Session: 07-02-01: Vibration and Acoustics in Biomedical Applications

Paper Number: 165744

An Instantaneous Frequency Based Approach to Estimate Heart Rate and Calculate Heart Rate Variability Metrics 

Heart rate variability (HRV) is an emerging diagnostic tool utilized to assess cardiac and various physiological conditions. These metrics are derived from heart rate measurements obtained from signals that detect cardiac activity, such as the electrocardiogram (EKG) and photoplethysmography (PPG) signals. Conventional methods for calculating heart rate predominantly rely on peak detection algorithms, which present several inherent limitations. In EKGs, anomalies, irregularities, and noise in the signal make it challenging for peak detection algorithms to correctly identify R peaks. Further, the amplitude and prominence of R peaks greatly depend on the electrocardiogram lead being assessed. Consequently, some EKG leads are unsuitable for heart rate estimation using R peaks. All this could result in missed R peaks and erroneous R peak detection. Such errors lead to incorrect heart rate estimation and, subsequently, inaccurate HRV metrics. Furthermore, RR intervals (interval between two neighboring R peaks) are not constant in an EKG. An instantaneous heart rate (IHR) signal constructed using RR interval measurements, can therefore, be thought of as a signal sampled at a variable sampling rate. As a result, frequency domain analysis of HRV using techniques such as the Fast Fourier Transform (FFT) cannot be performed directly on such IHR signals. This study explores an alternative approach to estimating heart rate from EKGs using the concept of instantaneous frequency. Additionally, we compare HRV metrics derived from the new method with those obtained through traditional R peak detection techniques. Here, we filter out a signal from the EKG containing constituent frequency components correlating with the instantaneous heart rate. The extracted signal is further processed to produce a new signal termed the IHR signal that shows the variation of the instantaneous heart rate with time. Since this IHR signal does not rely on R peaks, HRV metrics calculated using it do not suffer from errors resulting from incorrect R peak detection. Additionally, This IHR signal is sampled at a constant rate, enabling frequency domain analysis using algorithms such as the FFT. This makes it straightforward to calculate frequency domain HRV metrics. Additionally, the amount of information carried by the IHR signal calculated using the instantaneous frequency method is greater than the signal derived from RR interval measurements. This is due to the relatively low data points available to construct the IHR signal from RR intervals compared to the high temporal resolution of the IHR signal constructed using the instantaneous frequency method. In conclusion, the motivation for this study is to provide the medical community with a methodology to assess heart rate and heart rate variability from EKGs that potentially offers more information and addresses the shortcomings associated with algorithms reliant on R peak detection. We hope this study provides medical researchers with a new approach to calculating HRV metrics to better understand physiological functions and conditions.

Presenting Author: Don Jayasooriya Virginia Polytechnic Institute and State University

Presenting Author Biography: Don Prathap Jayasooriya is a Mechanical Engineer working in Power Electronics. He completed his Master of Science in Mechanical Engineering at Virginia Tech in 2024. His research interests lie in Digital Signal Processing, Dynamics and Control, and Mechatronics. He currently works as an Associate Engineer at TMEIC Corporation Americas.

Authors:

Don Jayasooriya Virginia Polytechnic Institute and State University
Alfred Wicks Virginia Polytechnic Institute and State University