A Step Towards Making VMAT TBI More Prevalent: Automating the Treatment Planning Process

Abstract

Compared to conventional TBI techniques, VMAT TBI has been shown to provide superior lung sparing while maintaining target coverage. However, one of the significant barriers preventing widespread VMAT TBI implementation is the time-consuming treatment planning process. To overcome this issue, we developed an in-house automated treatment planning process for VMAT TBI and evaluated the resulting dosimetric plan quality against manually generated treatment plans.Application Programming Interface (API) scripting was utilized to interface with a commercial treatment planning system. Two scripts were built to facilitate autoplanning: the first generates optimization structures, creates the plan, places the beams, and sets the optimization constraints using user-defined input. The second uses the output from the first script to run successive VMAT optimizations, where the plan quality is evaluated after each iteration and the optimization objectives are adjusted according to dosimetric deviation of each constraint from its objective. The optimization loop proceeds until either all plan objectives are met, or the number of iterations reaches a user-defined maximum. Nine patients previously treated in our clinic with VMAT TBI were used to evaluate the efficacy of the proposed autoplanning process. The plan quality of the produced auto plans was evaluated by comparing against the manually generated clinical plans. The clinically relevant metrics, including global Dmax, PTV V110%, lungs and lungs-1cm Dmean, kidneys Dmean, and bowel Dmax, were analyzed. All plans were normalized to deliver 100% of the prescription dose to 90% of the PTV. All dosimetric comparisons between the auto and manual plans were made with the dose expressed as a percentage of the prescription dose. Paired t-test was used to compare the dosimetric indices between auto and manual plans.No significant differences were observed between the manual and autoplan global Dmax (P < 0.893), PTV V110% (P < 0.734), kidneys Dmean (P = 0.351), and bowel Dmax (P < 0.473). Statistically significant decreases in the Dmean to the lungs and lungs-1cm volumes of 5.4% ± 6.4% (P < 0.024) and 6.8% ± 7.4% (P < 0.017), respectively, were obtained with the autoplans as compared to the manual plans. The required time for treatment planning (i.e., contouring and planning) ranged from approximately 2-3 days for the manual plans as compared to approximately 3-5 hours for the autoplans.Superior lung sparing with the same target coverage and similar global Dmax were observed with the autoplans as compared to the manual treatment plans. Large reductions in planning time were obtained using the developed autoplanning scripts as compared to manual planning, thus reducing the required effort of the treatment planning team and making VMAT TBI a more accessible treatment technique. The developed scripts will be made open-source and freely available to the public in the near future.