Wearables para a monitorização contínua da saúde cardiorrespiratória

Abstract

Background: The cardiovascular health (CH) or aerobic power can be represented by maximal the maximum oxygen uptake (VO2max) measured during the cardiopulmonary exercise testing (CPET). However, despite the extensive clinical relevance of CH evaluation, the CPET could be interrupted before obtaining the true VO2max due to many different reasons such as the lack of motivation, in addition, it is performed only occasionally and involving the use of medical equipment that are not available to the general population and in realistic settings. Thus, this thesis is composed of 3 studies, as shown below: Study 1- Aim: To verify if supra maximum exercise testing (SMET) is reliable to confirm the VO2max achieved during the CPET, in addition to using different equipment to investigate the integrated physiological responses during peak CPET and SMET. Methods: 74 volunteers of both sexes were evaluated, aged between 19 and 72 years, and with different levels of maximum aerobic power. They performed the CPET with incremental protocol followed by SMET with constant load at 110% of the peak workload of the CPET, both tests were performed on a cycle ergometer and interrupted by maximum volitional fatigue. During both tests the metabolic and ventilatory variables were collected, by the metabolic system, in addition, the cardiovascular variables were measured, by an electrocardiography and photoplethysmography system. Results: There is no significant difference between VO2max in CPET and SMET. While there were statically significant differences between the metabolic, ventilatory, and cardiovascular variables during the peak of CPET and SMET. Conclusion: The SMET at 110% of CPET workload should be reliable to confirm true VO2max reached during CPET, in just one day, with different integrated responses of metabolic, ventilatory, and cardiovascular variables. Study 2- Aim: To evaluate the prediction of CH using artificial intelligence (AI) methods from data obtained by wearable technologies, in addition to using an explainable method of AI to investigate how CH can be estimated from the longitudinal signals acquired by wearables during activities of daily living (ADLs). Methods: 43 volunteers of both sexes were evaluated, aged between 19 and 72 years, and with different levels of aerobic power. The first stage consisted of the evaluation of the CH by CPET with incremental protocol. The second stage was performed in non-supervised environments to collect the longitudinal data (for 7 days) with a smart shirt with three embedded sensors. Finally, the last stage was the training of AI algorithms to predict the CH in addition to the use of an explanatory method. Results: There is a positive, strong and statistically significant correlation between the VO2max measured during CPET and that predicted by AI techniques. The most important inputs were age, heart rate, hip acceleration, weight and height (considering the Hemodynamic and Anthropometric domains) used to predict VO2max, using the explainable models. Conclusion: The VO2max can be predicted by wearable technologies associated with AI, by a support vector machine. Explanatory models were used to extract clinical insights from these predictions. Study 3- Aim: To verify the correlation between the biological and accelerometry variables from a wearable system during different intensities and the CH, in addition, to verifying the differences in biological responses in front of the different intensities of ADLs. Methods: 43 volunteers of both sexes, aged between 19 and 72 years, and with different levels of aerobic power The stages 1 and 2 were conducted as the same as described before in Study 2. Finally, the last stage was the stratification of three free ADLs intensities: low, moderate, and vigorous. Results: Biological and accelerometry data during unsupervised vigorous-intensity ADLs correlated with CH. There are statistically significant differences between the intensities for these variables. Conclusion: The biological and environmental data from a wearable system is related to CH during unsupervised vigorous-intensity ADLs. General Conclusions: Therefore, it was possible to confirm the true CH in a single laboratory visit by an SMET at 110% of the CPET workload. Furthermore, it was possible to predict the CH using two AI models and with an explanatory method. In addition, the CH was correlated with vigorous-intensity ADLs. All these findings were performed in a population with a broad spectrum of maximum aerobic power (VO2max).