Non-invasive Diagnosis of COPD with E-nose Using XGBoost Algorithm

Abstract

This work presents the application of a self-developed electronic nose (e-nose) system in the detection of the respiratory disease chronic obstructive pulmonary disease (COPD). The volatile organic compound biomarkers of COPD are detected and analyzed with an array of gas sensors. The fabricated e-nose system is composed of eight gas sensors, a processing circuit, a control circuit, and a pattern recognition system. Two dimensionality reduction techniques principal component analysis (PCA) and linear discriminant analysis (LDA) were used and the classification model is developed with the ensemble learning method, XGBoost. The system is tested in thirty COPD patients and forty healthy controls to differentiate the breath samples. The PCA-XGBoost model obtained better classification results with 88.57 % accuracy, 86.67 % sensitivity, and 90.0% specificity.