A COVID-19 Non-contact Screening System Based on XGBoost and Logistic Regression (Preprint)

BACKGROUND COVID-19 is a new infectious disease with high infectivity. At present, body temperature detection is the main method for primary screening, but this single detection method has poor accuracy and is easy to miss detection. OBJECTIVE The objective of our study was to propose a non-contact, high-precision COVID-19 screening system. METHODS We used impulse-radio ultra-wideband (IR-UWB) radar to detect the respiration, heart rate, body movement, sleep quality, and various other physiological indicators. We collected 140 radar monitoring data from 23 COVID-19 patients in Wuhan Tongji Hospital, and compared them with 144 radar monitoring data of healthy controls. Then XGBoost and logistic regression(XGBoost+LR) algorithm was used to classify the data of patients and healthy people; feature selection was performed by SHAP value; using ten-fold cross-validation, XGBoost+LR algorithm was compared with five other classic classification algorithms, and the classification performance was evaluated by precision, recall, and the area under the ROC curve( AUC ). RESULTS The XGBoost+LR algorithm demonstrate excellent discrimination (precision=99.1 %, recall rate = 94.1 %, AUC=98.7 %), which is superior to several other single machine learning algorithms. In addition, the SHAP value indicate that number of apnea during REM(‘ REMSATims’) and mean heart rate(‘meanHR’) are important features for classification. CONCLUSIONS The COVID-19 non-contact screening system based on XGBoost+LR algorithm can accurately predict COVID-19 patients and can be applied in isolation wards to effectively help medical staff.