Dosimetric sensitivity of an enhanced leaf model (ELM) for individual versus averaged machines

With the introduction of a new multi-leaf collimator (MLC) enhanced leaf model (ELM) in the Varian Eclipse™ treatment planning system, there is currently limited data regarding the dosimetric sensitivity to real-world variation in the ELM parameters, and its clinical relevance.

To characterize the variation in ELM parameters across a large department with ten linear accelerators and investigate the feasibility of using a single machine-averaged ELM for treatment planning. This could achieve time and resource savings from reduced quality assurance, while allowing easy transfer of patients between machines.

Clinical plans of a range of sites (head and neck, prostate, breast, lung, and brain), techniques (VMAT, IMRT, SBRT, and SRS), and energies (6 MV, 6 MV FFF, 10 MV, and 10 MV FFF) were recalculated on Varian TrueBeam™ (120 MLC) and Varian EDGE™ (HD120 MLC), with machine-specific ELM beam models, an averaged machine and an outlier machine model. A range of clinically relevant metrics relating to target coverage (e.g. PTV D_98%, D_50%, D_2%) and OAR doses (dosimetric, volumetric, conformity, and gradient indices) were evaluated.

For the target metrics, the maximum percentage deviation from the mean was 0.422%, 0.157%, and 1.956% for the cases of the individual machines, the averaged machine and the outlier machine correspondingly, while the maximum absolute dose differences were 0.28 Gy, 0.07 Gy, and 0.38 Gy. For the OAR metrics, the maximum deviation from the mean was 1.833%, 0.204%, and 5.722% for the individual, averaged, and outlier machines, while the maximum absolute dose differences were 0.41 Gy, 0.10 Gy, and 0.97 Gy.

For machines that are well matched in terms of dosimetry for transmission and sweeping gap fields, the use of an averaged machine model is unlikely to introduce clinically significant dosimetric differences to treatment plans.