Kaji Elizabeth S, Grove Austin F, Lehtonen Eva, Mulford Kellen L, Rouzrokh Pouria, Hannon Charles P, Taunton Michael J, Wyles Cody C
Orthopedic Surgery Artificial Intelligence Laboratory, Department of Orthopedic Surgery, Mayo Clinic, Rochester, MN, USA.
Mayo Clinic Department of Orthopedic Surgery, Mayo Clinic, Rochester, MN, USA.
Arthroplast Today. 2024 Nov 12;30:101553. doi: 10.1016/j.artd.2024.101553. eCollection 2024 Dec.
A fully automated artificial intelligence-based tool was developed to detect and quantify femoral component subsidence between serial radiographs. However, it did not account for measurement errors due to leg position differences, such as rotation or flexion, between comparative radiographs. If there are small differences in rotation or flexion of the leg between comparative radiographs, the impact on subsidence measurement is unclear.
Twenty-five primary total hip arthroplasty procedures were performed by 3 fellowship-trained arthroplasty surgeons using a direct anterior approach. A Hana table allowed precise changes in femur position. Final fluoroscopic images were collected with rotational and flexion changes applied to the femur without moving the C-arm. Subsidence values were manually measured and compared across different positions.
Variations in greater trochanter to tip of the stem measurements between the neutral position and rotations were minimal, measuring <1 mm on an absolute scale and <1% on a relative scale. These differences decreased as the femur was rotated from an external rotation of 20° to an internal rotation of 20°. Notable variances exceeding 5 mm were observed in the 10° flexion position compared to neutral.
Minor changes (20° or less) in leg rotation between serial radiographs are unlikely to significantly affect the greater trochanter to tip of the stem measurement, whereas flexion is highly impactful. These findings suggest that the fully automated artificial intelligence-based tool for detecting and quantifying femoral component subsidence is robust against rotational variations but may be susceptible to significant measurement errors if there are considerable changes in leg flexion between comparative radiographs.
开发了一种基于人工智能的全自动工具，用于检测和量化连续X线片之间股骨假体的下沉情况。然而，该工具未考虑由于比较X线片之间腿部位置差异（如旋转或屈曲）导致的测量误差。如果比较X线片之间腿部旋转或屈曲存在微小差异，对下沉测量的影响尚不清楚。
3名接受过专科培训的关节置换外科医生采用直接前路入路进行了25例初次全髋关节置换手术。使用哈娜手术台可精确改变股骨位置。在不移动C形臂的情况下，对股骨施加旋转和屈曲变化后采集最终的透视图像。手动测量不同位置的下沉值并进行比较。
中立位与旋转位之间大转子至假体柄尖端的测量值变化极小，绝对尺度上测量值<1mm，相对尺度上<1%。随着股骨从外旋20°旋转至内旋20°，这些差异减小。与中立位相比，在10°屈曲位观察到超过5mm的显著差异。
连续X线片之间腿部旋转的微小变化（20°或更小）不太可能显著影响大转子至假体柄尖端的测量，而屈曲的影响很大。这些发现表明，用于检测和量化股骨假体下沉的基于人工智能的全自动工具对旋转变化具有较强的鲁棒性，但如果比较X线片之间腿部屈曲有相当大的变化，可能会产生显著的测量误差。
