Automated Landmark Selection for Pelvic and Femoral Coordinate Systems

Anatomical coordinate systems for the pelvis and femur were defined according to Wu et al.^{18 (link)} Bony landmarks were selected automatically or semi-automatically using PreView and PostView (Musculoskeletal Research Laboratories, University of Utah, Salt Lake City, UT). Specifically, principal curvature automatically defined the lunate surface of the acetabulum, iliac crest and superior border of the sacroiliac joint on the pelvis and the femoral head, articulating surface of the femoral condyles, and ridges on the medial and lateral femoral epicondyles. The pelvic and femoral joint centers (PJC_CT, FJC_CT) were calculated as the center of the best fit sphere of the lunate surface of the acetabulum and femoral head, respectively (Figure 3). For the medial-lateral axis of the femur and midpoint of the knee, a plane was fit to the medial and lateral epicondyle ridges to isolate the posterior region of the condyles, which was then automatically fit to a cylinder (Figure 4). The center of the cylinder defined the midpoint of the knee.
The posterior superior iliac spine (PSIS) was defined as the posterior intersection of the superior border of the sacroiliac joint and the medial border of the iliac crest (Figure 5). The anterior superior iliac spine (ASIS) was defined as the anterior intersection between the medial and lateral borders of the iliac crest (Figure 5). While each of these of these borders was defined automatically by curvature, there was a small region of nodes at their intersections, of which the user selected a single node to represent the landmark. As the process was not fully automatic, a repeatability study was completed for the ASIS and PSIS. Specifically, three observers selected the landmarks three times to calculate inter- and intra-observer precision following the definition used by Victor et al.^{19 (link)} The average position of each landmark across all nine selection trials was used in subsequent analyses. To evaluate the influence of landmark selection inconstancies, the pelvic coordinate system was calculated using the average landmark positions and the positions from each selection trial. For each axis of the coordinate system, the angle between the average and trial configurations was calculated.