Fig. 1

Flowchart of correlation diagram of Mean-SD. ECG RRI was converted from an observation time series into a time delay coordinate system. From the ECG RRI (1), the instantaneous heart rate (2) was calculated, and the attractor trajectory (RRI attractor) in a three-dimensional delay coordinate system and its residual density function (histogram) were created (3). (A-3) was a torus attractor (○), and (B-3) became a strange attractor (○). The histogram of the torus attractor became a normal distribution. The histogram of the strange attractor became a superposition of negative kurtosis and normal distribution. (A) and (B) were low-dimensional chaos, with (A) showing quasiperiodic behavior, and (B) showing chaotic behavior. With the analysis results of the 225 cases of the 137 subjects, the correlation diagram of Mean-SD (4) was created. Setting a 45° line on the correlation diagram of Mean-SD as the criteria, it was divided into two clusters of (A) and (B). The Polynomial A of the correlation diagram of Mean-SD was expressed with a quartic function closer to a linear function, while the Polynomial B was a Duffing’s equation and expressed with a quadratic function having a non-linear dead zone and a range closer to a linear function. The existence of outliers or abnormal values is suggested in the instantaneous heart rate time series data, with attractor trajectories having a Mean-SD relationship existing in the range of the dead zone. The procedures for the detection of outliers by using the instantaneous heart rate time series data and of the abnormal value test using ECG, Lorenz plot and RRI attractors are explained in Fig. 2