Objective: Very few studies have attempted to quantify the true (application) accuracy of image-guidance systems during craniotomy. This is, in part, because of the lack of millimetric intraoperative targets to allow such measurements. Few in vivo studies have compared the influence of mode of patient registration on subsequent true accuracy.
Methods: Seven modes of patient registration (anatomic landmarks, 5 or 10 adhesive fiducials, bone-implanted fiducials [Stryker-Leibinger], surface matching using 45 or 100 points over scalp convexity or nose/auditory meatus contours) were compared. Thirty patients were involved in the study. Millimetric targets (bone drill holes or deep 1-mm titanium hemoclips) were placed then localized and saved at surgery. These targets were then identified on postoperative volumetric computed tomography fused with operative data sets. Localization errors of the targets were measured for each registration on an optical image-guidance system (StealthStation).
Results: Only implanted cranial fiducials had a statistically significant accuracy advantage (1.7 +/- 0.7 mm). All other registrations had similar accuracies (approximately 4.0 +/- 1.7 mm) except anatomic landmarks, which were worse (4.8 +/- 1.9 mm). Calculated accuracies (root mean squared) had no predictive value for true (application) accuracies.
Conclusion: Not surprisingly, application accuracy of image-guidance is worse without implanted cranial markers. Unexpectedly, there was no major difference in localization of deep targets between the other registrations tested in this study. Care must be taken when using image-guidance tools to consider error introduced by registration. Cranium-implanted fiducials should be considered when high accuracy and reproducibility are needed.