Development and Evaluation of an AI System for COVID-19

Cheng Jin; Weixiang Chen; Yukun Cao; Zhanwei Xu; Xin Zhang; Lei Deng; Chuansheng Zheng; Jie Zhou; Heshui Shi; Jianjiang Feng

To further understand the performance of the AI system, we evaluated it on subsets divided by gender, age and number of CT scans. In the subsets divided by gender, the diagnostic performance of men was higher than that of women. We noticed an obvious difference in the size of lungs and lesions between men and women. This is consistent with the conclusion of Xiong et al. [22] that women have higher antiviral immunity than men, leading to a longer incubation period of COVID-19. In the subsets divided by age, the disease developed rapidly in the elderly patients and the CT findings were more obvious than young patients. This suggests that different thresholds should be adopted according to gender and age. The results on the subsets divided by the number of CT scans showed that diagnosis results by the AI system have little correlation with the stages of CT scans. These observations may provide decision-making references for the diagnosis of COVID-19 in different populations.