INTELIGÊNCIA ARTIFICIAL NA DETERMINAÇÃO DA CARGA DE TRABALHO NO TESTE DE ENDURANCE NA DOENÇA PULMONAR OBSTRUTIVA CRÔNICA: UM ESTUDO RETROSPECTIVO ANALÍTICO

• Ano de defesa: 2022
• Autor(a) principal: Fernanda Gabriela Dias
• Orientador(a): Paulo de Tarso Guerrero Muller
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Fundação Universidade Federal de Mato Grosso do Sul
• País: Brasil
• Palavras-chave em Português: DPOC; Teste de endurance; Taxa de Trabalho; Teste Cardiopulmonar ao exercício.
• Link de acesso: https://repositorio.ufms.br/handle/123456789/4791

The Endurance Test is an extremely important clinical and scientific instrument to assess the maximum time of tolerance to moderate/intense exercise in patients with Chronic Obstructive Pulmonary Disease (COPD). Recommendations from different Pulmonology and Cardiology societies guide the use of high intensity in constant load tests (CWT, 75-80% of the maximum load reached in the incremental test). However, when this load is applied, a large portion of the evaluated (~ 50%) do not fit the ideal time of exercise in CWT, that is, between 3-8 minutes, leading the evaluated ones to need to repeat the tests, with greater or lesser load. In this work, an artificial intelligence (AI) algorithm called M5P was tested, aiming to predict the ideal individualized load to be applied in a cycle ergometer, to reach the ideal time of 3-8 min. Accordingly, anthropometric and clinical data, pulmonary function data and physiological variables from the incremental test of 50 individuals with COPD GOLD II/III/IV in outpatient follow-up were retrospectively analyzed. When submitted to the M5P algorithm, these data, through a decision tree, found 2 models capable of estimating the individualized ideal load. Thus, the mean absolute error for Model 1 was 4.4 Watts, with a coefficient of determination of 79% and a 95% relative confidence interval of -36/+34%. The Model 2 performed worse, with a mean absolute error of 7.4 Watts, with a coefficient of determination of 36% and a 95% relative confidence interval of -40/+36%. For Model 1, only the maximum load in the incremental test was the chosen factor. For Model 2, various combinations of clinical, pulmonary function, and incremental testing variables entered the model in 3 different equations. The model chosen in this study was Model 1, which, however, showed a wide confidence interval. Further studies with larger samples and validation of Model 1 with individuals with COPD are necessary.