Lower extremity alignment remains one essential objective during total knee replacement. Implants positioning analysis requires selecting reliable skeletal landmarks. Our objective was to in vivo evaluate the precision of the implemented skeletal landmarks. This evaluation was based on multiple three-dimensional (3D) computer reconstructions of the lower extremity derived from an EOS^® biplanar low-dose X-ray system acquisition. A 3D angle measurement protocol was used.

Currently defined landmarks carry a tolerable uncertainty margin, which can still probably be further improved.

Nine lower extremity 3D computer reconstructions were obtained from an EOS protocol based on seven simultaneous A–P and lateral views performed in standing position. A database was established by four operators; finally, building up a total of 99 in vivo 3D reconstructions of these nine lower extremities. Specific algorithms were used for such 3D reconstructions of lower extremities based on bone points and pre-identified contours on X-ray. Four femoral landmarks and four tibial landmarks were thus defined. For each bone and each landmark studied, a mean landmark for the 11 consecutive series elements was established. The deviation from each constructed landmark to the corresponding mean landmark was calculated based on the anteroposterior (x), longitudinal (y) and mediolateral axes (z), in translation (Tx, Ty, Tz) and in rotation (Rx, Ry, Rz). Uncertainty was estimated by the 95% confidence interval (95% CI).

The landmarks located at the middle of the segment joining the center of each posterior condyle and at the barycenter of the plateaux showed a greater reliability; these landmarks uncertainty (95% CI) of Tx, Ty, Tz was less than 1, 0.5, 1.5 mm for the femur and 1.5, 0.6, 0.6 mm for the tibia, respectively. The femoral landmarks using the center or posterior edge of the posterior condyles to define the mediolateral axis were retained; for rotations Rx, Ry, and Rz, uncertainty remained less than 0.3, 4, and 0.5°. All of the tibial landmarks had a comparable reliability in rotation, 95% of the Rx and Rz deviations were under 0.5 and 1.3°, respectively, with a mean error less than 1°. For the tibial rotation Ry, the mean error was greater (4°), with uncertainty (95% CI) at 11.2°. All tibial translations showed a mean error of 1 mm. The 3D implantation angles were measured on two patients using preoperative 3D skeletal reconstructions and 3D geometric models of the implants repositioned on postoperative EOS^® knee X-rays.

The posterior condyles are rarely involved in the arthritic wear process, making them an anatomic landmark of choice in the analysis of the femoral component positioning. The femoral landmarks using the posterior condyles were sufficiently reliable for clinical use. However, the posterior contours of the tibial plateaux were less precise. The knees should be staggered from an anteroposterior perspective on the EOS^® lateral images so that they can be visualized separately. The anatomic zones on which the skeletal landmarks are based are usually removed by the bone cuts, making it preferable to save the preoperative computer reconstructions to analyze the postimplantation 3D reconstruction.

The lower extremity skeletal landmarks precision relates to the quality of the corresponding 3D reconstructions. Except for tibial rotation, all the translation and rotation parameters were estimated within a mean error margin inferior to 1.2 mm and 1.3°, respectively. Making the reconstruction algorithms more robust would render certain anatomic zones even more precise. Biplanar low-dose EOS^® X-ray system is a tool of the future to generate 3D knee X-rays that can improve the evaluation and follow-up of total knee arthroplasty patients.

Level of evidence: level IV, diagnostic study.