Background/Objectives: Radiostereometric analysis (RSA) is the gold standard for measuring implant migration, with CT-RSA increasingly used as an alternative. To evaluate CT-RSA, it is important to assess data that include the surrounding soft tissues, rather than data from simplified phantoms, while also avoiding unnecessary radiation from multiple scans. This study proposes a method for generating multiple follow-up CTs from a single post-operative CT (baseline CT) by simulating stem migration and uses it to assess an AI-based CT-RSA tool. Methods: The method involves extracting the stem implant voxels from the baseline CT, digitally translating them along the x-, y-, and z-axes, and storing the result as new follow-up CTs. The voxel spacing of the baseline CT is used to define the ground-truth translations, which are then compared with the AI-based CT-RSA results using descriptive statistics and Bland–Altman plots. Results: Using 10 patients’ baseline CTs, 780 follow-up CTs were generated. Bland–Altman analysis showed a mean difference of 0.00 mm, largest LoA −0.10 to 0.09 mm, and translational precision for zero-migration of 0.026 to 0.049 mm. Conclusions: The proposed method offers a practical alternative to phantom-based models, and the AI-based CT-RSA showed high accuracy and precision for stem translation. The study addresses translational migration only.
Nemati et al. (Thu,) studied this question.