摘要
Purpose: We developed an MRI system for three-dimensional planning in radiotherapy. Its contribution on gross tumor volume (GTV) delineation of central nervous system (CNS) diseases was evaluated. Methods and Materials: The MRI system, with corrected distortion, was registered on computed tomography (CT) by means of fiducial/anatomic landmarks. In 41 consecutive patients with various CNS diseases, GTVs determined by MRI/CT registration (MR/CT-GTV) and CT alone (CT-GTV) were compared. Hard copies of diagnostic MRI were shown to doctors when CT-GTV was determined to simulate a conventional planning situation. Multi-observer volumetric analysis was conducted, assessing interobserver deviations among four radiation oncologists and intermethodological deviations between MR/CT-GTV and CT-GTV. Results: Overall, the mean of geometric distortion was significantly reduced from 1.08 mm to 0.3 mm by distortion correction (p < 0.0001). The contribution of the correction was apparent at >12.0 cm radius from the center of the magnetic field. Interobserver deviation was significantly reduced by MR/CT registration (p = 0.005). The improvement was significant for acoustic neurinoma (p = 0.038), astrocytomas (p = 0.043), and lesions at the cerebellum/brainstem (p = 0.008). The regression coefficient between MR/CT-GTV and CT-GTV was <0.9 for cerebellum/brainstem lesions, suggesting that MRI/CT-GTV was smaller than CT-GTV. Conclusions: This system is feasible for three-dimensional planning and was shown to reduce interobserver deviations in GTV delineation for CNS diseases. Copyright © 2001 Elsevier Science Inc.