Caracterização Fenotípica do Sistema Visual em Modelos Animais de Doenças do Desenvolvimento Neurológico

• Autor(a) principal: Vigo, Inês José Borges Pinto
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/21785
• topic: Autism; Neurofibromatosis Type 1; Electroretinogram; Classification; Statistical Analysis; Machine Learning; Python; Autismo; Neurofibromatose Tipo 1; Eletrorretinograma; Classificação; Análise Estatística; Aprendizagem de Máquina

Neurofibromatosis Type 1 (NF1) is an inherited disease associated with neurodevelopmental disorders, including Autism spectrum disorder (ASD). One of the systems affected by this disease is the visual system, although it is little studied and little taken into consideration in the diagnosis of NF1. It is in this sense that the present investigation arises, where potential changes in retinal physiology are studied in a genetic model of NF1, using a non-invasive neurophysiological technique - the Electroretinogram (ERG) - to determine its diagnostic potential. Thus, the main goal of this work was to find an accurate way to distinguish individuals with NF1 from healthy individuals. To this end, we evaluated possible features of the ERG signals, for both genotypes, that could demonstrate differences between them, and therefore be able to differentiate them. In general, the aim of this study was to find an algorithm capable of classifying an individual as sick or healthy with the greatest possible accuracy. To achieve this goal, a Python program was created that, using statistical analysis and Machine Learning (ML) techniques, was capable of returning a complete analysis of the data and the most relevant characteristics, as well as estimating the classification of the data in terms of accuracy value. With this methodology, it was possible to conclude that the use of ERG in the diagnosis of NF1 in an individual is feasible, since the statistical analysis found two features with significant p-values and the ML classification obtained interesting accuracy values