Improved deep learning model for differentiating novel coronavirus pneumonia and influenza pneumonia

Min Zhou; Yong Chen; Dexiang Wang; Yanping Xu; Weiwu Yao; Jingwen Huang; Xiaoyan Jin; Zilai Pan; Jingwen Tan; Lan Wang; Yihan Xia; Longkuan Zou; Xin Xu; Jingqi Wei; Mingxin Guan; Jianxing Feng; Huan Zhang; Jieming Qu

The escalating crisis caused by SARS-CoV-2 with high infectivity and multiple routes of transmission is complicated by its co-occurrence with seasonal influenza, exactly as the things happening in the United States that some COVID-19 deaths have been misdiagnosed as influenza. The similarities in clinical symptoms between NCP and IP, along with shortage and high false negative rate of nucleic acid detection kits, make the differential diagnosis difficult (24) (25) (26) , prompting clinicians to investigate new diagnostic methods. Chest CT had high sensitivity in diagnosing NCP at early stage, giving it an advantage over nucleic acid detection in time of crisis. This is the reason why Hubei Provincial Government adopted characteristic chest CT finding as an important criterion for diagnosis of NCP at the peak of outbreak. However, the similar chest CT manifestations of NCP and IP will inevitably lead to inaccurate diagnosis even for experienced physicians, and increase the risk of over-diagnosis and cross infection (3, 8) . The main challenge to employ CT as a predominant diagnostic tool is to improve the accuracy and speed in identifying specific lesions on chest CT images.