A comparison of computer-aided surgery registration methods for endoscopic sinus surgery

Stereotactic computer-aided surgery has provided the surgeon with a means to navigate more safely through diseased or surgically altered sinus anatomy. Accurate registration is vital to successful image-guided surgery. This study compared the accuracy and performance of three registration methods: fiducial, anatomic landmarks, and surface registration. Ten fixed cadaveric heads underwent endoscopic computed tomography scan followed by middle meatal antrostomy and sphenoidotomy. Each registration method was performed, and the time required and mean registration error were recorded. Five anatomic sites were then identified and compared with the preoperative computed tomography images. The true distances between the known anatomic sites and the crosshair locations on the images were measured. Statistically significant differences were noted for mean registration error and time for registration. The mean +/-SEM time for registration for the fiducial, surface, and landmark methods were 5 minutes 24 seconds +/-27 seconds, 1 minute 1 second +/-5 seconds, and 11 minutes 46 seconds +/-45 seconds, respectively. The mean +/-SEM registration error for the fiducial, surface, and landmark methods were 0.48 +/- 0.21 mm, 1.05 +/- 0.06 mm, and 3.1 +/- 0.25 mm, respectively. When the true accuracy of the three registration methods were compared, no significant difference was found between fiducial and surface registration. However, fiducial registration was significantly more accurate than landmark registration at all points. When compared with landmark registration, surface registration was statistically more accurate at all anatomic sites except for the sella turcica and optic nerve. When the true accuracies of these methods were compared in fixed cadaveric specimens,fiducial and surface registration were statistically similar but were found to be significantly more accurate than landmark registration. Furthermore, when time of registration, accuracy, and ease of use were considered, surface registration was found superior.