SU‐GG‐J‐153: Voxel‐Based Phenomenological SUV‐Dose Response Model for the Human Parotid Glands

Purpose : The goal of this study is to evaluate SUV as an imaging biomarker of parotid function following radiotherapy (RT) and to develop a phenomenological model of SUV‐dose response, enabling prediction of the biomarker prior to RT. Method and Materials : As part of an ongoing study, patients (n = 8) with head and neck cancer (HNC) had 18 F‐FDG PET/CT studies performed before (within 30 days) and following RT (∼50 days). In addition, resting and stimulated whole mouth saliva was collected. Next, a deformable image registration algorithm was utilized to map parotid contours from planning CT (PCT) images to pre‐ and post‐RT PET/CT studies, separately. Mean SUV was then calculated for the parotid glands at both time points and compared to collected saliva. The SUV‐dose response relationship was assessed in a separate cohort (n = 60) also treated for HNC. Pre‐ and post‐RT PET/CT images were deformably aligned to PCT images to achieve correspondence between SUV and planned dose. SUV response was then modeled using a single parameter, incorporating planned dose and pre‐RT SUV; defined as metabolic dose. Results : Post‐RT stimulated saliva and SUV decreased (59% and 20%, respectively) in seven patients. For one patient, stimulated saliva and SUV increased (14% and 11%, respectively). Fractional SUV correlated significantly with xerostomia grade, P < 0.01 and Spearman's ρ = −0.964. A second order polynomial provided an adequate fit to the log transformed SUV‐dose response curve, adjusted R 2 = 0.95. Post‐RT SUV was predicted within 3% for an exemplary patient, but overestimated (50%) in the worst case. Conclusions : Pilot data suggest that SUV is significantly correlated with parotid saliva output. Furthermore, the feasibility of predicting reductions in SUV, prior to treatment delivery, has been demonstrated. Conflict of Interest : Research partially sponsored by Varian Medical Systems.