Federated semi-supervised learning for COVID region segmentation in chest CT using multi-national data from China, Italy, Japan

•First work to investigate solutions for collaborative medical AI model for COVID-19.•A novel federated semi-supervised learning technique is proposed to train a network.•Method fully utilizes available data from different sources, w/ or w/o annotations.•Multinational dataset of 1704 scans from 3 countries (China, Italy, and Japan).•Studies fully evaluate and understand different configurations of the method.