Evaluation of migration analysis with AI-based CT-RSA and preoperative 3D-planning in total hip arthroplasty.
Authors
Affiliations (4)
Affiliations (4)
- Department of Orthopedics, Skåne University Hospital, Clinical Sciences, Lund University, Lund, Sweden. [email protected].
- Ortoma AB, Gothenburg, Sweden.
- Radiation Physics, Department of Hematology, Oncology and Radiation Physics, Skåne University Hospital, Lund, Sweden.
- Department of Orthopedics, Skåne University Hospital, Clinical Sciences, Lund University, Lund, Sweden.
Abstract
Computed tomography (CT) has become a valuable tool for preoperative planning and perioperative, real-time navigation during total hip arthroplasty (THA). CT can also quantify postoperative implant migration without the need for implanted bone markers, making it a promising alternative to the current gold standard radiostereometric analysis (RSA). Our aim was to evaluate the accuracy of preoperative planning and postoperative implant migration of both cup and stem employing AI-based software using 3D CT-images (CT-RSA) compared with conventional RSA. 26 patients with primary THA were preoperatively 3D-planned and perioperatively navigated. They were followed and analyzed with AI-based CT-RSA within 2 days postoperatively and at 3, 12, and 24 months. 10 of the patients had implanted tantalum markers at surgery and were also followed up with conventional model-based RSA (MBRSA). The results were compared with CT-RSA. Prosthetic CAD models were used for both conventional RSA and AI-based CT-RSA analysis. Double CT and MBRSA scans were taken to evaluate precision. The preoperative plan was compared with actual perioperatively chosen implants. AI-based CT-RSA showed consistent migration patterns, with most migration in the first 3 months, which then levelled out. Bland-Altman plots indicated good agreement between MBRSA and AI-based CT-RSA. Overall, there was high correspondence between MBRSA and AI-based CT-RSA in translations, but more divergent rotation results. AI-based CT-RSA precision was consistently slightly better than MBRSA precision. The agreement between planned and actual size of cup was 25 out of 26, and 23 out of 26 for stems. AI-based CT-RSA demonstrated accuracy comparable to MBRSA, with slightly improved precision and reduced user-dependence. The same system also provided an accurate and predictable preoperative implant plan.