Background: The Poincare plot is a geometrical representation of an RR time series constructed by plotting successive RR intervals on a 2D phase space. In this article, the impact of age on the shape of Poincare plot of RR intervals and extracted features for quantification of this space is considered. Method: Fantasia database from Physionet databank is used in this paper. Two hours of ECG recording (sampling frequency: 250 Hz) for twenty young (21-34 years old) and twenty older adults (68-85 years old) subjects were used while all subjects remained in a resting state. After extraction of RR intervals from ECG, Poincare plot with 10 different lags (1-10) was constructed for each signal and five different features have been extracted in each lag of Poincare plots. The extracted features were standard descriptors of Poincare plot (SD1, SD2, and the SD1/SD2), the area of the estimated ellipse, and complex correlation measure (CCM). Results:
Extracted features from lagged Poincare plot were used as input to Multilayer Perceptron (MLP) neural network, SOM, and KNN classifier to discriminate two groups of young and older adults. The best result achieved was a sensitivity of 86.5%, specificity of 95.1%, and the accuracy of 91.4% using KNN.