Comparison of gamma index based on dosimetric error and clinically relevant dose-volume index based on three-dimensional dose prediction in breast intensity-modulated radiation therapy

Abstract

Background: Measurement-guided dose reconstruction has lately attracted significant attention because it can predict the delivered patient dose distribution. Although the treatment planning system (TPS) uses sophisticated algorithm to calculate the dose distribution, the calculation accuracy depends on the particular TPS used. This study aimed to investigate the relationship between the gamma passing rate (GPR) and the clinically relevant dose-volume index based on the predicted 3D patient dose distribution derived from two TPSs (XiO, RayStation). Methods: Twenty-one breast intensity-modulated radiation therapy plans were inversely optimized using XiO. With the same plans, both TPSs calculated the planned dose distribution. We conducted per-beam measurements on the coronal plane using a 2D array detector and analyzed the difference in 2D GPRs between the measured and planned doses by commercial software. Using in-house software, we calculated the predicted 3D patient dose distribution and derived the predicted 3D GPR, the predicted per-organ 3D GPR, and the predicted clinically relevant dose-volume indices [dose-volume histogram metrics and the value of the tumor-control probability/normal tissue complication probability of the planning target volume and organs at risk]. The results derived from XiO were compared with those from RayStation. Results: While the mean 2D GPRs derived from both TPSs were 98.1% (XiO) and 100% (RayStation), the mean predicted 3D GPRs of ipsilateral lung (73.3% [XiO] and 85.9% [RayStation]; p < 0.001) had no correlation with 2D GPRs under the 3% global/3 mm criterion. Besides, this significant difference in terms of referenced TPS between XiO and RayStation could be explained by the fact that the error of predicted V5Gy of ipsilateral lung derived from XiO (29.6%) was significantly larger than that derived from RayStation (-0.2%; p < 0.001). Conclusions: GPR is useful as a patient quality assurance to detect dosimetric errors; however, it does not necessarily contain detailed information on errors. Using the predicted clinically relevant dose-volume indices, the clinical interpretation of dosimetric errors can be obtained. We conclude that a clinically relevant dose-volume index based on the predicted 3D patient dose distribution could add to the clinical and biological considerations in the GPR, if we can guarantee the dose calculation accuracy of referenced TPS.