Establishment of machine learning-based tool for early detection of pulmonary embolism

Lijue Liu, Yaming Li, Na Liu, Jingmin Luo, Jinhai Deng, Weixiong Peng, Yongping Bai, Guogang Zhang, Guihu Zhao, Ning Yang, Chuanchang Li, Xueying Long

Abstract: Highlights•Screening features that passed the hypothesis test were established with reference to the 2019 ESC guidelines for diagnosis and management of acute pulmonary embolism. The data set of the study on pulmonary embolism was established by cleaning, sorting and screening the massive data obtained from hospitals.•Five machine learning (SVM, LogisticRregression, random forest XGBoost, and BP neural network) models for pulmonary embolism were developed. XGBoost model is approved as optimal among the five models, and its sensitivity, specificity and missed diagnosis rate are all superior to the comparison model, reaching the standard of assisting doctors in the clinical application.•The important features that constitute the XGBoost decision result are obtained. the 2019 ESC guideline has shown that these features are also important in the clinic, suggesting that our model has learned important information about screening for pulmonary embolism.•The model is used prior to pulmonary angiography and only requires the input of routine laboratory and test results to assess the patient's risk of pulmonary embolism and provides a reference for doctors in the next-step examination.

External IDs:dblp:journals/cmpb/LiuLLLDPBZZYLL24