Automatic Characterization of Pelvic and Sacral Measures from 200 Subjects.

BACKGROUND: An understanding of pelvic and acetabular morphology and orientation is required for accurate surgical reconstruction of the hip and spine, as well for component placement in a total hip arthroplasty. Our objectives were to develop an automated system for measuring pelvic and sacral orientations utilizing computed tomographic (CT) scans and to characterize these measures across 200 asymptomatic subjects. METHODS: An automated feature recognition algorithm was created to identify acetabular and pelvic orientation across 200 scans generated for non-musculoskeletal conditions. Three-dimensional models were generated from CT data to serve as inputs to the algorithm. Acetabular orientation was defined by comparing a plane fit to the acetabular rim with the anterior pelvic plane. Pelvic inclination, pelvic tilt, and sacral slope were defined as the angles between landmarks identified across the pelvis: pubic tubercles, acetabular center, left and right anterior superior iliac spines, and sacral plate. RESULTS: The mean sacral slope was 36.49°, the mean pelvic tilt was 15.60°, and the mean pelvic incidence was 52.05°. The mean sacropubic angle was 32.48° and the mean pelvic-Lewinnek angle was 8.93°. Significant differences between male and female subjects were observed in the sacral slope (mean difference, 4.72°; p < 0.05), pelvic tilt α (mean difference, 4.17°; p < 0.05), pelvic tilt γ (mean difference, 3.06°; p < 0.05), and the pelvic-Lewinnek angle (mean difference, 1.76°; p < 0.05). The comparison of acetabular orientation measures with those in a prior study of the same cohort yielded intraclass correlation coefficients (ICCs) all above 0.97. The validation of sacral orientation via manual measurement also yielded ICC values all at or above 0.97. CONCLUSIONS: Our algorithm showed a high degree of consistency in acetabular orientation measures with respect to a prior study of the same cohort. The measures of pelvic orientation were found to be accurate and reliable when compared with manual measurements of the same data set. All measurements of pelvic orientation were consistent with the means reported in the literature. CLINICAL RELEVANCE: An accurate and reproducible, automated technique for determining pelvic and acetabular orientation provides a way to characterize these measures as an aid in clinical diagnosis and preoperative planning.

Automated femoral version estimation without the distal femur.

Femoral version impacts the long-term functioning of the femoroacetabular joint. Accurate measurements of version are thus required for success in total hip arthroplasties and hip reconstructive surgeries. These are impossible to obtain without visualization of the distal femur, which is often unavailable preoperatively as the majority of imaging scans are isolated to the pelvis and proximal femur. We developed an automated algorithm for identifying the major landmarks of the femur. These landmarks were then used to identify proximal axes and create a statistical shape model of the proximal femur across 144 asymptomatic femora. With six proximal axes selected, and 200 parameters (distances and angles between points) from the shape model measured, the best-fitting linear correlation was found. The difference between true version and version predicted by this model was 0.00 ± 5.13° with a maximum overestimation and underestimation of 11.80 and 15.35°, respectively. The mean absolute difference was 4.14°. This model and its prediction of femoral version are a substantial improvement over pre-operative 2D or intra-operative visual estimation measures. © 2018 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 36:3161-3168, 2018.