ABSTRACT
When not corrected, geometrically distorted magnetic resonance images may be unsuitable for stereotactic intracranial neurosurgical procedures where accuracy of target localization is critical. On a GE Signa 1.5-tesla system, we implemented an imaging protocol designed to improve the accuracy of the determination of target coordinates by means of multiple scans utilizing reversal of the frequency-encoded readout gradient. Using a Cosman-Roberts-Wells (CRW) frame and a phantom, geometric shifts of important image features were found to occur. In patients undergoing functional neurosurgical procedures with the CRW system, localization of the posterior commissure by corrected MR images was compared to that obtained by intraoperative ventriculography. Unexpectedly, severe distortions in MR images were revealed by the studies, with shifts of some fiducial markers of 10 mm from their estimated true position. Most of this distortion was attributable to the magnetic properties of the stereotactic frame, and could be eliminated by appropriate design and manufacture of the frame system. Images obtained with an MRI-compatible Leksell stereotactic frame were found to be relatively free of major geometric distortion. This study points out that properties of frame systems used for stereotactic neurosurgery may greatly influence the accuracy of frame-based stereotactic neurosurgery, and that the accuracy of these frame systems is testable.
