Espectroscopia molecular acoplada com algoritmos de aprendizado de máquina em saliva: uma rápida e não-invasiva ferramenta de triagem diagnóstica para amelogênese imperfeita

• Ano de defesa: 2022
• Autor(a) principal: Felipe Morando Avelar
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; Programa de Pós-Graduação em Inovação Tecnológica e Propriedade Intelectual; UFMG
• Palavras-chave em Português: Amelogênese Imperfeita; Saliva; Espectroscopia de reflexão total atenuada no infravermelho com transformada de Fourier - ATR­FTIR; Biomarcadores; Amelogenesis imperfecta; ATR­FTIR; Biomarkers; Espectroscopia de infravermelho por transformada de Fourier FTIR; Inovações tecnológicas; Espectroscopia de infravermelho por transformada de Fourier; Amelogênese imperfeita
• Link de acesso: http://hdl.handle.net/1843/69589

Amelogenesis Imperfecta is a rare disease of the dental enamel that affects both primary and permanent dentition with the same severity, causing masticatory and self-esteem problems and demanding complex and prolonged treatment. Successful treatment depends on early detection and intervention. A series of innovative salivary diagnostic platforms with lower waste production, lower cost, and non-invasive and reagent-free use have potential applications for oral and systemic diseases. In this study, we evaluate the potential of ATR-FTIR associated with artificial intelligence algorithms to discriminate amelogenesis imperfecta from matched healthy controls. The pilot study included 06 patients with Amelogenesis Imperfecta and 06 controls matched by gender and age, previously healthy and with no history of any type of dental enamel alteration. Whole unstimulated saliva was collected in sterile polypropylene pots at least 1 hour after feeding or tooth brushing. Subsequently, they were stored and frozen at -80°C. The present study aimed to compare salivary vibrational modes between AI patients and matched controls using ATR-FTIR spectroscopy coupled with linear discriminant analysis (LDA), Random forest and support vector machine (SVM) algorithms. Classification of salivary infrared spectra by LDA showed a sensitivity of 83%, specificity of 67%, and accuracy of 75% between AI and matched controls. The SVM algorithm discriminates AI more than matched controls with 100% sensitivity, 83% specificity, and 92% accuracy. The spectral area between 1300 cm-1 to 1200 cm-1 can be considered a pre-validated salivary infrared spectral area as a potential biomarker for AI screening. In summary, ATR-FTIR spectroscopy coupled with machine learning algorithms can be used on saliva samples to discriminate AI and can be further explored as an additional screening tool for AI in dental settings.