OBJECTIVE To establish the automatic diagnosis system for varicella pneumonia based on multi-slice spiral computed tomography (MSCT) and verify its application value.

METHODS The clinical data were collected from 295 patients with varicella who were treated in the hospital from Jan. 2016 to Mar. 2023 and assigned as the training set, the pulmonary imaging findings were acquired from the MSCT chest scanning. The automatic diagnosis system for varicella pneumonia was established based on MSCT with the use of convolutional neural network technology. Totally 279 patients with varicella who were treated during the same period were chosen as the validation set, the result of comprehensive diagnosis was set as gold standard, and the efficiency of the above system in diagnosis of varicella pneumonia was observed.

RESULTS Totally 279 patients with varicella were included in the validation group, 243 of whom had varicella pneumonia, and 36 had simple varicella infection. The sensitivity of the automatic diagnosis system established based on MSCT was 97.53% in diagnosis of the varicella pneumonia of the validation group, with the specificity 91.67%, the accuracy 96.77%, respectively higher than 93.83%, 83.33% and 92.47% of MSCT, and it was highly consistent with the gold standard(Kappa=0.919, P < 0.001); there was consistency between MSCT and the gold standards(Kappa=0.675, P < 0.001).

CONCLUSION The automatic diagnosis system for varicella pneumonia established based MSCT can raise the sensitivity, specificity and accuracy in diagnosis of varicella pneumonia and have the advantages of automation and convenience, and serve as a new tool for clinical diagnosis of varicella pneumonia.