Accuracy evaluation of surface-based registration methods in a computer navigation system for hip surgery performed...

Many computer navigation systems have recently been developed for brain surgery, and the use of such systems in orthopedic surgery is increasing. Intraoperative registration of preoperative images is one of the most important steps in controlling the overall accuracy of computer navigation systems. Various parameters, such as CT-scan slice thickness, reconstruction pitch, intraoperative data sampling area, and data sampling volume, may affect the accuracy of registration. The purpose of this study was to evaluate the effect of the aforementioned parameters on the accuracy of registration for hip surgery performed through a posterolateral approach, and to find a clinically suitable trade-off between accuracy and surgical invasiveness.

One cadaveric pelvis and one cadaveric femur were used for this study. Four alumina ceramic balls with a diameter of 28 mm and within 1 micrometer of sphericity were attached to the pelvis, and three similar balls attached to the femur, to determine relative position. CT-scan images of the pelvis and femur were obtained with a helical scanner. Three sets of slice thickness and slice pitch were chosen for data acquisition, and two additional sets of reconstructed data were made. Bone contours were extracted by cutting out the surrounding substrate at a given CT number threshold, and surface models of the bone were made from the resultant data. The positions of the pelvis and femur were tracked by LED markers attached to the bone using an optical three-dimensional position sensor (OPTOTRAK). Registration of the computer models to the real objects was performed by measuring the position of a certain number of surface points on each object with an OPTOTRAK pen-probe.

Slice thickness and reconstruction pitch affected the accuracy of registration. As the sampling area was expanded from the periarticular area to the distant peripheral area, accuracy increased slightly. Accuracy did not increase when the whole area was used, but in fact decreased, especially in the femur. The positive effect of increasing the number of sampling points was saturated at 30 points when the surface of the periarticular area was sampled. The following trade-off between accuracy and invasiveness, in terms of various parameters of preoperative and intraoperative data, is proposed as clinically optimal: perform the CT scan with 3-mm slice thickness and 1-mm reconstruction pitch, and sample the periarticular area with 30 sampling points. With these parameters, the accuracy of registration was 1.2 mm and 0.9 degrees of bias with 0.7 mm and 0.3 degrees of RMS in the pelvis, and 1.4 mm and 0.6 degrees of bias with 1.3 mm and 0.3 degrees of RMS in the femur.