The effect of CT scanning and metal artefact reduction strategies on measuring knee implant displacement: iMAR – helpful or harmful?

Background: A recent development in diagnosing total knee arthroplasty (TKA) tibial component loosening uses Computed Tomography (CT) images of loaded knees to assess implant micro-motion, but metal artefacts can affect reliability. This study investigates the impact of increased exposure and tube voltage, iterative metal artefact reduction reconstruction (iMAR), Sn-filtering and an extended CT scale (ECTS), compared to a standard knee CT protocol on load-induced implant displacement measurement. Method: CT scans of a fresh frozen cadaver leg with a TKA were acquired using a standard protocol and four metal artefact reduction strategies: Increased exposure/tube voltage, iMAR, Sn-filtering, and ECTS reconstruction. Each scan was repeated 10 times with random leg repositioning ensuring no implant-bone motion. Image analysis included implant and proximal tibial cortex segmentation, object registration across scans, and quantification of apparent relative displacement. Grey-level similarity was assessed using the correlation coefficient (CC). Implant displacement error was measured using translation and rotation norms, mean target registration error (mTRE), and maximum total point motion (MTPM). Results: CT scans with metal artefact reduction showed less noise and streaking. Grey-level similarity of the implant was similar across protocols (CC = 0.88). iMAR had the highest bone similarity (CC = 0.99). Implant displacement errors were comparable for increased exposure/tube voltage, Sn-filtering and ECTS (median: 0.08 mm translation, 0.13° rotation, 0.09 mm mTRE, 0.14 mm MTPM). However, iMAR statistically significantly increased median rotation norm to 0.19˚. Conclusion: Increased exposure/tube voltage, Sn-filtering, and ECTS do not impact the measurement error of 3D CT-based knee implant loosening detection. iMAR showed increased rotational error while other displacement parameters remained unchanged, suggesting it may be less suitable for quantifying implant displacement.