Purpose: To develop automated tools and perform a cross-institutional study to evaluate patient anatomy trends using in vivo EPID transit images collected after every fraction of treatment. The tools were designed to forego a complex commissioning process and were deployed at two cancer centers, that have disparities in access to cancer care, to analyze head-and-neck cancer patients.
Methods: For 252 patients treated at academic institutions in South Africa (40 patients) and United States (212 patients), analysis of in vivo EPID images for head-and-neck patients was performed. Both centers treated patients on the Varian Halcyon which collects EPID images for every fraction of treatment. Deviations relative to the first fraction of treatment were assessed, for 5630 fractions, with a novel technique, Gradient-Dose Segmented Analysis (GDSA) which correlates with changes in the PTV mean dose. The change in PTV mean dose over a treatment course was compared between centers and individual patient trends were investigated.
Results: The mean and variance of the change in PTV mean dose were statistically equivalent for the two centers, demonstrating comparable magnitude and frequency of anatomical changes. Both centers observed patient weight loss/tumor shrinkage in the later fractions of treatment. One center observed cyclic patterns in the patient’s external body contour, for half of the patients analyzed. These trends were not observed at the other institution and may be driven by differences in the access to cancer care for the two centers.
Conclusion: We have developed an automated framework that can measure in vivo patient anatomy trends and how they compare between different clinics. In aggregate, anatomy deviations of head-and-neck patients at two different centers analyzed were equivalent. Further study is required to understand the source of different anatomy change trends observed. This framework is portable and can be deployed to other centers to perform similar analyses.
Funding Support, Disclosures, and Conflict of Interest: Presenting author receives grant funding from Varian Medical Systems.