PhysioNetPhysiology

Network Physiology for the Physio ecosystem — dynamic-interaction analysis of coupled physiological systems (brain, heart, respiration, muscle).

Single-modality tools cannot ask how different organ systems couple and reorganise together. This package does, on top of the multimodal PhysioExperiment data model, via Time Delay Stability (TDS) (Bashan et al. 2012): it tracks, over sliding windows, the time lag that maximises each signal pair's cross-correlation, and rewards periods where that lag stays put — the fingerprint of a genuine dynamic coupling. Stable couplings become links in an organ-interaction network, and the package measures how that network reconfigures across physiological states.

library(PhysioNetPhysiology)

# Four physiological systems on a common (slow) grid. Here a shared driver
# couples brain/heart/resp at fixed lags, while muscle stays uncoupled; in
# practice these come from PhysioEEG/PhysioECG/PhysioEMG.
set.seed(42)
n <- 1200
drive <- cumsum(rnorm(n)); drive <- (drive - mean(drive)) / sd(drive)
lag <- function(v, k) c(rep(v[1], k), head(v, -k))
X <- physioNodeMatrix(list(
  brain  = drive          + rnorm(n, sd = 0.05),
  heart  = lag(drive, 2)  + rnorm(n, sd = 0.05),
  resp   = lag(drive, 4)  + rnorm(n, sd = 0.05),
  muscle = rnorm(n)
))

# 1. Time Delay Stability between every pair
tds <- timeDelayStability(X, sampling_rate = 1,
                          window_sec = 60, max_lag_sec = 8)
summary(tds)

# 2. Organ-interaction network, links significance-tested vs surrogates
net <- tdsNetwork(tds, surrogate = TRUE, X = X)
plotTDSnetwork(net)

# 3. How the network reconfigures across states (e.g. sleep stages)
sleep_stage <- rep(c("wake", "sleep"), each = n / 2)
bystate <- tdsNetworkByState(X, states = sleep_stage, sampling_rate = 1,
                             window_sec = 60, max_lag_sec = 8)
plotTDSreconfiguration(bystate)

Installation

The ecosystem builds on Bioconductor, so its repositories have to be on the list as well -- without them the install stops at SummarizedExperiment.

install.packages("BiocManager", repos = "https://cloud.r-project.org")
install.packages(
  "PhysioNetPhysiology",
  repos = c("https://x-biosignal.r-universe.dev", BiocManager::repositories())
)

From GitHub instead:

install.packages("remotes", repos = "https://cloud.r-project.org")
remotes::install_github("x-biosignal/PhysioNetPhysiology")

Why TDS

TDS is robust to the amplitude non-stationarity of physiological signals and to differing signal types (it operates on the timing of coupling, not its strength), which is why it underlies the Network-Physiology programme. Links are significance-tested against phase-randomised surrogates that preserve each signal's own spectrum while destroying cross-signal timing.

Scope

The core engine works on any multivariate time series; the physioNodeMatrix() helper and the PhysioExperiment integration make organ-system assembly convenient. Reference: Bashan et al. (2012) Nat Commun 3:702; Bartsch et al. (2015) PLoS ONE 10:e0142143.