THZ SPECTROSCOPY OF EXHALED AIR FROM DIABETES MELLITUS

THZ SPECTROSCOPY OF EXHALED AIR FROM DIABETES MELLITUS PATIENTS Alexey V. Borisov, Tomsk State University, Tomsk, Russia, borisov@phys. tsu. ru Yu. V. Kistenev, 1, 2 Teteneva A. V. , 2 Sorokina T. V. , 2 A. I. Knyazkova, 1, 3 O. A. Zakharova 1, 3, E. S. Sim. 1, 2 1 National Research Tomsk State University, Lenin Ave. 36, 634050, Tomsk, Russia 2 Siberian State Medical University, Moskovskiy trakt, 2, 634050, Tomsk, Russia 3 Institute of Strength Physics and Materials Science SB RAS, Academichesky Ave. , 2/4, 634021 Saratov, September, 2019 Acknowledgements. This work was performed within the frame of the Fundamental Research Program of the State Academies of Sciences for 2013 -2020, line of research III. 23, and with partial financial support from the Russian Foundation for Basic Research (grant No. 17 -00 -00186).

The foundation of the approach Diabetes mellitus is a serious medico-social problem which constantly gaining momentum because of its prevalence. This disease, in importance, stands immediately after cardiac and oncological diseases. v Worldwide, more than 60 million people in the world suffer from diabetes, of which 4. 5 million in Russia. v Countries with the highest number of patients with diabetes: China, India, USA, Brazil, Russia. v The incidence rate increases by 6 -10% annually and doubles every 10 -15 years. 2

Diagnostic methods for diabetes: v v Venous plasma glucose test; TSH (with a clarification of the diagnosis); Hb. A 1 c hemoglobin level measurement (according to NGSP or IFCC standards); Determination of markers of autoimmune destruction of pancreatic β-cells (determining the risk of developing diabetes): v Genetic studies (HLA DR 3, DR 4 and DQ). Gas analysis in the diagnosis of diseases The main difficulty is the unambiguous relationship between specific biomarkers and their concentrations in exhaled air samples (EAS) with diseases. The solution is to register the marker profile (metabolic profile) and compare the results for different groups. Target group: 20 patients with diabetes Control group: 20 healthy volunteers 3

The equipment Time-domain THz spectrometer T-SPEC (EKSPLA, Estonia) Main parameters: - Tuning range 0. 2 -3. 5 THz - Dynamic range > 60 d. B 4

Data preprocessing methods and Source Data Issues THz spectra of diabetes 1. 2. 3. 4. 5. 6. 7. Exhaled samples Lifetime Emissions Noises Credibility Accuracy Range selection THz spectra of healthy 5

Data mining methods Selection of the informative features: the key idea of Principal Component Analysis Here is the matrix of accounts, is the matrix of loads, - error matrix 6

Data mining methods Classification: the key idea of Support Vector Machine - some nonempty set - the number of objects in the training set - output - classification objects 7

Data mining Mean norm of THz absorption spectra of diabetes and healthy PCA analysis SVM classification Binary classification Diabetes/ Healthy Sensitivity Specificity Mean Dispersion 0. 87 0. 05 0. 91 0. 04 8

Conclusion v A technique for sampling and analyzing EAS has been developed and its applicability in terahertz spectroscopy has been proved. v EAS of healthy volunteers and people with diagnosed diabetes were obtained. v From the analysis of spectral data, dependencies are obtained that characterize the relationship between clinical parameters and profiles of THz spectra of EAS. It was shown that THz EAS spectroscopy is a promising field for the diagnosis and non -invasive studies of diabetes mellitus, which correlate with clinical parameters. Acknowledgements. This work was performed within the frame of the Fundamental Research Program of the State Academies of Sciences for 2013 -2020, line of research III. 23, and with partial financial support from the Russian Foundation for Basic Research (grant No. 17 -00 -00186). 9

Thank you!