DeepPrognosis: Preoperative prediction of pancreatic cancer survival and surgical margin via comprehensive understanding of dynamic contrast-enhanced CT imaging and tumor-vascular contact parsing

•Present a multi-task learning framework to conduct a joint prediction of overall survival and resection margin for PDAC patients.•Propose a novel 3D Contrast-Enhanced ConvLSTM to learn the enhancement dynamics of tumor attenuation from multi-phase CE-CT images.•Exploit a self-learning approach for automated segmentation of pancreas and peripancreatic anatomies.•The new staging biomarker integrating both the proposed risk signature and margin prediction has evidently added values to be combined with the current clinical staging system.