Effects of Missing Data on Heart Rate Variability Metrics

Abstract

Heart rate variability (HRV) has been studied for decades in clinical environments. Currently, the exponential growth of wearable devices in health monitoring is leading to new challenges that need to be solved. These devices have relatively poor signal quality and are affected by numerous motion artifacts, with data loss being the main stumbling block for their use in HRV analysis. In the present paper, it is shown how data loss affects HRV metrics in the time domain and frequency domain and Poincaré plots. A gap-filling method is proposed and compared to other existing approaches to alleviate these effects, both with simulated (16 subjects) and real (20 subjects) missing data. Two different data loss scenarios have been simulated: (i) scattered missing beats, related to a low signal to noise ratio; and (ii) bursts of missing beats, with the most common due to motion artifacts. In addition, a real database of photoplethysmography-derived pulse detection series provided by Apple Watch during a protocol including relax and stress stages is analyzed. The best correction method and maximum acceptable missing beats are given. Results suggest that correction without gap filling is the best option for the standard deviation of the normal-to-normal intervals (SDNN), root mean square of successive differences (RMSSD) and Poincaré plot metrics in datasets with bursts of missing beats predominance (p<0.05), whereas they benefit from gap-filling approaches in the case of scattered missing beats (p<0.05). Gap-filling approaches are also the best for frequency-domain metrics (p<0.05). The findings of this work are useful for the design of robust HRV applications depending on missing data tolerance and the desired HRV metrics.

Optimizing support vector machines and autoregressive integrated moving average methods for heart rate variability data correction

“…However, even in these cases, deletion is probably a better choice, since it performs better than cubic interpolation for RMSSD and does not alter the RR interval lengths to the same degree. Deletion is also the recommendation given in [ 5 ] and [ 6 ] for these instances. For the methods in this study, there does not seem to be a difference in the results when applied to data from healthy subjects or data from subjects with atherosclerosis.…”

“…When measured during movement, either by photoplethysmography (PPG) or with a heart rate chest strap, motion artifacts present a significant problem for the analysis of the HRV data. These artifacts need to be handled correctly, as they can greatly affect the analysis [5] , [6] . Bourdillon et al.…”

“…The most common method for gap filling is cubic spline interpolation [ 8 ]. Other methods have also been proposed [5] , [9] , including linear interpolation and simple deletion, but the error for longer gaps is usually large and the beat-to-beat variability is in many cases unrealistic for the corrected HRV. We have previously tested recurrent neural networks (RNNs) for HRV data artifact correction [ 10 ], but did not achieve satisfactory results.…”

Advances in Wearable Photoplethysmography Applications in Health Monitoring

“…Motion artifacts and data loss are two really frequent issues to be addressed during the pre-processing of PPG signals and derived parameters provided by wearables. Cajal et al investigated data loss effects in HRV metrics using both simulated and real missing data [ 10 ]. The PPG-derived heart rate series were provided by Apple Watch during relaxing and stress-inducing experimental conditions.…”

A Novel Adaptive Noise Elimination Algorithm in Long RR Interval Sequences for Heart Rate Variability Analysis

“…Consequently, this results in different methods being used to process these signals. Overall ECG artifact identification and removal are made possible by various algorithms and are closely dependent on the aim [ 19 ]: a data-driven mechanism of empirical mode decomposition [ 20 , 21 ], deep-learning-based models [ 22 ], wavelet-based models [ 23 ], and sparsity-based, Bayesian-filter-based, and Hybrid models [ 24 ]. The majority of them focus on short-duration artifacts, which are often treated in the same way as ectopic beats [ 25 , 26 , 27 ].…”