Loading...
Publication
Sistema Point of Care baseado em espetroscopia Raman para a deteção de doenças renais
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.26 MB | Adobe PDF |
Abstract(s)
Este trabalho insere-se no projeto LUMINA, cujo objetivo futuro é o desenvolvimento de um sistema portátil para determinar o estado renal do paciente através da análise de urina. Este sistema terá três abordagens de deteção, a presente dissertação concentra-se exclusivamente no desenvolvimento da parte da espetroscopia Raman. O sistema Raman foi construído com base na metodologia de desenvolvimento da Starter Edition, sistema projetado pelos engenheiros responsáveis do site OpenRAMAN. A otimização do sistema foi dividida em duas partes: otimização da fonte de radiação, o laser, e otimização dos parâmetros de aquisição do software utilizado para obter os espetros Raman. Numa primeira fase, determinou-se a Largura à Meia Altura do pico do laser a diferentes temperaturas de funcionamento, de forma a selecionar a temperatura que minimizava este parâmetro. Para determinar o impacto que a temperatura de funcionamento do laser tem no espetro Raman, adquiriu-se o espetro do etanol a diferentes temperaturas. Com esta análise concluiu-se que a temperatura de funcionamento do laser que o otimiza é 30ºC. Na segunda fase de otimização, obteve-se o espetro do etanol ajustando os parâmetros de aquisição do software utilizado, com o intuito de identificar quais os parâmetros que mais significativamente impactam a qualidade do espetro final. Observou-se que estes parâmetros são o tempo de aquisição e o número de imagens utilizadas. Para validar o sistema Raman desenvolvido, adquiriu-se 5 espetros de diferentes amostras de urina. Com esta validação, concluiu-se que se o sistema desenvolvido é consistente nas suas medições e sensível às variações de concentração dos compostos presentes na urina. Por último, desenvolveu-se uma rede de aprendizagem automática para a classificação de espetros Raman. O treino e teste da rede foram efetuados com soluções com várias concentrações de álcool. A otimização da rede foi efetuada variando os seus hiperparâmetros e determinando a sua precisão, exatidão e tempo de treino.
Abstract This work is part of the LUMINA project, whose future goal is the development of a portable optoelectronic system to determine the patient's renal status through urine analysis. This system will have three detection approaches, the present dissertation focuses exclusively on the development of the Raman spectroscopy part. The Raman system was built based on the development methodology of the Starter Edition, a system designed by the engineers responsible for the OpenRAMAN website. The optimization of the system was divided into two parts: optimization of the radiation source, the laser, and optimization of the acquisition parameters of the software used to obtain the Raman spectra. In the first phase, the Width at Mid-Height of the laser peak at different operating temperatures was determined, in order to select the temperature that minimized this parameter. To determine the impact that laser operating temperature has on the Raman spectrum, the ethanol spectrum was acquired at different temperatures. With this analysis, it was concluded that the operating temperature of the laser that optimizes its operation is 30ºC. In the second phase of optimization, the ethanol spectrum was obtained by adjusting the acquisition parameters of the software used, to identify which parameters most significantly impact the quality of the final spectrum. It was observed that these parameters are the acquisition time and the number of images used. To validate the developed Raman system, 5 spectra of different urine samples were acquired. With this validation, it was concluded that the developed system is consistent in its measurements and sensitive to variations in the concentration of compounds present in urine. Finally, a machine learning network for Raman spectrum classification was developed. The training and testing of the network was carried out with solutions with various concentrations of alcohol. The optimization of the network was carried out by varying its hyperparameters and determining its precision, accuracy and training time.
Abstract This work is part of the LUMINA project, whose future goal is the development of a portable optoelectronic system to determine the patient's renal status through urine analysis. This system will have three detection approaches, the present dissertation focuses exclusively on the development of the Raman spectroscopy part. The Raman system was built based on the development methodology of the Starter Edition, a system designed by the engineers responsible for the OpenRAMAN website. The optimization of the system was divided into two parts: optimization of the radiation source, the laser, and optimization of the acquisition parameters of the software used to obtain the Raman spectra. In the first phase, the Width at Mid-Height of the laser peak at different operating temperatures was determined, in order to select the temperature that minimized this parameter. To determine the impact that laser operating temperature has on the Raman spectrum, the ethanol spectrum was acquired at different temperatures. With this analysis, it was concluded that the operating temperature of the laser that optimizes its operation is 30ºC. In the second phase of optimization, the ethanol spectrum was obtained by adjusting the acquisition parameters of the software used, to identify which parameters most significantly impact the quality of the final spectrum. It was observed that these parameters are the acquisition time and the number of images used. To validate the developed Raman system, 5 spectra of different urine samples were acquired. With this validation, it was concluded that the developed system is consistent in its measurements and sensitive to variations in the concentration of compounds present in urine. Finally, a machine learning network for Raman spectrum classification was developed. The training and testing of the network was carried out with solutions with various concentrations of alcohol. The optimization of the network was carried out by varying its hyperparameters and determining its precision, accuracy and training time.
Description
Dissertação para obtenção do grau de Mestre em Engenharia Biomédica
Keywords
Espetroscopia Raman Sensor Otimização Instrumentação Diagnóstico Doença renal Raman spectroscopy optimization Instrumentation Diagnosis Kidney disease
Citation
DOMINGOS, Catarina Filipa da Silva – Sistema Point of Care baseado em espetroscopia Raman para a deteção de doenças renais. Lisboa: Instituto Superior de Engenharia de Lisboa. 2024. Dissertação de Mestrado.