Abstract
Introduction
Subtle variations in hip morphology associate with risk of hip osteoarthritis (OA). However, validated accurate methods to quantitate hip morphology using plain radiography are lacking. We have developed a Matlab-based software-tool (SHIPs) that measures 19 OA-associated morphological-parameters of the hip using a PACS pelvic radiograph. In this study we evaluated the accuracy and repeatability of the method.
Methods
Software accuracy was assessed by firstly measuring the linear ratio of 2 fixed distances and several angles against a gold-standard test radiograph, and secondly by repeated measurements on a simulated AP radiograph of the pelvis (reformatted from CT-data) that was digitally rotated about 3-axes to determine the error associated with pelvic mal-positioning. Repeatability was assessed using 30-AP Pelvic radiographs analysed twice (intra-observer), by 2 readers (inter-observer), and finally, using 2 pelvic radiographs taken in 23 subjects (n=46 radiographs) taken same day after re-positioning (short-term clinical-practice variability), and was expressed as coefficient of variation (CV%).
Results
Software accuracy was 0.1% for linear measurements, and 0.2, 0.4, and 0.1 degrees, for angular measurements of 30, 60, and 90 degrees, respectively. Anterior rotation of the pelvis in the sagittal plane beyond 10 degrees produced a decrease in acetabular-tilt (-11 degrees at 20 degrees rotation) and acetabular-index-of-depth-to-width-ratio (-9.3% at 20 degrees rotation). Conversely, femoral-head-to-neck-ratio increased with both anterior and transverse rotation (+9% to +14% at 20 degrees rotation).
The intra-observer CV was between 0.3-6.3%, and inter-observer CV was between 0.7-14.9% for all measurements with the exception of the measurement of horizontal-toit-externa (HTE) that had intra and inter-observer CVs of 33.4 and 29.1%, respectively. Short-term clinical repeatability was between 0.4-8.5%, with the exception of HTE that was 20.7%).
Discussion
This software showed good accuracy and precision for the measurement of OA-associated hip morphological-parameters from plain radiographs of the pelvis, and may be useful in clinical research studies quantitating the relationship of these parameters to the development of hip OA. The method is, however, sensitive to large variations in pelvic positioning and use of the HTE measurement is associated with poor repeatability that is likely due to poor definition of the bony landmarks used for this parameter.
