Assessment of image guided accuracy in a skull model: comparison of frameless stereotaxy techniques vs. frame-based localization.

INTRODUCTION: The use of image-guided systems (IGS) for brain biopsy has increased in neurosurgical practice. We sought to evaluate the accuracy of a plastic, disposable burr hole mounted guide for stereotactic biopsy using an IGS and compare the results of different targeting methods with those of frame based localization.
METHODS: MRIs were performed on a skull model with mounted fiducials with a stereotactic frame in place and data was loaded onto the Stealth IGS. The model was placed in a Mayfield head holder and fixed to the OR table. Registration of imaging to physical space was carried out. Using three different targeting methods on the Stealth IGS, the distance between the target and the predicted position of the target, the offset error, was measured in three dimensions and confirmed by 2 observers. A sum of squares for the 3 offset errors in all planes was used to calculate the summed vector error. The same MRI dataset used with the Cosman-Roberts-Wells (CRW) stereotactic frame for comparison. The summed vector error was calculated in the same manner to compare the accuracy of targeting with these guides to the frame-based CRW system.
RESULTS: For frameless stereotaxy using the "Straight- guide 4 2D" targeting method the mean error was 2.58 +/- 0.51 mm (n=12). The vector error was 5.23 +/- 0.54 (n=4). For the registration set and target using the "Offset- guide 4 2D" targeting method the mean error was 1.66 +/- 0.36 mm (n=12). The vector error was 3.32 +/- 0.72 (n=4). The best localization was obtained with the "probe's eye" planning and targeting. The mean error was 0.33 +/- 0.16 mm (n=12). The vector error was 1.0 +/- 0.28 (n=4). We found a statistical difference between the different techniques (P<0.001) (Kruskal-Wallis One Way Analysis of Variance on Ranks). An all pairwise multiple comparison procedure (Holm-Sidak method) found an overall significance level = 0.05. For the frame-based CRW the mean error from the target was 1.03 +/- 0.19 mm (n=18) and the mean target localization error vector was 2.23 +/- 0.14 (n=6). We found a statistically significant difference between NDT guide "Probes Eye" vs. the MR-CRW (P=0.003, Mann-Whitney Rank Sum Test).
CONCLUSIONS: These results indicate that using MR imaging, surgical planning software and the skull mounted Navigus-DT with the probe's eye view option for targeting, localization accuracy appears to fall within acceptable ranges compared with frame-based methods which have been the standards for stereotactic brain biopsy and functional neurosurgery. Furthermore, there may be considerable differences in accuracy between different targeting methods.