AI-assisted quantitative analysis of chest CT progression patterns of COVID-19

To evaluate retrospectively lung changes and progression patterns on the chest CT of Coronavirus Disease 2019(COVID-19) by using of computer artificial intelligence(AI) system quantitatively. Chest CT scans were performed during the treatment of 71 patients with confirmed COVID-19(41 men,30 women;mean age,48.2 years) from January 1 to March 5,2020. Recognizable patterns of radiographic progression were determined by comparing the volume and volume percentage of lung involvement calculated by computer-aided detection software for each patient on serial chest CT scans. Three patterns of chest CT progression were recognized:type 1(initial CT opacity followed by radiographic improvement) in 21 of 71 patients(29.6%),type 2(initial CT opacity deterioration to peak level followed by radiographic improvement) in 47 cases(66.2%),and type 3(progressive radiographic deterioration) in 3 cases(4.2%). Lesions on chest CT were completely absorbed in 14 of 71 patients, partially absorbed in 54 cases, and 3 cases were in deteriorating progress. Sixty-one patients(85.9%) had been discharged,8(11.3%) were still in hospital, and 2 cases died(2.8%). There are three types of CT progression in COVID-19. COVID-2019 has a long course and slow absorption, and most cases have good prognosis.