Submitted by Cristiane Chim (cristiane.chim@ucpel.edu.br) on 2019-05-24T13:44:22Z No. of bitstreams: 1 Mateus Beck Fonseca.pdf: 917017 bytes, checksum: 8d4582361a9da2b48dd9938561cfe410 (MD5) Made available in DSpace on 2019-05-24T13:44:22Z (GMT). No. of bitstreams: 1 Mateus Beck Fonseca.pdf: 917017 bytes, checksum: 8d4582361a9da2b48dd9938561cfe410 (MD5) Previous issue date: 2019-02-28 Bipolar disorder and schizophrenia are disorders of difficult diagnosis and differentiation, and studies point to the alteration of levels of inflammatory biomarkers with the diagnosis of diseases. Artificial neural networks (ANNs) are computational tools of artificial intelligence for modeling based on biological neural systems, which use mathematical formulas mimicking neural behavior. The objective of this work is to propose a model of ANN to aid in the diagnosis of bipolar disorder and schizophrenia using biomarkers and simple characteristics of the sampled population. The method of analysis is ANN training applied to a free distribution database of the Stanley Neuropathology Consortium, which consists of inflammatory biomarkers and characteristics of the population with diagnoses of schizophrenia, bipolar disorder and a control (without mental disorders) group. The RNA training program used is OpenNN, and is also freely distributed. As a result, it is expected to train a ANN with more than 80% accuracy in the classification of bipolar disorder diagnoses, schizophrenia and control group. Bipolar disorder, major depression and schizophrenia are disorders of difficult diagnosis and differentiation. Studies indicate that altered levels of inflammatory and neurotrophic biomarkers may be associated with the diagnosis of these diseases. Artificial neural networks (ANN) are computational tools of artificial intelligence for modeling based on biological neural systems, which use mathematical formulas mimicking neural behavior. The objective of this work is to propose an ANN model to aid in the diagnosis of bipolar disorder, major depression and schizophrenia, using biomarkers and simple characteristics of the population sampled. The method of analysis for the first article is ANN training applied to a free distribution database of the Stanley Neuropathology Consortium, which consists of inflammatory biomarkers and characteristics of the population with diagnoses of schizophrenia, bipolar disorder and one control group (without disorders); for the second article, another database was used, with biochemical variables, population characteristics and questionnaire responses with diagnoses of major depression, bipolar disorder and a control group (without disorders). The RNA training program used is OpenNN, and it is also freely distributed. As a result, trained RNAs with more than 80 % accuracy in diagnostic classifications O transtorno bipolar, a depressão maior e a esquizofrenia são transtornos de difícil diagnóstico e diferenciação. Estudos apontam que a alteração de níveis de biomarcadores inflamatórios e neurotróficos podem estar associados com o diagnóstico dessas doenças. Redes neurais artificiais (RNA) são ferramentas computacionais de inteligência artificial para modelagem baseadas em sistemas neurais biológicos, as quais utilizam fórmulas matemáticas mimetizando o comportamento neural. O objetivo deste trabalho é propor um modelo de RNA para auxiliar no diagnóstico de transtorno bipolar, da depressão maior e da esquizofrenia, utilizando biomarcadores e características simples da população amostrada. O método de análise para o primeiro artigo é o treinamento de RNA aplicada à um banco de dados de distribuição livre da Stanley Neuropathology Consortium, o qual consiste de biomarcadores inflamatórios e características da população com diagnósticos de esquizofrenia, transtorno bipolar e um grupo controle (sem transtornos); para o segundo artigo utilizou-se outro banco de dados, com variáveis bioquímicas, características da população e respostas de questionários com diagnósticos de depressão maior, transtorno bipolar e um grupo controle (sem transtornos). O programa de treinamento da RNA utilizado é o OpenNN, e também é de distribuição livre. Como resultado tem-se RNAs treinadas com mais de 80% de acurácia nas classificações dos diagnósticos