Publicação
Evaluation of spectroscopic-based methodologies for early detection of urinary tract infection
| Resumo: | Despite all the remarkable advances in healthcare, healthcare-associated infections (HCAI) are still a critical public health problem, with 30 to 40% of infections related to the urinary tract system. These urinary tract infections (UTIs) are considered one of the most common bacterial infections in hospital settings and everyday community context, where about 80% are highly correlated with urinary catheter insertion - Catheter-associated urinary tract infection (CAUTI). Considering that 15 to 25% of hospitalised patients need to be catheterised during their treatments and most CAUTIs are asymptomatic, it results in a tremendous challenge to early diagnosis CAUTI, therefore, to initiate its treatment. The present work aimed at exploring the potential of absorption and fluorescence spectroscopic methodologies to detect UTIs. The urine samples were used without any previous treatment to target the most straightforward testing protocol possible. The emission spectrum with excitation fixed at 280 nm, when combined with the transmittance value at 600 nm, was found to be a valid and powerful methodology to distinguish healthy from unhealthy (with UTI) samples. In addition, it was developed an interactive application based on machine learning algorithms capable of evaluating the data to identify, autonomously, if the sample is healthy or not. Multiple classifiers were optimised and trained with a previously selected ideal subset of features to achieve this goal. The best classifier performance was the K-Nearest Neighbour (KNN) with a confusion matrix accuracy and area under curve (AUC) value of 0.902 and 0.750, respectively, indicating a good capability of discriminating the two defined groups of samples. In conclusion, the results presented in this work indicate the potential of urine’s fluorescence and transmittance spectra to be developed as a simple and rapid diagnostic tool. However, further exploration must be given to improve the studied methodology and the ML classifier accuracy from the input data. |
|---|---|
| Autores principais: | Mendes, Ana Filipa Neves de Sousa |
| Assunto: | Infeções do trato urinário Testes de diagnóstico Espectroscopia de Fluorescência Transmitância Aprendizagem automática Teses de mestrado - 2023 |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório da Universidade de Lisboa |
| Resumo: | Despite all the remarkable advances in healthcare, healthcare-associated infections (HCAI) are still a critical public health problem, with 30 to 40% of infections related to the urinary tract system. These urinary tract infections (UTIs) are considered one of the most common bacterial infections in hospital settings and everyday community context, where about 80% are highly correlated with urinary catheter insertion - Catheter-associated urinary tract infection (CAUTI). Considering that 15 to 25% of hospitalised patients need to be catheterised during their treatments and most CAUTIs are asymptomatic, it results in a tremendous challenge to early diagnosis CAUTI, therefore, to initiate its treatment. The present work aimed at exploring the potential of absorption and fluorescence spectroscopic methodologies to detect UTIs. The urine samples were used without any previous treatment to target the most straightforward testing protocol possible. The emission spectrum with excitation fixed at 280 nm, when combined with the transmittance value at 600 nm, was found to be a valid and powerful methodology to distinguish healthy from unhealthy (with UTI) samples. In addition, it was developed an interactive application based on machine learning algorithms capable of evaluating the data to identify, autonomously, if the sample is healthy or not. Multiple classifiers were optimised and trained with a previously selected ideal subset of features to achieve this goal. The best classifier performance was the K-Nearest Neighbour (KNN) with a confusion matrix accuracy and area under curve (AUC) value of 0.902 and 0.750, respectively, indicating a good capability of discriminating the two defined groups of samples. In conclusion, the results presented in this work indicate the potential of urine’s fluorescence and transmittance spectra to be developed as a simple and rapid diagnostic tool. However, further exploration must be given to improve the studied methodology and the ML classifier accuracy from the input data. |
|---|