SVM-based screening tool for assessing increased risk of ischemic stroke

In recent years, increasing attention has been paid to the role of bacterial infections in systemic diseases. Among these, Streptococcus mutans, an oral pathogen which is the main cause of dental caries, has been linked to an increased risk of ischemic strokes when the Cnm-positive (Cnm+) mutation is present. Therefore, the ability to distinguish between Cnm-negative and Cnm-positive strains of Streptococcus mutans is essential for accurate risk assessment of ischemic stroke. Raman spectroscopy can be used to distinguish between the Cnm-positive and Cnm-negative variants of this bacterium thanks to the presence of sulfur exclusive to the Cnm+ ones, which is linked to an easily distinguishable peak in Raman spectra. However this method, while faster and cheaper than the more standard PCR, requires the presence of an expert who can interpret the spectra and it is thus not easily applicable on a large scale. As an alternative, this thesis aims to combine Raman spectroscopy with machine learning techniques, using the spectra as input data for a radial basis function support vector machine. This approach enables us to automatically classify S. mutans strains into Cnm-positive and Cnm-negative categories, thus reducing the need for expert interpretation. This thesis is based on the spectral data collected from 30 patients, 4 of which are infected with the mutated strain. Given the limited amount of data, this work serves as a proof of concept, illustrating how applied mathematics and machine learning can contribute to developing practical screening tools in biomedicine.