1. A practical strategy for incorporating the convolution algorithm in Leksell GammaPlan for routine treatment planning.
- Author
-
Yoichi Watanabe, Mathew, Damien, and Natanasabapathi, Gopishankar
- Subjects
- *
COMPUTED tomography , *ALGORITHMS , *RADIOSURGERY - Abstract
Purpose: This study aims to establish criteria for convolution dose calculations and an efficient procedure to include the heterogeneity effects in GammaKnife radiosurgery (GKRS) treatment plans. Methods and materials: We analyzed 114 GKRS cases of various disease types, tumor locations, sizes, the number of fractions, and prescription doses. There was a total of 205 tumors. CT scans were performed in addition to routine MRI scans for all treatments. All treatment plans were created using the TMR10 algorithm (TMR10). We repeated the dose calculations for this study with the convolution algorithm (Conv). We calculated the ratios between Conv and TMR10 of the treatment volume (TxtVol), the volume covered by half of the prescription dose (TxtVol2), the minimum, maximum, and mean doses in the tumor (minDose, maxDose, and meanDose), and the volume of tumor covered by the prescription isodose (covVol). We then categorized those quantities for locations of tumors represented by the shortest distance of the skull surface from the tumor center (distC) and the tumor edge (distE). Results: All six ratios increased with increasing distC and distE. For example, the median minDose ratio increased from 0.885 to 0.933 as distE increased. There was a statistically significant difference in the minDose ratio between tumors of distE < 2 cm and distE ≥ 2 cm. On the other hand, the median maxDose ratio was about 0.933 [0.928-0.939], being almost independent of distE. This suggested a 6.1% overestimation of the delivered dose with TMR10. Conclusions: The heterogeneity effects must be considered for the volume dose calculations by applying the convolution algorithm when the distance of the skull surface from the closest point of the tumor is less than 2 cm to achieve less than 3% accuracy. [ABSTRACT FROM AUTHOR]
- Published
- 2022