A diagnostic report supervised deep learning model training strategy for diagnosis of COVID-19

Shiqi Deng, Xing Zhang, Shancheng Jiang

Abstract: Highlights•A training strategy that trains the model under diagnostic report supervision is introduced to improve the model generalization performance.•A COVID-19 related image-text pairs dataset composed of chest CT scans and diagnostic reports are constructed.•We demonstrate the advantages on model interpretability of the proposed model training strategy.