ECG-derived respiration to explain resting-state BOLD fluctuations and respiratory modulations: the question, the experiment, and why it answers — a commentary on Esteves et al. (Scientific Reports, 2025)

1) The scientific question

The core question is: Can we extract a useful respiration signal from ECG recorded inside the MRI environment (ECG-derived respiration, EDR) without additional respiratory hardware, and still explain BOLD fluctuations and estimate cerebrovascular reactivity (CVR) with performance comparable to measured respiration?
This matters especially for EEG-fMRI, where ECG is typically available, but respiratory recordings can be missing, corrupted, or logistically difficult.

ECG-derived respiration to explain resting-state BOLD fluctuations and respiratory modulations

2) The experiment

The authors acquired EEG-fMRI data from 15 healthy participants across three conditions:

• Resting state (RS)

• Slow-paced breathing (SPB, 0.1 Hz)

• Breath-holding (BH)

They recorded both:

• ECG (already standard in many EEG-fMRI setups), and

• measured respiration (ground truth)

They then extracted multiple EDR signals spanning several method families (ENV, HRV, AM, QRS-AM, PCA, kPCA, EMD), and evaluated:

1. EDR vs measured respiration similarity (time-domain correlation with optimal lag and frequency-domain coherence)

2. Physiological regressor quality for:

• BOLD denoising (RETROICOR respiratory terms, RVT-based models)

• CVR estimation, particularly during BH

3. Direct comparison of EDR-based regressors vs respiration-based regressors in variance explained and voxelwise mapping.

3) Why this experiment answers the question

It answers the question by testing EDR at three practical levels:

1. Does EDR resemble respiration at all?
   They evaluate both temporal similarity (correlation) and spectral fidelity (coherence), which matters because many fMRI physiological regressors depend heavily on respiratory frequency content.

2. Does EDR “work” for the fMRI use-cases we care about?
   They move beyond resemblance and test what matters: how much variance in BOLD is explained and how regressor choice changes RVT/CVR estimates.

3. Does EDR survive strong respiratory manipulations?
   SPB and BH are deliberately included to stress-test the approach outside passive rest.

The key result that closes the loop: amplitude-based EDR methods performed poorly under MRI ECG distortions, while an HRV-based EDR approach performed most consistently, yielding denoising and CVR-related estimates broadly comparable to those obtained with measured respiration.

4) BrainLatam reading — APUS (extended proprioception)

We read this paper as experimental infrastructure. It improves how we model the body’s contribution to BOLD without adding equipment. From an APUS perspective, BOLD fluctuations are not purely “brain”—they are also shaped by how the body is organized in the scanner (comfort, micro-adjustments, subtle movement constraints). When respiration is missing or unreliable, the physiological model loses part of that embodied context.

A robust ECG-based proxy for respiratory variability—especially HRV-derived EDR—functions as a practical shortcut that helps preserve interpretability without increasing setup complexity.

5) BrainLatam reading — Tekoha (extended interoception)

Tekoha is central here: respiration, HRV, and BOLD are interconnected faces of internal regulation. A major strength of the study is demonstrating that—even with MRI-induced ECG distortion—meaningful respiratory information can be recovered well enough to support:

• physiological noise modeling and correction

• RVT-related BOLD mapping

• CVR estimation during BH (with reduced sensitivity but comparable amplitude patterns)

A key nuance: HRV-based EDR may capture not only respiration, but also a broader autonomic component that itself modulates BOLD. That is not a flaw; it is a reminder that “physiological noise” is often Tekoha operating through vascular and autonomic pathways.

6) Limits that define the next experiment

• Variable time lag between EDR and measured respiration: correlation improves with optimal lag, but the optimal lag varies across subjects/tasks, limiting “blind” deployment when ground-truth respiration is absent.

• Strong dependence on ECG quality in MRI: morphology/amplitude-based methods degrade more under distortion.

• During BH, EDR-derived RVT produced fewer significant voxels, although average BOLD percent signal change was similar—suggesting reduced sensitivity but still usable signal.

• Conceptual caution: indiscriminate removal of EDR-related variance can remove physiologically meaningful BOLD components depending on the scientific aim (because respiration also carries metabolic and autonomic information).

7) BrainLatam translation to the organic world

BrainLatam translation to the organic world: this work shows that we can reduce experimental burden and still preserve much of the physiological modeling needed for BOLD interpretation by leveraging a signal already present in EEG-fMRI (ECG). In real-world scenarios—missing respiratory belts, corrupted respiratory traces, retrospective datasets—HRV-based EDR becomes an operational alternative that supports denoising and CVR-related modeling without additional hardware.

8) Open BrainLatam question

If HRV-based EDR reflects both respiration-linked variability and broader autonomic regulation, should we treat EDR primarily as:

• a “respiration substitute,” or

• explicitly as a composite cardiorespiratory marker, modeling separately what is respiration-driven versus autonomic-driven in BOLD fluctuations?

The body does not need belief to function.
It needs space, movement, and regulation.

Ref.:

‌Esteves, I., Fouto, A. R., Ruiz-Tagle, A., Caetano, G., & Figueiredo, P. (2025). Using ECG-derived respiration for explaining BOLD-fMRI fluctuations during rest and respiratory modulations. Scientific Reports, 15(1). https://doi.org/10.1038/s41598-025-23131-7