3% of the series across eight types of implants. For local scoring task, AiMAR 4 showed better metal artifact and anatomy visualization than the original and AiMAR 2. AiMAR 4 was comparable in anatomy visualization but inferior to AiMAR 5 in metal artifact scores. AiMAR 4 led to 56.3% noise reduction around the implant areas compared with the original images, and AiMAR 5 68.1% but also resulted in anatomy blurring in 40% of the implants. ICC and Krippendorff’s alpha revealed at least substantial reliability in the local scores among the readers. Conclusions: AiMAR was evaluated in WBLDCT skeletal surveys. AiMAR 4 demonstrated the highest overall quality ranking and improved local scores with noise reduction around implant areas.