Assessing the human cardiovascular response to moderate exercise: feature extraction by support vector regression.

This study aims to quantitatively describe the steady-state relationships among percentage changes in key central cardiovascular variables (i.e. stroke volume, heart rate (HR), total peripheral resistance and cardiac output), measured using non-invasive means, in response to moderate exercise, and the oxygen uptake rate, using a new nonlinear regression approach-support vector regression. Ten untrained normal males exercised in an upright position on an electronically braked cycle ergometer with constant workloads ranging from 25 W to 125 W. Throughout the experiment, VO(2) was determined breath by breath and the HR was monitored beat by beat. During the last minute of each exercise session, the cardiac output was measured beat by beat using a novel non-invasive ultrasound-based device and blood pressure was measured using a tonometric measurement device. Based on the analysis of experimental data, nonlinear steady-state relationships between key central cardiovascular variables and VO(2) were qualitatively observed except for the HR which increased linearly as a function of increasing VO(2). Quantitative descriptions of these complex nonlinear behaviour were provided by nonparametric models which were obtained by using support vector regression.