Assessment of acetabular cup position with computed tomography

Abstract

Background: Revision of Total Hip Arthroplasty (THA) due to loosening is necessary in many patients. Diagnosis is usually late with significant bone and soft tissue destruction. Available diagnostic techniques are either not sensitive enough or too complicated and expensive for routine use. These studies comprise the first results from a development project aiming towards a new non-invasive method for accurate determination of position and migration of orthopedic implants using Computed Tomography (CT) intended to be applicable in a routine clinical setting.

Specific aim: Volumetric detection of pelvic orientation and acetabular cup position and orientation using CT as input modality.

Method: Some additions to a volume fusion tool were made. Mathematical algorithms were implemented in computer software. An image handling tool for bringing a volume to a standard orientation was built. These tools were validated in three studies of 10 patients with double post-operative CT scans after cemented THA, and one phantom study where CT acquisitions in different positions of a pelvic model with tantalum markers implanted for fusion and an uncemented cup were combined into pairs together with corresponding fused volumes from a previous study. Image analysis were performed by two independent observers. For fusion, manually picked co-homologous point pairs in the pelvis in conjunction with an affine transform were used (Study 1). Implant orientation and position were visually evaluated, and numerically calculated based on manually picked points corresponding to the implant (study I-IV). For direct measurement of cup orientation relative to the pelvis, the bone was reoriented to an anatomical orientation, and this information was used to minimize the effect of patient positioning on measurements of cup orientation (study IV). Parametric statistical methods were used after data had been tested for normality.

Results: Fusion effectively minimized the positioning error of the pelvis on repeated CT examinations to a visual mismatch of the pelvis in patients to less than one cortical width. The visual accuracy for detecting motion of the cup was at least doubled compared to what could be expected from routine radiographs. Using a minimum of landmarks after fusion, the mean error of the angle between cup openings was 2.5° (range 0.3-5.3°, SD 1.5°), of the acetabular opening centerpoint 2.5 min (range 0.4-9.9, SD 2.8), and of the femoral component head was 1.0 min (0.7-1.8, SD 0.3) (study I). Using an over-determined pattern for detection, repeatability of measurements of cup/head center 3D coordinates within the same patient volumes was 2.2/1.7 mm (examiner 1), 2.4/2.2 min (examiner 2), These error zones in 3D localization were smaller than the basic picture element (the voxel) (study II). Orientation of an uncemented cup in a pelvic model could across volumes after fusion be assessed with a 95 % interobserver precision confidence interval of 0.7°. Mean angle error of the cup after fusion (the bias from a 0° golden standard) was less than 0.4° in this model study (study III). Orientation of a cemented cup in patients could be assessed relative to the pelvis in 3D with a 95 % interobserver precision confidence interval of 3° for inclination and 6° for anteversion across patient scans (study IV). In the patient studies, most of the errors occurred during the cup detection step. Image analyses were done within a clinically reasonable time-frame of 1030 minutes. There were no significant differences in results between examiners. The inferior results on cup detection in the patients studies were probably due to the design of the radiolucent thread in the cup.

Conclusions: Volume fusion effectively aids in differentiating true movement of the cup relative to the pelvis from projectional errors. Reorienting the pelvis after the CT-scanning facilitates measurements of cup orientation directly in relation to the bone. These two techniques are therefore less dependent on correct patient positioning than conventional radiographs. In patient studies both visual and numerical results was significantly better than what can be expected from simple measurements on radiographs, but did not match the results of validated methods for migration such as Radiostereometry or EBRA. Precision and accuracy were better than I' in a model study. The present studies cannot establish the system's accuracy and reproducibility in patients, since they were small and performed with a series of limitations, and longitudinal patient studies are lacking. However the present studies shows exciting potential for the monitoring of THA.