Gamma knife (GK) surgery is an important part of the treatment armamentarium for benign and malignant brain tumors. In general, quantitative volumetrical analysis of the tumor on neuroimaging studies is the most reliable method of assessment of the tumor's response and is critical for accurate dose planning. This study evaluated various factors contributing to volumetric data error of tumors treated with GK radiosurgery. Three differently shaped phantoms (spherical, rectangular, and irregular morphology) were created by immersing like shaped objects into 2% agarose gel. The volumes of phantoms were measured by laser scanning with errors <1%. MRI sequence and parameters including time of flight (TOF), T(1), T(2), different slice thickness, size of field of view (FOV), phase FOV as well as different position and axis of phantoms were retrieved and transferred to a Perfexion Gamma Knife Workstation (PGK-WS) and Picture Archiving and Communication System (PACS) for data analysis. The volumetric data errors were presented as the volume difference between those computed on the PGK-WS and actual volume measured by laser scanning divided by the actual laser scanning volume. The systemic error was defined as volume discrepancy between Perfexion and PACS over that in Perfexion. One-way ANOVA was used for evaluation of data errors between different methods as well as for factor analysis. The MRI-computed volume of the various phantoms approached the laser-scanned volume within 2% when the slice number was >or=30. The volumetrical data errors (10/5 slices) associated with various MRIs for phantoms were 6.94 +/- 0.04%/9.45 +/- 0.35% (spherical phantom), 12.3 +/- 0.2%/ 20.06 +/- 0.7% (rectangular phantom), and 9.29 +/- 0.078%/ 15.67 +/- 0.6% (irregular phantom) (p < 0.001 and p < 0.001), respectively. The system errors (10/5 slices) associated with various MRIs for the phantoms were 3.17 +/- 0.11%/3.9 +/- 0.13% (spherical phantom), 3.61 +/- 0.12%/4.01 +/- 0.12% (rectangular phantom), and 4.39 +/- 0.07%/4.75 +/- 0.13% (irregular phantom) (p < 0.001 and p = 0.01), respectively. The volumetric data errors were related to the number of slices and the shape of phantom, but the systemic errors were only related to the irregularity of phantom morphology. The volumetrical data errors were not related to size of the FOV, phase FOV, sequence of T(1), T(2), TOF, and position of phantom. For the rectangular phantom, the data error was related to slice orientation of imaging acquisition (p < 0.001). Volume discrepancies existed between those volumes computed by the PGK-WS and volumes determined by laser scanning. The volumetric data errors were reduced through the acquisition of more slices through the phantom and a more spherical morphology of the phantom. Relatively few system volume errors were observed between those by the PGK-WS and PACS except for a significant discrepancy for the irregular surface phantom. For the rectangular-shaped phantom, the volumetric data errors were significantly related to slice orientation of measurement. When measuring the tumor response in GK radiosurgery or follow-up, an error of as large as 20% is possible for irregularly shaped object and with MRIs using