Purpose: To develop computationally efficient and fast process of plan generation for use in the optimization of multi-leaf collimator dosimetric leaf gap and transmission.
Methods: Eighteen sliding window intensity modulated radiotherapy treatment plans with varying number of fields, collimator angles, and monitor units were selected. The planar doses corresponding to a set of transmission and dosimetric leaf gaps were calculated using the treatment planning system. Each pixel value in these series of planar doses was fit with an independent linear function for transmission and dosimetric leaf gap. The interpolated dose planes within the bounding values of DLG and transmission were compared to the calculated dose planes from the treatment planning system. Mean absolute difference (MAE), mean squared difference (MSE), and structure similarity metric (SSM) were used to compare the interpolated dose planes to the TPS generated and exported dose planes.
Results: The interpolation method produced 116,688 dose planes in 2.5x10^5 seconds, or 2.16s per dose plane, as opposed to 13s per dose plane when calculating in the treatment planning system (without accounting for read-write operations required to export the dose plane). The use of multivariate linear interpolation to calculate dose planes produces nearly identical results to the dose planes calculated in the treatment planning system with MAE of 0.0207 ± 0.0286 cGy, MSE of 0.0484 ± 0.1768 cGy², and SSM of 0.9999 ± 0.0001.
Conclusion: Using multivariate linear interpolation provides an accurate and time efficient process for generating dose planes required for optimization of dosimetric leaf gap and transmission.