Gender and age classification using a new Poincare section-based feature set of ECG

Abstract

In the medicine, the implication of individual differences has frequently been emphasized. Gender- and age-related differences can be mentioned as the most important individual parameters. On the other hand, electrocardiogram (ECG) signals are the subject of these differences. However, limited information is available regarding these individual dissimilarities in ECG dynamics. This study was aimed to evaluate gender and age differences by means of novel Poincare section indices. Our focus was to detect and classify dynamical behaviors of the ECG trajectories using three binary classification strategies: (1) gender-, (2) age-, (3) gender- and age-based classification. After constructing the 2D phase space of ECG, linear Poincare sections in distinct angles were developed and some geometric indices were extracted. The effect of delayed phase space on ECG measures was also inspected. We tested our algorithm on 79 healthy subjects. Using support vector machine, the maximum correct rate of 93.33% was achieved for the gender- and age-based classification strategies. Considering the information of both age and gender, the highest rate was 94.66%. The best results were achieved with delays of 5 and 6. In conclusion, our results showed that basin geometry of the ECG phase states is affected by individual differences.