Institutional Repository of Vilnius University
Mikroorganizmų identifikavimas Infraraudonosios ATR spektrometrijos metodais /
Abstract
dc:descriptionMicroorganism identification is a major issue in medicine and food safety. Although current biochemical methods of microorganism identification have excellent accuracy, they are rather expensive and require a lot of time and labour. Vibrational spectroscopy has been proposed as a cost-effective and label-free alternative method for identifying microorganisms. Because different species of microorganisms differ in their biochemical composition, they have distinct vibrational spectra that can be used to identify them. One of the obstacles for applying vibrational spectroscopy in this field are the natural variations in the vibrational spectra of microorganisms that make it harder to develop reliable classification models. The goal of this study is to find the best methods for measuring microorganism vibrational spectra and creating classification models for microorganism spectra. For testing spectra measurement techniques, vibrational spectra of two species of bacteria - Streptococcus pyogenes and Staphylococcus aureus - and two species of fungi - Candida guilliermondii and Candida parapsilosis – were measured using three different methods. These methods were: attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR), Fourier transform infrared micro spectroscopy and Fourier transform Raman scattering spectroscopy. Out of those methods ATR-FTIR was chosen as the most optimal, for its convenience and reproducibility. Classification models utilizing logistic regression and various spectra preprocessing techniques were tested using a data set comprised of ATR-FTIR spectra of 5 individual microorganisms and 3 mixtures of two microorganisms. It was noted that the normalization of the standard deviations of intensity intervals in the spectra data set resulted in a significant increase in the classification accuracy of the logistic regression model, going from 68.1 % for the model without any preprocessing to 80.6 %. Furthermore, spectra normalization based on the integral intensity of certain ranges of the spectra also improved the classification accuracy. Additionally, it was discovered that a novel application of a normalization algorithm used in X-ray absorption spectroscopy for preprocessing microorganism ATR IR spectra resulted in a consequential increase in the classification accuracy, outperforming normalization by integral intensity. The highest microorganism spectra identification model accuracy achieved in this study is 89.2 %, obtained by preprocessing spectra with X-ray absorption spectroscopy normalization, normalization of standard deviation of intensities and principal component analysis, then applying resulting data to a logistic regression model with L2 regularization.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Taraškevičius, Tautvydas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210578022&prefLang=en_US