Intraobserver errors in obtaining visually selected anatomic landmarks during registration process in nonimage-based navigation-assisted total knee arthroplasty: a cadaveric experiment.

This study investigated the intraobserver errors in obtaining visually selected anatomic landmarks that were used in registration process in a nonimage-based computer-assisted total knee replacement (TKR) system. The landmarks studied were center of distal femur, medial and lateral femoral epicondyle, center of proximal tibia, medial malleolus, and lateral malleolus. Repeated registration in the above sequence was done for 100 times by a single surgeon. The maximum combined errors in the mechanical axis of the lower limb were only 1.32 degrees (varus/valgus) in the coronal plane and 4.17 degrees (flexion/extension) in the sagittal plane. The maximum error in transepicondylar axis was 8.2 degrees. The errors using the visual selection of anatomic landmarks for the registration technique of bony landmarks in nonimage-based navigated TKR did not introduce significant error in the mechanical axis of the lower limb in the coronal plane. However, the error in the transepicondylar axis was significant in the "worst-case scenario."