Background: Combining electromagnetic tracking of instruments with preoperatively acquired images can provide detailed visualization for intraoperative guidance and reduce the need for fluoroscopy and contrast. In this study, we investigated the accuracy of a vessel-based registration method designed for matching preoperative image and electromagnetically tracked positions for endovascular therapy.
Material and methods: An open-source registration method was used to match the centerline extracted from computed tomography (CT) to electromagnetically tracked positions within a vascular phantom representing the abdominal aorta with bifurcations. The target registration error (TRE) was calculated for 11 fiducials distributed over the phantom. Median and intra-quartile range (IQR) for 30 registrations was reported. TRE < 5 mm was claimed sufficient for endovascular navigation, evaluated using the Wilcoxon signed-rank test. TRE was also compared to a 3D-3D registration method based on intraoperative cone-beam CT, using the Mann-Whitney U-test.
Results: The TRE was 3.75 (IQR: 3.48-3.99) mm for the centerline registration algorithm and 3.21 (IQR: 1.50-3.57) mm for the 3D-3D method (p < .001). For both methods, the TRE was significantly < 5 mm (p < .001).
Conclusion: The centerline registration method was feasible, with an accuracy sufficient for navigation in endovascular therapy. The centerline method avoids additional image acquisition for registration purpose only.
Keywords: Registration; centerline; image fusion; navigation.