A spike correction approach for variability analysis of heart rate sick infants

In critical care monitoring, the heart rate (HR) offers valuable insight into the autonomic function of sick infants. However, the intensity of monitoring and clinical care such as intubation, suctioning, and venesection as well as routine movement, create a hostile environment for contamination of continuous signals. These artifacts usually present as spikes in the HR signal, which interfere with the characterization and subsequent evaluation of the HR. Post hoc spike removal is commonly required in research studies but is not feasible in clinical monitoring. We propose a two-step process to correct spikes in HR data. Step 1 comprises of two sub-steps to remove the spikes with upward deflection and downward deflection. In Step 2, we repeat Step 1, for different ϵ values and calculate root mean square (RMS) of the difference between the uncorrected HR and the corrected HR. The corrected HR that displayed either the smallest RMS value or the same RMS values for two or more ϵ values is considered optimally corrected data. We demonstrate the application of this approach to HR data collected from 5 preterm infants. We show that there is a significant difference between the spectral powers obtained for spike uncorrected and spike corrected HR.