Objectives: Local recurrence in head and neck cancer after chemoradiotherapy is associated with poor outcome. The ability to predict such local recurrence could be of clinical value in order to identify patients that may benefit from intensified treatment and/or more frequent follow-up. The objective of this study was to investigate the value of proliferative tumor volumes characterization from 18F-FLT PET images for prognosis of disease free survival (DFS) and prediction of locoregional recurrence. Methods: 45 head and neck cancer patients treated with (chemo)radiotherapy who underwent a pre-treatment 18F-FLT PET were retrospectively included. Proliferative tumor volumes were characterized after automatic delineation with the Fuzzy Locally Adaptive Bayesian algorithm using image analysis providing parameters such as SUV measurements, as well as 3D shape descriptors and textural features. DFS prognostic value was assessed using Kaplan-Meier analysis. Locoregional recurrence predictive value was assessed using a random forest (RF) classifier that was subsequently validated using the leave-one-out strategy. Results: Accuracy of the RF classifier in predicting locoregional recurrence using a subset of the most optimal features was 86 %. Kaplan-Meier analysis demonstrated that textural features and proliferative volume are prognostic factors of DFS, larger, more heterogeneous proliferative volume being associated with shorter DFS. Conclusions: Our results suggest that 3D shape descriptors and textural features characterizing shape and heterogeneity of head and neck tumors proliferative volumes in baseline 18F-FLT images may help identify subgroups of patients at increased risk of developing a locoregional recurrence after head and neck cancer (chemo)radiotherapy.