Automated Treatment Planning System for Uveal Melanomas Treated With Proton Therapy: A Proof-of-Concept Analysis

Abstract

By precalculation of an entire set of planning solutions for protons, penalizing them and providing a graphical navigator tool (Automated Treatment Planning [ATP]), we aim to improve the efficiency of the planning procedure for uveal melanoma (UM) and make it independent of treatment planner experience. A phase space of plans is evaluated by transforming the eye model in each gaze angle, calculating cumulative dose-volume histograms for each position, and defining a dose-volume constraint for each considered structure. The final result is a map of the plan phase space, displaying how many criteria are fulfilled for each gaze angle. To test its usability and performance, ATP was used retrospectively on 48 UM patients treated with protons. In 36 of 48 cases (75%), the planning result was either the same (13 of 48, 27%) or comparable (23 of 48, 48%). In 11 of 48 evaluated cases (23%), ATP plans showed improvements. In 1 case (2%) the patient's visual acuity had been impaired, and an optimization was not possible. We have developed a dose calculation and planning engine that prepares a set of treatment plans covering a wide range of theoretical clinically feasible gaze angles for a given patient, by precalculating the dose distributions for each gaze angle. By considering different structures and adapting their constraints, the identification of the optimal gaze angle can be realized. With a better understanding of the dose-volume constraints and the development of strategies to react to the trade-offs between considered structures, ATP may lead to a complete automation of the planning process for UM treated with proton therapy.