LAUSR

PURPOSE The double-stacked design of the Halcyon multileaf collimator (MLC) presents new challenges for treatment planning systems (TPSs). The leaf trailing effect has recently been described as the result of the interplay between the fluence transmitted through the leaf tip ends of each MLC layer. This effect makes the dosimetric leaf gap (DLG) dependent on the distance between the leaves of different layers (trailing distance) and is not adequately modeled by the Eclipse TPS. The purpose of our study was to investigate and report the dose discrepancies produced by these limitations in clinical plans and to explore how these discrepancies can be mitigated and avoided. METHODS The integrated platform with the Halcyon v2 system, Eclipse and Aria v15.6, was used. The dose discrepancies were obtained with EPID images and the portal dosimetry software and validated using radiochromic film dosimetry. The results for the AIDA commissioning test and for nine selected clinical beams with the sliding window intensity modulated radiotherapy (dIMRT) technique were thoroughly analyzed and presented. First, the DICOM RT plans were exported and the fluences were computed using different leaf tip models, and then were compared. Second, the detailed characteristics of the corresponding leaf sequences were investigated. Finally, modified DICOM RT plans were created in which the non-collimating (backup) leaves were retracted 2 mm to increase the leaf trailing distance, the modified plans were imported back into the TPS and the measurements were repeated. Dedicated in-house tools were developed in Python to carry out all analyses. RESULTS Dose discrepancies greater than 10% and regions of gamma failure were found in both the AIDA test and clinical beams using static-gantry dIMRT. Fluence analysis highlighted that the discrepancies were due to limitations in the MLC model implemented in the TPS. Analysis of leaf sequences indicated that regions of failure were associated with very low leaf speeds and virtually motionless leaves within the beam aperture. Some of these discrepancies were mitigated by increasing the trailing distance of the non-collimating leaves without affecting the beam aperture, but this strategy was not possible in regions where the leaves from both layers actively defined the beam aperture. CONCLUSIONS Current limitations of the MLC model in Eclipse produced discrepancies between calculated and delivered doses in clinical beams that caused plan-specific quality assurance failures and interruptions in the clinical workflow. Careful evaluation of the clinical plans produced by Eclipse for the Halcyon is recommended, especially for static gantry dIMRT treatments. Some characteristics of leaf sequences are problematic and should be avoided in clinical plans and, in general, a better leaf tip model is needed. This is particularly important in adaptive radiotherapy treatments, where the accuracy and reliability of TPS dose calculations are of the utmost importance.