Breath acetone classification using XGBoost algorithm for diabetes detection

Anna Paleczek, Dominik Grochala, Artur Rydosz

Institute of Electronics, Faculty of Computer Science, Electronics and Telecommunications, AGH University of Science and Technology, al. A. Mickiewicza 30, 30-059 Krakow, Poland

Poster PDF

Abstract: 

Currently, the analysis of exhaled air is becoming more and more popular as a non-invasive method that enables the medical diagnosis of various diseases, e.g. diabetes, asthma or various cancers. In the research artificial breath samples were analyzed using a designed e-nose system containing various gas sensors. Breathing simulations were focused on acetone, which has been a known diabetes biomarker for years. The classification of the prepared breath samples, for the purpose of diabetes detection, was carried out with the use of machine learning algorithms. The results showed that the Extreme Gradient Boosting algorithm obtained the highest accuracy and recall of 99% and 100%, respectively, but the other classic machine learning algorithms such as Decision Trees and Support Vector Machines also showed promising results. The analysis showed that the set of sensors used is highly selective for acetone, also in variable relative humidities that characterize human breathing. According to research, the humidity of the exhaled air fluctuates between 50-90%. The prepared e-nose system is a promising introduction to the prepar tion of a non-invasive device for measuring blood sugar levels