Use of standardized uptake value thresholding for target volume delineation in pediatric Hodgkin lymphoma

Purpose: A limitation of [¹⁸F] 2-fluoro-2-deoxy-d-glucose positron emission tomography (FDGPET) in radiation planning for Hodgkin lymphoma (HL) is significant variability in delineation of tumor volume. One approach to reduce variability is to apply automatic or semiautomatic segmentation methods such as thresholding based on a percent tumor maximum standardized uptake value (SUV_max). Here, we apply various tumor SUV_max thresholds and examine their effects in involved field radiation therapy (IFRT) and involved site radiation therapy (ISRT) target volumes. Methods and materials: PET/computed tomography data sets were reviewed for 16 pediatric and young adult patients treated with IFRT. The following percent tumor SUV_max thresholds were applied to the prechemotherapy PET: 15%, 20%, 25%, 30%, 35%, and 40%. Clinical target volumes for IFRT and ISRT plans were manually generated based on these threshold volumes (CTV_PET) and compared with clinical target volumes generated using the standard qualitative visual method (CTV_QVM). Treatment plans were generated, doses to normal structures were compared, and the optimum threshold, defined as the CTV_PET that corresponded to the percent overlap closest to 100% when compared with the CTV_QVM, was determined. Results: On average, there was a 7.6-fold increase in PET volume between 40% and 15% SUV_max. When the 6 SUV_max thresholds were applied in the design of target volumes for IFRT, 2 of 16 patients had a change in treatment volume. There was a 2.4-fold increase in ISRT CTVs generated based on the 15% and 40% SUV_max, which translated into a clinically significant decrease in dose to normal structures when the ISRT plans that were generated using the 15% SUV_max volumes were compared with the 40% SUV_max. In most patients, the optimum threshold was SUV_max 15%. Conclusions: Accurate target volume delineation with [¹⁸F] 2-fluoro-2-deoxy-d-glucose PET in HL is challenging and may require more precise and reproducible segmentation methods as we move toward more conformal therapies.