The accuracy of an optimized three-point transformation method and its usefulness for integrating multiple independent coordinate systems has been described. Such integration can be implemented to accomplish complex stereotactic procedures which may require the use of multiple image data sets and combinations of frame-based and frameless stereotactic systems. This report details the application of an optimized transformation for intracranial lesion biopsy and/or resection, radiofrequency pallidotomy for treatment of Parkinson's disease, and fractionated stereotactic radiosurgery in a total of 68 patients. For approach to intracranial lesions, a noninvasive definition of image coordinate systems with multiple radiodense scalp markers was cross-registered with a standard stereotactic guidance system. This method allowed for elective acquisition of stereotactic image sets without requiring head frame fixation until the time of the operative procedure. In planning pallidotomy procedures, spatial cross-registration of CT and MR image coordinates were performed to target the posteroventral pallidum. CT coordinates were defined by the standard picket fence algorithm while MR images were referenced with multiple scalp markers. The addition of MR data sets improved anatomic resolution in the regions of the basal ganglia and commissures. Fractionated radiosurgery was accomplished by cross-registration of CT, MR, and plain radiographs using BRW localizers coupled with multiple scalp markers. A daily check of target positioning was performed with the BRW angiograph localizer. The average calculated error was 2.83 mm with a standard deviation of 1.66 mm which remained within the average scan slice thickness of 3.63 mm. In all cases surgical targets were reached without complication.(ABSTRACT TRUNCATED AT 250 WORDS)