Ultra-short-term HRV analysis in a dynamic scenario using the traditional and functional approach

Abstract

Objective. Heart rate variability (HRV) is increasingly used to study short, dynamic, stimulus-aligned physiological responses, but recordings shorter than one minute are vulnerable to artefacts, non-stationarity and loss of information when reduced to single-number metrics. The objective of this study was to evaluate whether selected ultra-short-term (UST) HRV metrics and functional data analysis (FDA) can be used to characterize cardiovascular arousal during a brief dynamic simulated interpersonal threat/self-defence scenario. Approach. We first reviewed HRV metrics that are theoretically suitable for UST recordings and then examined their sensitivity to physiologically plausible trend components using simulated 30 s RR-interval series. We subsequently applied both traditional HRV metrics and FDA to a real-world dataset from a controlled self-defence experiment. For the real-world analysis, 20 s windows were aligned to the externally annotated onset of physical contact, covering 10 s before and 10 s after the event. RR-interval series were cleaned, converted to heart-rate time series, represented as smoothed B-spline functions, and compared using mixed-effects models and functional ANOVA. First-derivative functions were analysed to separate absolute HR level from the rate of HR change. Main results. In the simulation, mean and median HR primarily reflected vertical shifts in the HR trajectory, whereas SDNN, the HRV triangular index and SD2 were sensitive to the rate of increase. RMSSD and SD1 were the least affected by the imposed trend, consistent with their mathematical relationship and their focus on short-term beat-to-beat variability. In the real-world dataset, both approaches detected group- and scenario-related differences. The FDA approach additionally localized these differences across the event-aligned time window and distinguished differences in absolute HR from differences in the rate of HR increase. Significance. FDA should not be interpreted as a real-time classifier or as a replacement for established HRV metrics. Rather, it provides a complementary framework for analysing short, non-stationary, event-aligned recordings in dynamic settings, especially when the timing and shape of the physiological response are scientifically relevant.