Aprendizado de máquina e dispositivo vestível na identificação dos diferentes estágios da COVID-19 utilizando a variabilidade da frequência cardíaca.

• Ano de defesa: 2024
• Autor(a) principal: Sanches, Carlos Alberto
• Orientador(a): Librantz, Andre Felipe Henriques
• Banca de defesa: Librantz, Andre Felipe Henriques , Jorge, Luciana Maria Malosa Sampaio , Aletti, Federico , Belan, Peterson Adriano , Araújo, Sidnei Alves de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Nove de Julho
• Programa de Pós-Graduação: Programa de Pós-Graduação em Informática e Gestão do Conhecimento
• Departamento: Informática
• País: Brasil
• Palavras-chave em Português: aprendizado de máquina; coronavírus; COVID-19; dispositivos vestíveis; oxímetro; SARS-CoV-2; VFC; variabilidade da frequência cardíaca
• Palavras-chave em Inglês: coronavirus; COVID-19; HRV; heart rate variability; machine learning; oximeter; SARS-CoV-2; wearable devices
• Área do conhecimento CNPq: CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO
• Link de acesso: http://bibliotecatede.uninove.br/handle/tede/3505

The COVID-19 pandemic has required multidisciplinary efforts in its combat, including the development and distribution of diagnostic tests and vaccines. In 2023, the COVID-19 pandemic continued to pose a threat to global public health, with millions of people infected and thousands of deaths, in addition to a large number of individuals affected by long COVID syndrome. Although effective, mass testing remains complex and time-consuming. Recent studies have explored the relationship between heart rate variability (HRV) and COVID-19, using physiological data to retrospectively assess individual health. Motivated by the growing popularity of personal devices, such as smartwatches and smart bracelets, that monitor physiological data, this study investigated the feasibility of using these devices for real-time diagnosis of COVID-19. Machine learning techniques were employed to identify patterns in HRV indices for healthy individuals, COVID-positive individuals, and those with long COVID syndrome. Several algorithms were analyzed and tested, and the decision tree was selected as the main algorithm, achieving an overall average accuracy of 77%, improving to 96,7%, when information on recent COVID-19 infections were available. An automated oximeter was developed to collect and transmit HRV data for processing on a remote server. A web-based real-time monitoring system was designed to provide immediate diagnoses based on oximeter readings. This work is the first to offer real-time assessment and indication of the health status related to COVID-19 infection using physiological data. Furthermore, it introduces another innovative approach by providing, for the first time, a diagnosis of long COVID syndrome not based on clinical examinations. Real-time diagnosis can help prevent the spread of the disease and monitor its progression in individuals at higher risk of complications.