3-D landmarking repeatability emphasizes challenges in scan positioning during weight bearing CT of the knee.

OBJECTIVE: Maintaining repeatability of positioning during knee weight bearing CT (WBCT) acquisition can be challenging. We evaluated a manual bone landmarking system and automatically derived angulation metrics to advance understanding of joint positioning repeatability. DESIGN: Bilateral knee WBCTs for 14 recruited individuals taken at a mean ± SD interval of 14.9 ± 8.1 days using a frame and laser guidance for positioning were used in this technical study. One observer manually placed 33 bony landmarks at each knee, verified by a second observer who also placed landmarks at each end of the scan range in the centre of the bone. Landmark repeatability error was the mean ± SD absolute distance (mm) between baseline and follow-up. Automated methods for knee flexion angle (KFA), valgus and vertical positioning alignment were developed from the landmarks. RESULTS: Most repeatable landmarks were around the patella and proximal fibula with best performance at the proximal fibula tip (1.3 ± 0.5 mm). Least repeatable landmarks were along shallow curves, e.g. central tibial margins and around the trochlea, with worst performance at the tibial tuberosity (5.5 ± 4.7 mm). Mean repeatability for 85% of placed landmarks was in the 1-4 mm range. Valgus alignment absolute repeatability error ranged from 0.1-2.1° and for KFA was 0.3-15.5°, neither dependent on KL grade severity. Worst vertical scan range placement error was ∼60 mm. CONCLUSION: Despite guidance systems, repeatability of KFA and vertical positioning could be optimised at the time of acquisition. This would provide a robust platform for evaluation of metrics such as meniscal extrusion and 3-D joint space width from WBCT that have shown promise in the evaluation of knee OA.