Comparison of atlas-based and deep learning methods for organs at risk delineation on head-and-neck CT images using an automated treatment planning system

•DL methods are faster and more performing than ABAS methods.•DL model can reach high performances with a limited training dataset.•ABAS methods give consistent results but are less time efficient.•Small dosimetric impact observed on plans generated with automatic contours.•All solutions allow reducing the inter-observer variability in contouring.