Research Information System
Thesis Type: Postgraduate
Institution Of The Thesis: Gazi University, Fen Bilimleri Enstitüsü, -, Turkey
Approval Date: 2022
Thesis Language: Turkish
Student: MEHMET KEREM ÖZEZEN
Supervisor: Selim Acar
Open Archive Collection: AVESIS Open Access Collection
Abstract:
The olfactory diagnosis method, which has been used by mediciners for centuries in the field
of health, has become the focus of attention for engineering in recent years, and electronic
noses that are biologically inspired by the olfactory systems of animals have been developed.
In the literature, electronic noses have been proposed for a wide variety of fields, mainly in
health, food sector, industry and military sevices. However, the analysis of organic volatiles
in human breath remains pioneering work. In this study, concentrations of 𝐶3𝐻6𝑂 (acetone),
𝐶4𝐻8𝑂 (2-butanone) and 𝐶2𝐻6𝑆 (dimethyl sulfide) analytes in human breath were associated
with non-alcoholic fatty liver disease and liver cirrhosis, attempted to be detected with quartz
crystal microbalance sensors. Data sets were created for machine learning algorithms that
mathematically model the time-dependent frequency changes of sensors and the
performances of various algorithms were tested. According to the results, an ensemble model
was established with Logistic Regression, XGBoost, Support Vector Machines and Random
Forest algorithms with the highest score, and with this model the success rate, AUC: 0.97,
F1 Score: 0.87 for 𝐶3𝐻6𝑂, AUC: 0.97, F1 Score: 0.89 for 𝐶4𝐻8𝑂, and AUC: 0.99, F1 Score:
0.91 for 𝐶2𝐻6𝑆 was captured. In this study, the ppm values of the analytes were calculated
by a regression process with RMSE: 2.396874, 2.853136 and 2.775164 error rates,
respectively, according to the classification results.
| E-Nose, QCM Sensors, Machine learning, NAFLD, Cirrhosis |