Automated proton therapy planning: clinical validation for head-and-neck cancer and chordoma

Abstract

Intensity-modulated proton therapy (IMPT) enables highly conformal radiation dose delivery while reducing radiation exposure to surrounding healthy tissues. However, treatment planning for head and neck cancer (HNC) and skull base chordomas is complex, time-consuming, and highly dependent on planner expertise. This thesis aimed to develop and clinically validate fully automated multicriterial IMPT planning using Erasmus-iCycle.

A treatment site-specific wish-list was developed and iteratively optimized for HNC. Automated IMPT plans for HNC patients showed comparable target coverage and consistently lower doses to organs at risk than clinical plans, with significantly lower predicted risks of xerostomia and dysphagia. In a blinded clinical assessment, three experienced radiation oncologists considered all automated plans clinically acceptable and preferred them over the clinical plans in the vast majority of cases. Automated planning was subsequently developed for challenging nasopharyngeal cancer and skull base chordoma cases. For nasopharyngeal cancer, automated planning achieved similar or improved target coverage and lower doses to critical organs, while transmission beams provided additional sparing of serial organs at risk. For skull base chordomas, automated plans generally achieved similar or improved plan quality, although patient-specific adjustments were required in some cases.

Overall, fully automated IMPT planning is clinically feasible and can improve efficiency, consistency, and plan quality for several tumor sites.