henhao
Sign in
PhysiologyThe Journal of Physiology
Peer-reviewed journal article

Non-invasive assessment of integrated cardiorespiratory network dynamics after physiological stress in humans

Cecilia Morandotti, Louise RignyiD, Thomas B. WilliamsiD, Juan I. BadariottiiD, Matt DicksiD, Amar S. BhogaliD, Shachula Han, Jo CorbettiD, Mike TiptoniD, Joseph T. CostelloiD, Ali Reza ManiiD

PDFDOIPreprintCode
In-depth full-text digest available · Jump to digest ↓

Abstract

BackgroundPhysiological variables provide critical insights into the integrated control of the cardiorespiratory system, reflecting the bodys real-time responses to internal and external perturbations. Visualizing the exchange of information between different components of the cardiorespiratory system is beneficial for monitoring individuals in various clinical settings and extreme environments. This study aimed to develop a non-invasive method, using principles of information theory, to visualize the flow of information between physiological variables, with a focus on the integrated cardiorespiratory responses to different physiological stressors. MethodsHeart rate, respiratory rate, minute ventilation, respiratory frequency, tidal volume, capillary oxygen saturation (SpO2), end-tidal oxygen, and end-tidal carbon dioxide concentrations were recorded from 22 healthy participants. Transfer entropy, which reflects measures of causal relationships between parallel time-series, was used to compute the flow of information between cardiorespiratory signals into network maps. Network mapping was performed after rest in a control condition and following exposure to the isolated and combined effects of normobaric hypoxia (FIO2: 0.12), moderate intensity cycling exercise (100W), and overnight sleep deprivation. For each intervention, 10-minute segments of physiological signals (from minutes 5 to 15 after the commencement of hypoxia and/or exercise) were used for analysis. ResultsEach physiological stressor was associated with a distinctive pattern of information flow between physiological variables. Hypoxia led to the engagement of SpO2 a hub in the network, facilitating the exchange of information with end-tidal oxygen concentration and heart rate. Sleep deprivation was associated with a shift in the flow of information from SpO2 to other nodes, such as respiratory rate, during hypoxia. During exercise, heart rate emerged as the central node for receiving information, while SpO2 acted as the primary node disseminating information to other nodes. Increased connectivity within the networks was observed during exercise alone or when combined with other stressors. ConclusionThis non-invasive network mapping technique visualizes the interaction of various cardiorespiratory variables following exposure to physiological stressors. By mapping normative responses, this approach may help identify outliers associated with various disease states.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
12
FWCI
8.38
Field percentile
top 2%
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (18 Mar 2025) · CC BY 4.0

Transfer entropy network maps of seven breath-by-breath cardiorespiratory signals in 22 healthy adults differ by stressor; only exercise raised total transfer entropy, while hypoxia and sleep deprivation reshaped node roles.

Patient monitoring typically relies on mean values of single signals such as heart rate or capillary oxygen saturation (SpO2), and the authors note there is no robust way to map directed interactions among non-invasively recorded signals. Here pairwise transfer entropy is paired with shuffled-surrogate significance testing and directed centrality, unlike the group's earlier sepsis analysis, which used routine signals without a surrogate test.

Key results

  1. ShownExercise was the only stressor that increased total information transfer, the summed transfer entropy (TE) across all signal pairs; hypoxia or sleep deprivation without exercise left it unchanged.

    Three-way ANOVA on total TE: exercise effect P<0.001, with no significant interactions among hypoxia, exercise and sleep deprivation; the increase held in all four exercise conditions relative to their non-exercise counterparts (Figure 4, medians with interquartile ranges). · Results; Discussion

  2. ShownDuring exercise, heart rate became the main receiver of information and SpO2 a leading sender, with SpO2 to heart rate the most prominent added link; adding hypoxia reduced heart rate's centrality.

    Group-median maps (visual reading, untested): heart rate had the highest indegree with exercise alone or plus sleep deprivation; SpO2 the highest outdegree with exercise alone, and SpO2 to heart rate the most prominent link. Three-way ANOVA, the only test, on heart-rate indegree: exercise P<0.001, exercise by hypoxia interaction P<0.001; heart rate received and sent less when hypoxia accompanied exercise. · Results; Discussion

  3. SuggestedHypoxia at rest (inspired O2 fraction 0.12) recruited SpO2 as a hub exchanging information with heart rate and end-tidal O2; combining hypoxia with exercise produced more links, with SpO2 the top receiver.

    Visual comparison of group-median maps versus normoxic rest: a new bidirectional heart rate and SpO2 link, more inflow to SpO2 from heart rate and end-tidal O2, stronger end-tidal CO2 to O2 transfer. SpO2 outdegree generally rose under hypoxia except with sleep deprivation. No statistical test of these patterns; hypoxia data come from 12 participants. · Results; Discussion; Study limitations

  4. SuggestedSleep deprivation changed network structure without changing the mean of any recorded variable; combined with hypoxia, information flow shifted away from SpO2 toward respiratory rate.

    Three-way ANOVA: no significant sleep-deprivation effect on any variable's mean. In the hypoxia plus sleep deprivation map, respiratory rate sent the most information and tidal volume, end-tidal O2 and end-tidal CO2 received the most. The authors could not explain the loss of SpO2 centrality and say further research must determine whether it was an outlier or a real functional change. · Results; Discussion

  5. ShownResting normoxic networks were dominated by ventilatory and gas-exchange links, providing the reference against which each stressor's map was compared.

    Normoxia control map: bidirectional exchange between end-tidal O2 and end-tidal CO2, minute ventilation and tidal volume, and tidal volume and end-tidal CO2; one-way transfer from heart rate to end-tidal O2 and from respiratory rate to tidal volume; end-tidal O2 and CO2 had the highest indegree. · Results

Methods, data and reproducibility

Seven cardiorespiratory signals, sampled once per breath, were recorded under 8 conditions crossing hypoxia (inspired O2 fraction 0.12), 100 W cycling and sleep deprivation. Pairwise transfer entropy on 10-minute windows (lag 5 breaths) was kept if it beat the 95th percentile of 100 shuffled surrogates; group-median matrices gave directed networks scored by indegree and outdegree.

  • Experiment 1 at the University of Portsmouth Extreme Environments Laboratory: normobaric hypoxia, moderate cycling (100 W) and one night of total sleep deprivation, isolated or combined · human · n = 12 participants
  • Experiment 2: exercise and three nights of partial sleep deprivation, isolated or combined · human · n = 12 participants
  • Analysed cohort pooled from both experiments: healthy, physically fit volunteers (age 25.1 ± 3.6 years) after two exclusions for incomplete data; 10-minute windows per condition · human · n = 22 (5 females)
Methods & tools
MATLAB, Transfer entropy MATLAB function (PhysioNet, Darbellay-Vajda adaptive partitioning), SPSS Statistics, Quark CPET metabolic cart (Cosmed), Nonin 7500 pulse oximeter
Code
https://www.physionet.org/conTEt/tewp/1.0.0/
Reproducibility gaps
Data only on reasonable request. Only the third-party TE MATLAB function is linked (printed path reads 'conTEt'); the authors' scripts for surrogate testing, network mapping and centrality are not shared. No MATLAB or SPSS versions given.
Limitations (6)
  • Participants were young, healthy, physically active volunteers screened for ethical reasons, so findings may not hold in a wider population; older people and those with comorbidities remain to be studied. · stated by the authors
  • Variables were sampled once per respiratory cycle, so information transfer at higher frequencies, such as flow driving high-frequency heart rate fluctuations, could not be assessed. · stated by the authors
  • Only 12 participants underwent the hypoxic challenge, ruling out paired or repeated-measures analysis; the authors call for a larger, power-calculated, more gender- and age-balanced sample. · stated by the authors
  • Why SpO2 lost centrality under hypoxia with sleep deprivation is unexplained, and the authors say further work must determine whether this was an outlier or a real functional change. · stated by the authors
  • The sleep-deprived conditions pool one night of total and three nights of partial sleep deprivation from two separate experiments, and hypoxic conditions come only from the first, so the sleep factor mixes protocols and cohorts. · noted in this summary
  • Most condition-specific network differences rest on visual comparison of group-median maps; only total TE and heart-rate indegree were tested, and missing values were replaced with the series mean before TE computation. · noted in this summary

Who should read:Read this if you work with multichannel physiological monitoring and want a surrogate-tested transfer entropy recipe for directed cardiorespiratory networks, plus healthy-volunteer reference maps under hypoxia, exercise and sleep loss.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISBarà '25Faes '20Zaccaro '22Zaccaro '26Candia‐Rive '22Chen '20
67% semantic matchBioengineering
Chiara Barà, Laura Sparacino, Luca Faes+1 more·bioRxiv·Mar 26, 2025· 1 citations