Abstract
The use of adaptive radiotherapy (ART) may improve care for those receiving head and neck cancer treatment by reducing dose to organs at risk (OARs) and allowing margin reduction for targets. Current literature investigates predetermined and triggered adaptation, but how to trigger and how often to assess remain open questions. So, a study of adaptive assessment schemes may help fill this gap in literature. One way to achieve this is through the creation of a reliable evaluation framework.
This study investigated the performance of midpoint and biweekly assessment of H&N treatment plans through a novel evaluation framework. Midpoint and biweekly assessment of H&N treatment plans was retroactively performed for three patients who received radiation treatment at the University of Nebraska Medical Center. These assessment schemes use a validated workflow based in Varian’s Velocity software. Assessments triggered adaptive replanning if the dose to any of seven organs-at-risk was projected to exceed physician-determined dose constraints.
Once completed, the schemes were evaluated on their trustworthiness, dosimetric benefit, and efficiency. Trustworthiness was quantified through the ability of schemes to catch necessary replans and project final patient dose. Both schemes performed similarly when catching when replans were necessary, showing equal sensitivity and specificity metrics of 0.75 and 0.94-0.95, respectively. Midpoint and biweekly schemes predicted final delivered dose given replans with sensitivities of 0.95 and 0.86, respectively. Midpoint assessment had a specificity of 0.5, while the biweekly assessment had insufficient data to calculate specificity.
Dosimetric benefit was quantified through comparison of the number of OARs whose dose constraints were violated without ART, with the total number seen when an assessment scheme was used. From four total OAR constraints being violated without adaptation, midpoint and biweekly assessment decreased this number to two and zero, respectively.
Lastly, the efficiency of both schemes was quantified as time-cost-effectiveness, where the cost was physicist time in hours. Regardless of the way benefit was quantified, the midpoint assessment performed more efficiently (i.e., with higher time-cost-effectiveness). This work shows that a clear picture of different modes of ART is given by this framework.