Online-only Table 1 Heart rate variability approach.

fFdP

Fano factor distance from a Poisson distribution

Describes fractal-like point processes via a modified estimate of the Fano factor, based on the average distance to a Homogeneous Poisson Point Process.

small = less complex

aFdP

Allan factor distance from a Poisson distribution

Describes fractal-like point processes via a modified estimate of the Allan factor based on the average distance to a Homogeneous Poisson Point Process.

small = less complex

MultiFractal_c1

MultiFractal spectrum cumulant of the first order

Maximum of the multifractal spectrum that can be viewed as the most common variability within a window, similar to a global variability.

small = more variability

MultiFractal_c2

MultiFractal spectrum cumulant of the second order

Width of the multifractal spectrum i.e. how variability departs from C1 value. Small C2 = variability changes little over time. High C2 = variability of the signal can differ greatly over time

small = more variability

Correlation dimension

Correlation dimension

Overall complexity; related to fractal dimension; minimum number of variables required to characterize system dynamics.

small = less complex

DFA Alpha 1

Detrended Fluctuation Analysis: α exponent

Scaling analysis method to represent correlation properties. α≈1: fractal-like temporal structure; α = 0.5 uncorrelated data (white noise); α = 1.5 Brownian noise (smooth fluctuations).

closer to 0.5 = roughest; closer to 1.5 = smoothest; healthy ≈ 1

DFA Alpha 2

Detrended Fluctuation Analysis: α1 exponent

Scaling analysis method to represent short-term correlation properties. α≈1: fractal-like temporal structure; α = 0.5 uncorrelated data (white noise); α = 1.5 Brownian noise (smooth fluctuations). Thought to reflect baroreceptor reflex or parasympathetic modulation.

closer to 0.5 = roughest; closer to 1.5 = smoothest; healthy ≈ 1

DFA AUC

Detrended Fluctuation Analysis: Area Under the Curve

Area Under the DFA curve estimating the total variance of the signal across time scales

small = low variability

Power Law Slope LombScargle

Power Law Slope (LombScargle method)

Slope of the linear portion of the power spectrum on a log-log plot. Represents long term scaling of fractal like processes with long range dependence (i.e. 1/f power-law exponent)

small = less complex

Hurst exponent

Hurst exponent

Index of long-range dependence and related to the fractal dimension of a system i.e. the minimum number of degrees of freedom of the dynamical system.

small = higher fractal dimension

eScaleE

Embedding Scaling exponent

Estimates fractal scaling by looking at how much the covariance of embedded vectors changes with embedding dimension. It relates to the Hurst exponent.

small = less complex

AsymI

Multiscale time irreversibility asymmetry index

Degree of temporal asymmetry and lack of invariance of the statistical properties of a signal over multiple scales; Pathologic signals are more symmetric than healthy ones.

small = less complex

Largest Lyapunov exponent

Largest Lyapunov exponent

Represents the average exponential growth rate of the distance between two neighbouring points in the dynamical system trajectory as time evolves. Measure of predictability, entropy rate, chaotic behavior. It is positive for chaotic data.

small = less complex/chaotic

SDLEalpha

Scale dependent Lyapunov exponent slope

Multiscale complexity measure estimating the main slope of the log-log plot of Lyapunov exponents at multiple scales, which characterizes the speed of loss of information i.e. the uncertainty involved in predicting the value of a random variable. Initial slope (lower scales) typically indicates noisy dynamics.

small (steeper descending slope) = more complex

SDLEalpha2

Scale dependent Lyapunov exponent slope (after break)

Multiscale complexity measure estimating the main slope of the log-log plot of Lyapunov exponents at multiple scales, which characterizes the speed of loss of information i.e. the uncertainty involved in predicting the value of a random variable. Typically the second slope is flat and indicates chaotic regime.

if flat = chaotic regime

SDLEmax

Scale dependent Lyapunov exponent max value

Multiscale complexity measure estimating the maximum of Lyapunov exponents across multiple scales. The maximum typically occurs at small scales and is related to entropy measures.

small = less complex

gcount

Grid transformation feature: grid count

Trajectory of a dynamical system is represented on a discrete grid and the number of “visited” pixels is counted. Represents the degree of complexity or chaotic dimension.

small = less complex

sgridAND

Grid transformation feature: AND similarity index

Estimates self-similarity (regularity of a signal) between discretized delayed versions of dynamical system trajectories. Measures both the spread of trajectories as well as their similarity between two time instants.

small = less similar

sgridTAU

Grid transformation feature: Time delay similarity index

Estimates self-similarity (regularity of a signal) between discretized delayed versions of dynamical system trajectories.

small = less similar

shannEn

Shannon Entropy

Average information content of a signal

small = less complex

Sample entropy

Sample Entropy

Represents the signal unpredictability or regularity within short time segments by estimating the entropy rate

small = less complex

QSE

Quadratic Sample Entropy

Normalized version of the sample entropy (with respect to the tolerance window); Represents the signal unpredictability or regularity within short time segments by estimating the entropy rate

small = less complex

Multiscale Entropy

Multiscale Entropy

Represents signal unpredictability/regularity within short time segments (akin to sample entropy) but over multiple scales

small = less complex

KLPE

Kullback Leibler permutation entropy

Deviation from randomness. With increasing complexity of the time series, KLPE decreases until it reaches zero for noise. Higher values indicate better predictability of the system.

small = more complex

Control entropy

Control Entropy

Entropy measure designed to be robust to non-stationary signals; Sample entropy of the successive differences

small = less complex

ARerr

Predictive feature: error from an autoregressive model

Measure of predictability

small = low variability

LF Power LombScargle

LF Power (Lomb-Scargle method)

Power contained in the low frequency band (0.04–0.2 Hz for fetal recordings) of the ECG spectrum. Reflects both sympathetic and vagal activity, and blood pressure regulation via baroreceptors.

small = less power

HF Power LombScargle

HF Power (Lomb-Scargle method)

Power contained in the high frequency band (0.2–2 Hz for fetal recordings) of the ECG spectrum. Mostly reflects vagal modulation of HR and is related to the respiratory cycle.

small = less power

LF/HF ratio LombScargle

LF/HF ratio (Lomb-Scargle method)

Represents sympathetic and parasympathetic modulation but its interpretation is experiment dependent and unclear in general.

small = less power

VLF Power LombScargle

VLF Power (Lomb-Scargle method)

Power in the Very Low Frequency Band. Thought to relate to thermoregulation and to be sympathetically mediated

small = less power

pD

RQA: percentage of determinism

% of recurrent points forming diagonal lines, with a minimum of two adjacent points (deterministic). Measures predictability

small = more chaotic

pDpR

RQA: determinism/recurrences

ratio of %Determinism over %Recurrence. Quantifies transition/nonstationary periods in a system.

small = more stationary

pL

RQA: percentage of laminarity

Laminarity detects laminar states (chaos-chaos transitions) and rapid changes in RRI

small = more complex

pR

RQA: percentage of recurrences

Global measure of recurrence (% of plot filled with recurrent points). Periodic dynamics have higher % of recurrence than aperiodic dynamics.

small = more complex

dlmax

RQA: maximum diagonal line

Maximal length of the diagonal structures in a recurrence plot, representing exponential divergence of the trajectories. Detects transitions from periodic to chaotic behavior and Inversely related to the largest lyapunov exponent.

small = more chaotic

dlmean

RQA: average diagonal line

Average length of the diagonal structures in a Recurrence Plot. Mean prediction time of the system.

small = more chaotic

sedl

RQA: Shannon entropy of the diagonals

Rough measure of the information content of the trajectories (diagonal lines) on a Recurrence plot.

small = less complex

vlmax

RQA: maximum vertical line

Max length of vertical lines on a Recurrence plot. Information about the time duration of the laminar states and marker of intermittency

small = more chaotic

Mean

Mean of RR-intervals

Sample mean of the RR interval time series.

small = low variability

Standard Deviation

Standard Deviation

Standard deviation of Normal-Normal (NN) intervals.

small = low variability

RMSSD

Root mean square of successive RR interval differences

Represents beat-to-beat variance in HR and estimates vagally mediated changes in HRV. Mathematically equivalent to Poincare SD1

small = low variability

Coefficient of variation

Coefficient of variation

Standard deviation normalized by the mean

small = low variability

histSI

Similarity index of the distributions

Similarity index between the statistical distribution of two consecutive data blocks (0: lowest similarity or highest variability; 100: highest similarity or lowest variability)

small = high variability

Complexity

Hjorth’s Form Factor (Complexity)

Measure of “excessive details with reference to sine wave”, assesses complexity in the signal; Complexity of the first order. Sensitive to noise. Can be used to detect arrhythmias?

small = less complex

Poincaré SD1

Poincaré plot SD1

Poincaré plot standard deviation perpendicular the line of identity. Represents short interval variations and reflects parasympathetic index of sinus node control. Mathematically equivalent to RMSSD.

small = low variability

Poincaré SD2

Poincaré plot SD2

Poincaré plot standard deviation along the line of identity. Represents long interval variation and is linked to both parasympathetic and sympathetic tones. Related to Standard Deviation

small = low variability

CSI

Poincare plot Cardiac Sympathetic Index

Ratio between SD2 and SD1, represents unpredictability of the RR time series

small = more random/scattered

CVI

Poincare plot Cardiac Vagal Index

Log of the Area of the ellipse fitted to Poincare plot: Total variability

small = low variability

SymDce_2

Symbolic dynamics: modified conditional entropy, non-uniform case

Characterizes entropy rate

small = less complex

SymDfw_2

Symbolic dynamics: forbidden words, non-uniform case

High number of forbidden words indicates a more regular behaviour of time series

small = more complex

SymDp0_2

Symbolic dynamics: percentage of 0 variations sequences, non-uniform case

Patterns with no variation i.e. all the symbols are equal. Thought to reflect cardiac autonomic modulation, predominantly sympathetic modulation.

small = more complex

SymDp1_2

Symbolic dynamics: percentage of 1 variations sequences, non-uniform case

Patterns with one variation i.e. two consecutive symbols are equal and the remaining one is different

small = less complex

SymDp2_2

Symbolic dynamics: percentage of 2 variations sequences, non-uniform case

Patterns with two variations (either like or unlike variations). Thought to reflect cardiac autonomic modulation, predominantly parasympathetic modulation.

small = less complex

SymDse_2

Symbolic dynamics: Shannon entropy, non-uniform case

Characterizes entropy, i.e., complexity of the pattern distribution.

small = less complex