Non-linear data mining methods to assess the impact of physical training on the cardiovascular system of subjects from different age groups

A set of non-linear data mining methods have been applied to ECG signals and other cardiovascular and blood parameters to evaluate the cardiovascular response to exercise in young and master athletes, compared with control groups of untrained subjects of the same age. Methods include PNN calculation, Multiscale Entropy analysis (MSE), and a comparison between clustering and an Artificial Neural Network analysis performed by means of chaotic attractors. After recruiting four groups of healthy athletes and sedentary subjects, with age under and over 40, we analyzed the collected data, obtaining cross-validated classifications and significant variable differences among clusters. The analyses lead to a good stratification of the subjects, establishing some important relationships between physical activity, age, sex, and cardiovascular parameters. In particular the existence of significant differences in the cardiovascular status of hese groups was shown, depending in particular on the MSE1, PNN20, VO and FC variables. This will make it possible a follow-up of the subjects, analyzing the above specified parameters over time, in order to identify possible markers of increased arrhythmic risk, useful to prevent fatal cardiac events.