Home > Research > Publications & Outputs > Identification and Prediction of Pathogenic Mic...

Electronic data

  • 2022MesaRamirezPhD

    Final published version, 24.7 MB, PDF document

    Embargo ends: 1/08/27

Text available via DOI:

View graph of relations

Identification and Prediction of Pathogenic Microorganisms with Vibrational Spectroscopy and Machine Learning

Research output: ThesisDoctoral Thesis

Publication date1/08/2022
Number of pages387
Awarding Institution
  • Lancaster University
<mark>Original language</mark>English


Identification pathogenic micro-organisms has time constraints , as it requires culturing of the bacteria which can take up to 48 hours . This time gap has a major impact on the health care system, making timely diagnosis of pathogenic microorganisms a significant challenge. . Current methods of microbial diagnosis are limited to molecular tests, such as biochemical assays, Polymerase Chain Reaction, antigen testing, etc. Recently, the use of artificial intelligence called machine learning (ML) has been explored to perform screenings in different areas of medicine. ML is carried out by using different algorithms; Support Vector Machines (SVM), Logistic Regression (LR), Partial Least Squares (PLS), among others. This study aims to explore the potential application of vibrational spectroscopy in conjunction with different ML models to screen for different bacteria, including S. aureus, S. epidermidis, K. pneumoniae, and P. aeruginosa, as well as the SARS-CoV-2 virus, which causes COVID-19 disease. These algorithms aim to detect patterns in the spectra that are fed into the model. Each ML model performs differently, and although they all aim to detect similarities between the data, each model clusters the output data differently, providing varying prediction results depending on the ML model used. In this study, it was possible to obtain specificities and sensitivities ranging from 90% to 100%, bringing into the spotlight the promising application of vibrational spectroscopy in conjunction with ML in the area of clinical diagnostics.
In addition to the application of ML models in the spectra, the different spectral bands that characterise each pathogenic microorganism used were identified in this study. On the other hand, it was also possible to evaluate the limit of detection of residual nasopharyngeal samples of COVID-19, as well as the temperature stability after 3 freeze-thaw cycles of samples with SARS-CoV-2 virus.