NICeSim: um simulador interativo baseado em técnicas de aprendizado de máquina para avaliação de recém-nascidos prematuros em UTI neonatal

• Ano de defesa: 2014
• Autor(a) principal: Ferreira, Tiago Geraldo
• Orientador(a): Cerqueira, Fábio Ribeiro
• Banca de defesa: Siqueira-batista, Rodrigo
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Ciência da Computação
• Departamento: Metodologias e técnicas da Computação; Sistemas de Computação
• País: BR
• Palavras-chave em Português: Aprendizado do computador; Inteligência artificial; Bioinformática; Simulação computacional
• Palavras-chave em Inglês: Learning the computer; Artificial intelligence; Bioinformatics; Computer simulation
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://locus.ufv.br/handle/123456789/2673

This work describes the NICeSim, a freely available and open source simulator that uses machine learning techniques (ML) to assist health professionals in obtaining a bet- ter assessment in terms of prognosis of preterm infants in neonatal intensive care units (neonatal ICUs). The application was developed and tested using a database collected in a public teaching hospital located in the city of Viçosa, Minas Gerais. The available data were used to feed a pipeline of ML that was designed to create a simulator able to predict the probability of death for newborns admitted in neonatal ICUs. Here, we present the techniques used as base for the development of the simulator, artificial neural network and support vector machines algorithms were used as the machine learn engines of the application. We ll also discuss some results from statistical and practical tests using the data set mentioned above. Our statistical experiments showed that the resulting model to predict death achieved an accuracy of 86.7% in best case scenarios. This significant accuracy demonstrates that NICeSim can be used for hypothesis tes- ting. In fact, in an experiment conducted by two doctors, three main attributes were evaluated to understand how they affect the risk of death. The results showed that the model provides predictions that are in good agreement with the literature, showing that NICeSim can be an important tool to support decision making in clinical practice. More than that, the method can be used as a template for the creation of similar solu- tions to other computing scenarios of interest, even if those belong to different problem domains, given that appropriate data is provided.