Prediction of volumetric data errors in patients treated with gamma knife radiosurgery

Purpose: Volumetry is the most commonly used method to measure tumor response in patients who receive Gamma Knife radiosurgery. We calculated the data errors in measurement made by different methods based on the stereotactic fiducials (Gamma Knife workstation), surface area multiplied by thickness (PACS), and product of maximum diameter in x, y, and z dimensions (geometric method) to more precisely evaluate tumor response in patients treated with Gamma Knife radiosurgery. Materials and Methods: From 2003 to 2006, 210 tumors were enrolled in this study. MRIs obtained from these patients were transferred to Gamma Knife and PACS workstations. Data errors were defined as the difference between the volume calculated by various methods and the Gamma Knife workstation divided by the volume obtained from the Gamma Knife workstation. Linear regression was used for data analysis. Results: There were 27 lesions with volume < 0.5 cm(3), 97 lesions between 0.5 and 4 cm(3), 68 lesions between 4 and 14 cm(3), and 18 lesions larger than 14 cm(3). A strong linear correlation was found between the volume measurement by Gamma Knife workstation and PACS and the geometric method (r = 0.993, 0.967, respectively). Poor correlation between the Gamma Knife and PACS workstation volume measurement was observed in tumors less than 0.5 cm(3) (r = 0.763), but better correlation was found between the Gamma Knife workstation and geometric method (r = 0.871). Fewer data errors were observed in the PACS than in the geometric method (p <0.001) in tumors with volumes of 0.5-4 cm(3) and 4-14 cm(3); whereas in tumors less than 0.5 cm 3, fewer data errors were observed in the geometric method (p=0.01). The data error in the whole series was 6 +/- 15% in the PACS system and was relatively correlated with the volume (p = 0.03) and the number of slices (p = 0.021), but not with the Pearsonian coefficient of skewness (p = 0.81). Conclusion: The different methods of measurement of tumor volume (> 0.5 cm(3)) demonstrated strong linear correlation. In tumors with volume less than 0.5 cm 3, the most reliable method was the geometric method. When using the PACS system in the evaluation of tumor response, a data error as high as 21% should be considered. Copyright (c) 2007 S. Karger AG, Basel