FocusNetv2: Imbalanced large and small organ segmentation with adversarial shape constraint for head and neck CT images

•Proposing a two-stage segmentation framework for OAR segmentation for HaN CT images.•Segmenting large organs with multi-scale convolutional neural network, while first localizing and then segmenting small organs.•Adding adversarial shape constraint to regularize the estimated mask, which makes the prediction consistent with medical prior.•Testing our method on both large real data from hospitals and an open source dataset and shows superior performance.