VALIDATION OF A NEW METHOD FOR STANDARDIZED EVALUATION OF PELVIC RADIOGRAPHS

Abstract

Introduction: It could be shown that an ample number of classical hip parameters for radiographic quantification of hip morphology on anteroposterior (AP) pelvic radiographs vary significantly with individual pelvic tilt and rotation. This could be proven not only for classical hip parameters (e.g. the lateral centre edge angle) but also for more recently described radiographic features such as acetabular retroversion. The resulting misdiagnosis and misinterpretation can potentially impair a correct therapy for the patient.

We developed fast and easy-to-use computer software to perform three-dimensional (3D) analysis of the individual hip joint morphology using two-dimensional (2D) AP pelvic radiographs. Landmarks extracted from the radiograph were combined with a cone beam x-ray projection model and a strong lateral pelvic radiograph to reconstruct 3D hip joints. Twenty-five parameters including quantification of femoral head coverage can be calculated for a neutral orientation. The aim of the study was to evaluate the validity of this method for tilt and rotation correction of the acetabular rim and associated radiographic parameters.

Methods: The validation comprised three steps:

1. External validation;

2. internal validation; and

3. intra-/interobserver analysis.

A series of x-rays of 30 cadaver pelves mounted on a flexible holding device were available for step 1 and 2. External validation comprised the comparison of radiographical parameters of the cadaver hips when determined with our software in comparison with CT-based measurements or actual radiographs in a neutral pelvic orientation as gold standard. Internal validation evaluated the consistency of the parameters when each single pelvis was calculated back from different random orientations to the same neutral pelvic position. The intra-/interobserver analysis investigated the reliability and reproducibility of all parameters with the help of 100 randomized, blinded AP pelvic radiographs of a consecutive patient series.

Results:

1. All but one parameter (acetabular index) showed a substantial to almost perfect correlation with the CT-measurements.

2. Internal validity was substantial to almost perfect for all parameters.

3. There was a substantial to almost perfect reliability and reproducibility of all parameters except the acetabular index.

Conclusion: The software could be shown to be an accurate, reliable and reproducible method for correction of AP pelvic radiographs. This computer-assisted method allows standardized evaluation of all relevant radiographic parameters for detection of anatomic morphologic differences. It will be used to study the influence of pelvic malorientation on the radiographic appearance of each individual parameter. In addition, it allows evaluating the clinical significance of standardizing pelvic parameters.