Computed Tomography scanning in patients with COVID-19: artificial intelligence analysis of lesions volume and outcome.

OBJECTIVE

The aim of this study was to summarize the computed tomography (CT) chest scanning results of COVID-19 patients, and to assess the value of artificial intelligence (AI) dynamics and quantitative analysis of lesion volume change for the evaluation of the disease outcome.

PATIENTS AND

METHODS

First chest CT and reexamination imaging data of 84 patients diagnosed with COVID-19 who were treated at Jiangshan Hospital of Guiyang, Guizhou Province from February 4, 2020, to February 22, 2020, were retrospectively analyzed. Distribution, location, and nature of lesions were analyzed according to the characteristics of CT imaging and COVID-19 diagnosis and treatment guidelines. Based on the results of the analysis, patients were divided into the group without abnormal pulmonary imaging, the early group, the rapid progression group, and the dissipation group. AI software was used to dynamically measure the lesion volume in the first examination and in the cases with more than two reexaminations.

RESULTS

There were statistically significant differences in the age of patients between the groups (p<0.01). The first chest CT examination of the lung without abnormal imaging findings mainly occurred in young adults. Early and rapid progression was more common in the elderly, with a median age of 56 years. The ratio of the lesion to the total lung volume was 3.7 (1.4, 5.3) ml 0.1%, 15.4 (4.5, 36.8) ml 0.3%, 115.0 (44.5, 183.3) ml 3.33%, 32.6 (8.7, 98.0) ml 1.22% in the non-imaging group, early group, rapid progression group, and dissipation group, respectively. Pairwise comparison between the four groups was statistically significant (p<0.001). AI measured the total volume of pneumonia lesions and the proportion of the total volume of pneumonia lesions to predict the receiver operating characteristic (ROC) curve from early development to rapid progression, with a sensitivity of 92.10%, 96.83%, specificity of 100%, 80.56%, and the area under the curve of 0.789.

CONCLUSIONS

Accurate measurement of lesion volume and volume changes by AI technology is helpful in assessing the severity and development trend of the disease. The increase in the lesion volume proportion indicates that the disease has entered a rapid progression period and is aggravated.