Related Experiment Video
Updated: Jan 7, 2026
Radiation Planning Assistant - A Streamlined, Fully Automated Radiotherapy Treatment Planning System
Published on: April 11, 2018
Qualitative Blinded Clinical Assessment of Automated and Manual IMPT Plans for Head and Neck Cancer
Merle Huiskes1, Eleftheria Astreinidou1, Martin de Jong1
1Department of Radiation Oncology, Leiden University Medical Center, Leiden, The Netherlands.
Purpose:
To clinically validate fully automated intensity-modulated proton therapy (IMPT) planning for head and neck cancer (HNC) through blinded comparison with manual plans by experienced radiation oncologists (ROs).
Methods And Materials:
Thirty patients with HNC treated with a manually created IMPT plan were retrospectively included. Dose prescription was 7000 cGy (relative biologic effectiveness) to the primary tumor and 5425 cGy (relative biologic effectiveness) to the bilateral elective nodal volumes. Fully automated IMPT plans were generated using Erasmus-iCycle, applying the same clinical 4-beam configuration and dose constraints. Three experienced HNC ROs independently assessed the automated and corresponding manual plans in a blinded manner. Each RO evaluated both plans, blinded to the planning method, to determine clinical acceptability and indicate their preference, if any. Interobserver differences in preference were evaluated using pairwise McNemar tests. A generalized linear mixed model was used to assess overall preference for automated plans.
Results:
All 30 automated IMPT plans were considered clinically acceptable, whereas 28 of the 30 manual plans were rated clinically acceptable by all 3 ROs. In 28 out of 30 cases, the ROs unanimously (P = 1.00 between each pair of observers) preferred the automated plan. In the remaining 2 cases, 1 of the 3 ROs had no preference. The manual plan was never preferred. The overall preference for automated plans over no preference was statistically significant (P< .001). Target coverage was considered equivalent, whereas target conformality and sparing of both serial and parallel organs at risk were generally rated superior in the automated plans.
Conclusions:
All fully automated IMPT plans for HNC were considered clinically acceptable by experienced HNC ROs. Automated plans were unanimously preferred over manually created plans in nearly all cases, with statistical significance, highlighting the clinical potential of automated IMPT planning tools.
