Date of Award

2023

Publication Type

Thesis

Degree Name

M.Sc.

Department

Physics

Keywords

Artificial neural networks, Bacteria, Blood, Laser-induced breakdown spectroscopy, Machine learning, Urine

Supervisor

S.J.Rehse

Supervisor

T.J.Hammond

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract

The aim of this thesis is to expand on and improve the existing techniques used for detecting and identifying bacterial pathogens in clinical specimens with laser-induced breakdown spectroscopy (LIBS). Specifically, the existing experimental procedures, including bacterial sample preparation and data acquisition, as well as the data analysis with chemometric algorithms were investigated. Substantial reductions in LIBS background signal were achieved by implementing rigorous cleaning steps and the introduction of the use of ultrapure water. Following this, a database of LIBS spectra was acquired from specimens of E. coli, S. aureus, E. cloacae, M. smegmatis, and P. aeruginosa. The use of both discriminant function analysis (DFA) and partial least squares discriminant analysis (PLSDA) were compared. A PLSDA model built using the sum of all spectra acquired from 21 filters of E. coli and deionized water resulted in a sensitivity and specificity of 100% and 100%, respectively, in an external validation. To optimize the classification accuracy of the single-shot spectra for E. coli, E. cloacae, and S. aureus, outlier rejection schemes and data pre-processing methods were investigated. Classification errors of 30% motivated the use of artificial neural network analysis with principle component analysis pre-processing (PCA-ANN). The average sensitivity and specificity obtained using a randomized 80:20 split validation of the data was 94.18% and 97.01%, respectively. External validation was done on 52 filters of E. coli, E. cloacae, and S. aureus, giving an average sensitivity of 65.7%, and on 49 filters of E. coli, S. aureus, and M. smegmatis giving an average sensitivity of 87.2%.

Samples of blood and urine were obtained from a hospital and spiked with the same species listed above. 98.9% sensitivity and 100% specificity were achieved for detection of bacteria in urine. 96.3% sensitivity and 98.6% specificity were achieved for detection of bacteria in blood. Discrimination using PCA-ANN on species in urine using an 80:20 split resulted in an average sensitivity and specificity of 97.2% and 98.6%, respectively. External validation on 16 filters gave an average sensitivity 77.5%. Applying PCA-ANN using an 80:20 split on species in blood resulted in 100% sensitivity and specificity. External validation on 19 filters gave an average sensitivity of 82.3%. These results indicate the potential usefulness of LIBS in the clinical setting.

Share

COinS