A Pragmatic Approach to Assessing Structure-specific Dosimetric Error of VMAT Delivery

Abstract

Listen

Translate article

VMAT adopts a multi-dimensional modulation methodology. However, multi-dimensional modulations present multi-dimensional challenges. Firstly, electro-mechanically, the multi-dimensional modulation paradigm tends to perform less reliably than its one-dimensional pre-cursor. Secondly, for plans demanding extensive gantry angular speed modulation, persistent acceleration and deceleration could lead to dosimetric error due to mechanical constraint. Temporal dose rate modulation with high frequency and amplitude results in the same dosimetric consequences because of the time delay and imperfect beam pulses. More significantly, beam hold-off signals are not asserted for VMAT due to the excessive rotational inertia of the LINAC head. Consequently, it is not surprising to observe larger leaf RMS errors for VMAT plans than those for IMRT plans. The compounding effects of all these impose a significant technical challenge in clinical situations where steep dose gradients need to be accurately verified and faithfully delivered. Although there are QA tools to measure delivered doses, they, in general, either lack the required spatial resolution or detect the peripheral dose only or require extensive post-irradiation processing and are not structure-specific. Here, we present a practical and efficient approach to retrospectively calculating actually delivered dose using VMAT Dynalog files. The VMAT DynaLog file was an extensive treatment log file that contained the theoretical and actual leaf positions at the MLC physical plane at a given gantry angle and was automatically saved on the 4-Dimensional Integrated Treatment Console (4DITC). The leaf positions were sampled at a time interval of 50 ms. Due to beam divergence and rounded-leaf-end design, the physical leaf positions were converted to the projected positions at the iso-plane using a polynomial model. The actually delivered DVA file was then reconstructed iteratively and fed back to our homegrown VMAT planning system for forward dose calculation. Dosimetric parameters of clinical significance between the planned and reconstructed were then compared. It was found that there were observable differences between the planned and delivered doses in several important dosimetric parameters. For D95, the planned and delivered doses were 98.97% and 96.55%, respectively. The same trend was also observed for V95, which was 99.04% and 97.12% for the planned and delivered. The PTV Dmin was 105.09% and 103.46%, respectively. The same dosimetric behavior was detected for a group of randomly sampled VMAT patients, revealing potential delivery hardware deficiency. We developed a pragmatic and efficient technique to assess delivered VMAT doses to specific structures. The method provides concrete dosimetric evidence for adaptive VMAT treatment planning.