Decision Tool to Select Prostate Patients for Fully Adaptive Prostate Treatment vs. Intra-Fractional Monitoring and Intervention

Abstract

The use of on-line adaptive radiation therapy (OART) for prostate patients involves a lengthy procedure. It is of clinical interest to select those patients who benefit most from this procedure and to select the optimal planning constraints. Previous attempts use fitted functions as a way to convert geometric parameters to dose constraints. We propose to use the relative overlap volume between PTV and rectum (Vrel) and hypothesize that this represents dosimetric data well. A group of 87 hypofractionatedly (20 fractions) treated prostate patients were categorized in three subgroups: A) Prostate + seminal vesicles as a single target, B) Prostate + seminal vesicles SIB, and C) Prostate only. Using ROC analysis, calculating the AUC and maximizing accuracy the optimal surrogate level for the relative overlap volume between PTV and Rectum was found by varying the threshold of Vrel to generate the ROC-curve. The specific value of Vrel representing the dosimetric constraint is found by identifying the maximum accuracy as a function of threshold when the data permits to do that as in some choices of dosimetric constraints Vrel shows a monotone behavior. The different groups provided in our planning study provided data on dose levels constraints with respective ranges for groups A, B, and C of [3.5:28]%, [6:25]%, and [2:15.7]%. For the different groups the optimal values predicting different constraint levels are presented. Using ROC analysis, the threshold exhibiting maximal accuracy in predicting constraint levels are chosen. The reduction of dosimetric constraints to geometric constraints allows to categorize patients at time of target contouring before lengthy planning procedures are initiated. In the online adaptive process, where the overlap volume can change every fraction, these geometric measures can be used for comparing the original plan quality with newly created plans.