Objective To investigate the computed tomography (CT) characteristics and diagnostic value of novel coronavirus pneumonia (NCP or COVID-19) in pregnancy. Methods This study included ten pregnant women infected with COVID-19, treated in the Zhongnan Hospital of Wuhan University from January 20, 2020 to February 6, 2020. Clinical and chest CT data were collected and clinical symptoms, laboratory indicators, and CT images were analyzed to explore CT characteristics and diagnostic value for COVID-19 during pregnancy. Results Laboratory examination showed that white blood cell count was normal in nine patients, and slightly higher in one patient (10.23 × 109). The lymphocyte ratio decreased in two patients by 12% and 14%, respectively. The levels of C-reactive protein was elevated in seven patients (range, 21.16-60.3 mg/L) and the levels of D-dimer was increased in eight patients (range, 507-2141 ng/mL). Six patients had low levels of total protein (range, 35.3-56.5 mg/ L). Two patients showed small patchy ground glass opacity (GGO) involving single lung, while eight patients showed multilobe GGO in both the lungs, with partial consolidation. Peripheral and non-peripheral lesion distributions were seen in ten (100%) and four (40%) patients, respectively. There were four patients who had signs of intra-bronchial air-bronchogram, six patients had small bilateral pleural effusions, while none had lymphadenopathy. Dynamic observations were performed in four patients after COVID-19 treatment. Among these four patients, one patient showed normal on the initial examination, and new lesions were observed after 3 days; 1 patient showed progression after 7 days of treatment, with expansion of the lesion area; and the other 2 patients showed improvement after 14 days of

Across four studies, we document a novel and unexpected effect such that COVID-19 data gathered by AI (vs. humans) reduces consumers' intentions to take preventive measures. This effect is driven by greater perceived dehumanization, such that consumers view AI-generated data as numbers, rather than as human lives. We find that this effect is mitigated when humanness is primed, is moderated by AI type (narrow AI leads to greater dehumanization than general AI), and is attenuated by task type (perceived vaccine effectiveness is similar for AI and humans when the vaccine development process highlights AI-advantageous attributes).