Definition of the prostate in CT and MRI: A multi-observer study

Purpose: To determine, in three-dimensions, the difference between prostate delineation in magnetic resonance (MR) and computer tomography (CT) images for radiotherapy treatment planning. Patients and Methods: Three radiation oncologists, considered experts in the field, outlined the prostate without seminal vesicles both on CT, and axial, coronal, and sagittal MR images for 18 patients. To compare the resulting delineated prostates, the CT and MR scans were matched in three-dimensions using chamfer matching on bony structures. The volumes were measured and the interscan and interobserver variation was determined. The spatial difference between delineation in CT and MR (interscan variation) as well as the interobserver variation mere quantified and mapped three-dimensionally (3D) using polar coordinates. A urethrogram was performed and the location of the tip of the dye column was compared with the apex delineated in CT and MR images, Results: Interscan variation: CT volumes were larger than the axial MR volumes in 52 of 54 delineations. The average ratio between the CT and MR volumes was 1.4 (standard error of mean, SE: 0.04) which was significantly different from 1 (p < 0.005), Only small differences were observed between the volumes outlined in the various MR scans, although the coronal MR volumes were smallest. The CT derived prostate was 8 mm (standard deviation, SD: 6 mm) larger at the base of the seminal vesicles and 6 mm (SD 4 mm) larger at the apex of the prostate than the axial MRI. Similar figures were obtained for the CT and the other MRI scans. Interobserver variation: The average ratio between the volume derived by one observer for a particular scan and patient and the average volume was 0.95, 0.97, and 1.08 (SE 0.01) for the three observers, respectively. The 3D pattern of the overall observer variation (1 SD) for CT and axial MRI was similar and equal to 3.5 to 2.8 mm at the base of the seminal vesicles and 3 mm at the apex. Conclusion: CT-derived prostate volumes are larger than MR derived volumes, especially toward the seminal vesicles and the apex of the prostate, This interscan variation was found to be larger than the interobserver variation. Using MRI for delineation of the prostate reduces the amount of irradiated rectal wall, and could reduce rectal and urological complications. (C) 1998 Elsevier Science Inc.