CC BY
12B-I-25
BIOMEDICAL PHOTONICS
Breathomics using laser spectroscopy and machine learning
Yury V. Kisteneva , Alexey V. Borisova, Viktor E.Skibaa, Igor K. Lednevb, Han Jinc' d
aTomsk State University, Tomsk 634050, Russian Federation bUniversity at Albany, SUNY, Albany, NY 12222, USA cInstitute of Micro-Nano Science and Technology, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, P. R. China dNational Engineering Research Center for Nanotechnology, Shanghai, 200241, P. R. China
Yv. kistenev@gmail.com
Analysis of volatile molecular biomarkers in the exhaled air, called breathomics, is suitable for operative non-invasive medical screening tests. The report is devoted to applications of laser spectroscopy and machine learning for evaluation of volatile molecular biomarkers' profile to detect a specific disease. Breath air analysis can be conducted through the chemical-composition-based and pattern-recognition-based approaches. For the former approach implementation, we use deep neural networks [1] and original chemometrics' methods: (a) a combination of the standard addition method with multivariate curve resolution called HAMAND [2]; (b) criterium based on reducing a spectrum complexity (RSC) [3] to provide exhaled air chemical composition. The latter approach is typical for supervised machine learning algorithms. We will compare both approaches.
We also presented results of absorption spectra resolution improving using computer super-resolution (SR) reconstruction, using several machine learning models based on different network architectures, including sequential ensemble architecture.
The research was carried out with the support of a grant under the Decree of the Government of the Russian Federation No. 220 of 09 April 2010 (Agreement No. 075-15-2021-615 of 04 June 2021).
[1] V. V. Prischepa, et al. Proc. SPIE 11582, 2020. P. 115821J; doi:10.1117/12.2581568
[2] M. Ando, I.K. Lednev, and H.-o Hamaguchi. In: Frontiers and Advances in Molecular Spectroscopy. 2018. P.369-378. doi:10.1016/B978-0-12-811220-5.00011-3
[3] A.Borisov, et al. Journal of Breath Research 2021. P.027104. doi:10.1088/1752-7163/abebd4
*
ALT'22
