Diagnostic value of CT-based machine learning model for stage I pneumoconiosis

YAN Chengfeng; JIAO Tianyu; ZENG Qingshi

Abstract：Objective To investigate the diagnostic value of CT-based machine learning model for stage I pneumoconiosis.Methods We retrospectively collected clinical data and CT images of 202 patients diagnosed with stage I pneumoconiosis and 199 normal individuals from our hospital.We selected regions of interest(ROI)on each patient's CT lung window image using 3D-slicer software,and the patients were randomly split into training and validation cohorts in the ratio of 7:3.We used the least absolute shrinkage and selection operator algorithm to screen the features extracted from each case.The support vector machine algorithm was then used to build a CT-based machine learning model.The area under the ROC curve(AUC)and decision curve analysis(DCA)were applied to evaluate the performance of the model.Results 851 features were extracted from the CT im-ages and 9 features were filtered to build a CT-based machine learning model.The AUC of the model was 0.930(95%CI 0.901～0.963)in the training cohort and 0.820(95%CI 0.742～0.895)in the validation cohort.DCA showed the net benefit of the model.Conclusion CT-based machine learning model can effectively differentiate between normal and stage I pneumoconio-sis,which is of great diagnostic value for stage I pneumoconiosis.