Session S22.3

Novel Feature for Quantifying Temporal Variability of Poincaré Plot: A Case Study

C Karmakar*, J Gubbi, A Khandoker, M Palaniswami

The University of Melbourne
Melbourne, Australia

The Poincaré plot of RR intervals is one of the most popular techniques used in heart rate variability (HRV) analysis. Initially it was extensively used for qualitative analysis but later two standard descriptors SD1 and SD2 were proposed to quantify the variability of the heart rate. These descriptors represent variability perpendicular to the line of identity and along the line of identity respectively. Fundamentally, SD1 and SD2 of Poincaré plot are directly related to basic statistical measures: standard deviation of RR interval (SDRR) and standard deviation of the successive difference of RR interval (SDSD). This implies that SD1 and SD2 are not affected by the temporal variations of the signal and represents the distribution of signal in a 2D space by quantifying spatial (shape) information. It should be noted that many possible RR interval series results in identical plots with exactly similar SD1 and SD2 values in spite of different underlying temporal structure. The present study proposes a novel descriptor to discriminate Poincaré plots having similar 2D shape with different temporal structures. Moreover, the performance of new descriptor is presented and compared with standard descriptors using real world case study. In this study, we have used the novel descriptor called as Complex Correlation Measure (CCM) to quantify the temporal information of a Poincaré plot. To compare performance of CCM with standard Poincaré descriptor SD1 and SD2, we have calculated ROC area for each descriptor between Normal Sinus Rhythm (NSR) and Congestive Heart Failure (CHF) subjects. The RR intervals of 54 NSR subjects and 29 CHF subjects from Physionet NSR and CHF database are used. The (mean ± sd) values of SD1, SD2 and CCM for NSR subjects are found as (0.03 ± 0.02), (0.19 ± 0.04) and (0.05 ± 0.03) respectively. Similarly for CHF subjects these values are (0.04 ± 0.02), (0.11 ± 0.06) and (0.14 ± 0.06). The p value obtained from chi-square analysis between two groups for SD1 and SD2 are found insignificant whereas for CCM it is significant with p-value of 9.07E-14. The largest ROC area between two groups was for CCM (0.92) compared to SD1 (0.71) and SD2 (0.90). Results indicate that CCM can be used as a significant feature for detecting pathology.

(Abstract Control Number: 154)