A novel optimization algorithm for enabling dynamically collimated proton arc therapy

Abstract

The advent of energy-specific collimation in pencil beam scanning (PBS) proton therapy has led to an improved lateral dose conformity for a variety of treatment sites, resulting in better healthy tissue sparing. Arc PBS delivery has also been proposed to enhance high-dose conformity about the intended target, reduce skin toxicity, and improve plan robustness. The goal of this work was to determine if the combination of proton arc and energy-specific collimation can generate better dose distributions as a logical next step to maximize the dosimetric advantages of proton therapy. Plans were optimized using a novel DyNamically collimated proton Arc (DNA) genetic optimization algorithm that was designed specifically for the application of proton arc therapy. A treatment planning comparison study was performed by generating an uncollimated two-field intensity modulated proton therapy and partial arc treatments and then replanning these treatments using energy-specific collimation as delivered by a dynamic collimation system, which is a novel collimation technology for PBS. As such, we refer to this novel treatment paradigm as Dynamically Collimated Proton Arc Therapy (DC-PAT). Arc deliveries achieved a superior target conformity and improved organ at risk (OAR) sparing relative to their two-field counterparts at the cost of an increase to the low-dose, high-volume region of the healthy brain. The incorporation of DC-PAT using the DNA optimizer was shown to further improve the tumor dose conformity. When compared to the uncollimated proton arc treatments, the mean dose to the 10mm of surrounding healthy tissue was reduced by 11.4% with the addition of collimation without meaningfully affecting the maximum skin dose (less than 1% change) relative to a multi-field treatment. In this case study, DC-PAT could better spare specific OARs while maintaining better target coverage compared to uncollimated proton arc treatments. While this work presents a proof-of-concept integration of two emerging technologies, the results are promising and suggest that the addition of these two techniques can lead to superior treatment plans warranting further development.