Neurosurgeons are often faced with a lesion that is obvious on magnetic resonance imaging (MRI), but not well-defined on computerized tomography (CT). However, questions remain regarding the distortion inherent in MRI images. Therefore the following comparative study was conducted. Five patients with intracranial lesions (1 lymphoma, 1 multiple sclerosis, 2 glioblastomas, and 1 AVM), underwent both CT and MRI for purpose of stereotactic biopsy or radiosurgery. The Brown-Roberts-Wells CT and MRI compatible localizing rings were used. Coordinates of the left optic nerve-globe junction, the aqueduct, pineal and optic chiasm were recorded from both CT and MRI. With MRI, all three imaging planes, axial, coronal and sagittal, were used. Coordinates were calculated in millimeters and submitted to statistical analysis using Pierson correlation coefficients. In all, there were 17 CT, 17 MRI axial, 13 MRI coronal, and 13 MRI sagittal coordinates. The analysis revealed that of the MRI coordinates, the axial coordinates were more available and retrievable. Lateral targets, such as the left optic nerve-globe junction, were more difficult to identify on coronal and sagittal images and often lacked a full compliment of fiducial points for calculation. The correlation with CT was best for axial, followed by coronal and then sagittal MRI planes. Geometric image distortion occurs owing to nonlinearity of the magnetic field as well as magnet susceptibility to different tissues at an interface. We feel that MRI stereotaxy can be utilized in conjunction with CT for verification of lesion coordinates in relatively large lesions, and particularly in those not well delineated on CT.