Related Experiment Video
Updated: May 28, 2025
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
Machine Learning Methods Based on Chest CT for Predicting the Risk of COVID-19-Associated Pulmonary Aspergillosis
Jiahao Liu1, Juntao Zhang2, Huaizhen Wang3
1Department of Radiology, The First Affiliated Hospital of Shandong First Medical University & Shandong Provincial Qianfoshan Hospital, Jinan, China (J.L., H.W., L.W., J.C., M.L., Q.Z.); Shandong First Medical University, Jinan, China (J.L., M.L., S.W.).
Rationale And Objectives:
To develop and validate a machine learning model based on chest CT and clinical risk factors to predict secondary aspergillus infection in hospitalized COVID-19 patients.
Materials And Methods:
This retrospective study included 291 COVID-19 patients with complete clinical data between December 2022 and March 2024, and some (n=82) of them developed secondary aspergillus infection after admission. Patients were divided into training (n=162), internal validation (n=69) and external validation (n=60) cohorts. The least absolute shrinkage and selection operator regression was applied to select the most significant image features extracted from chest CT. Univariate and multivariate logistic regression analyses were performed to develop a multifactorial model, which integrated chest CT with clinical risk factors, to predict secondary aspergillus infection in hospitalized COVID-19 patients. The performance of the constructed models was assessed with the receiver operating characteristic curve and the area under the curve (AUC). The clinical application value of the models was comprehensively evaluated using decision curve analysis (DCA).
Results:
Eleven radiomics features and seven clinical risk factors were selected to develop prediction models. The multifactorial model demonstrated a favorable predictive performance with the highest AUC values of 0.98 (95% CI, 0.96-1.00) in the training cohort, 0.98 (95% CI, 0.96-1.00) in the internal validation cohort, and 0.87 (95% CI, 0.75-0.99) in the external validation cohort, which was significantly superior to the models relied solely on chest CT or clinical risk factors. The calibration curves from Hosmer-Lemeshow tests showed that there were no significant differences in the training cohort (p=0.359) and internal validation cohort (p=0.941), suggesting the good performance of the multifactorial model. DCA indicated that the multifactorial model exhibited better performance than others.
Conclusion:
The multifactorial model can serve as a reliable tool for predicting the risk of COVID-19-associated pulmonary aspergillosis.
