Desenvolvimento de plataforma biofotônica para triagem diangóstica da sepse pela saliva baseada em algoritmos de inteligência artificial

• Ano de defesa: 2022
• Autor(a) principal: Moura, Douglas Vieira de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Odontologia
• Palavras-chave em Português: Saliva; Sepsis; Diagnosis; ATR-FTIR; Artificial Intelligence; Sepse; Diagnóstico; ATR­FTIR; Inteligência Artificial; CNPQ::CIENCIAS DA SAUDE::ODONTOLOGIA; Odontologia; Septicemia - Diagnóstico; Saliva - Doenças - Diagnóstico; Inteligência artificial - Aplicações médicas
• Link de acesso: https://repositorio.ufu.br/handle/123456789/38864; http://doi.org/10.14393/ufu.di.2022.653

Sepsis is characterized as a clinical syndrome resulting from a systemic inflammatory response due to a focus of infection. The hypothesis of this study was that saliva can be used as a diagnostic biofluid associated with a sustainable and rapid biophotonic platform for the diagnosis of sepsis. The aim of present study was to develop a diagnostic platform using Fourier transformed infrared spectrometry with total saliva reflectance (ATR­FTIR) associated with univariate and multivariate equipment and artificial intelligence. Materials and Methods: Wistar rats (~260g) were divided into control (n=7) and sepsis (n=7). The sepsis group underwent cecal ligation and puncture surgery (CLP) and controls underwent SHAM surgery. After 24 hours, with the animal anesthetized, saliva was collected for 7 minutes and analyzed in the ATR­FTIR. A clear separation between controls and sepsis was observed through principal component analysis (PCA). The sum of PC­1 and PC­2 was responsible for 95.4% of the total explained variance between samples. The vibrational mode of CH2 of lipids was reduced in the sepsis group (2933cm­1) by second derivative analysis, showing possible alterations in lipid metabolism as a result of sepsis. Analysis by the Support Vector Machine (SVM) algorithm showed sensitivity of 0.72%, specificity of 100% and accuracy of 0.87%. Analysis using the Linear Discriminant Analysis (LDA) algorithm showed a sensitivity of 0.67%, specificity of 0.95% and accuracy of 0.82% Conclusion: Therefore, the use of the ATR­FTIR platform coupled with artificial intelligence algorithms can be an alternative tool for diagnostic screening of sepsis using saliva