Disentangling Neural Activity from Vascular Dynamics in Neuroimaging

43 concepts37 papers1 proposal

Non-invasive functional neuroimaging techniques do not measure electrical neural signaling directly. Instead, methods like functional Magnetic Resonance Imaging (fMRI), functional near-infrared spectroscopy (fNIRS), and functional ultrasound imaging (fUSI) rely on hemodynamic proxies—such as the blood-oxygen-level-dependent (BOLD) signal, cerebral blood volume, or local blood flow.

This proxy relies on neurovascular coupling: the biological cascade where active neurons signal nearby blood vessels to dilate, delivering oxygen and glucose to meet metabolic demands. However, because the measured signal passes through this vascular "filter," a fundamental ambiguity arises. When an imaging signal changes, does it reflect a shift in underlying neural computation, a change in the physical structure and reactivity of the blood vessels, or systemic physiological noise? Disentangling these factors is critical for drawing valid conclusions about brain function.

The Linear Assumption and the Nonlinearity Trap

To translate vascular changes back into neural activity, researchers historically relied on mathematical simplifications. In a foundational study, Boynton et al. (1996) validated a "linear transform model" for fMRI, demonstrating that the sluggish hemodynamic response could be treated as a predictable linear system acting on underlying neural firing. This framework allowed scientists to separate signals in time and contrast, inferring neural properties directly from non-invasive data.

However, this linear assumption breaks down when modeling complex, non-linear brain behaviors. When an advanced neural encoding model encounters a non-linearity, it faces an interpretational wall. As Gao et al. (2025) point out, indirect measurements capture a composite of both neural processing and vascular responses, both of which possess non-linear characteristics. This makes it difficult to isolate whether the complexity observed in an encoding model is driven by actual neural computation or by the non-linear nature of the blood-oxygenation signal itself.

This ambiguity is highly localized. By analyzing voxel-wise responses during contrast adaptation, Zhou et al. (2021) demonstrated that signal decrements exhibit complex, voxel-specific variations that standard linear models fail to capture. Without a clear way to separate these drivers, it remains uncertain whether these non-linearities stem from distinct neuronal subpopulations (such as off-center cells) or the physical architecture of the local microvasculature.

Moving Targets: Lifespan Development and Aging

The vascular-neural ambiguity becomes particularly acute when studying brains that are undergoing structural transformations, such as during development or aging. In these populations, the vascular filter itself is a moving target.

During fetal and pediatric development, the brain is simultaneously building its neural circuits (synaptogenesis) and its blood vessel networks (angiogenesis). Hagmann et al. (2012) note that because functional imaging relies on blood oxygenation, developmental changes in the vascular system itself act as a major confounding variable, easily leading researchers to misinterpret vascular growth as shifts in neural connectivity.

When mapping early functional networks in utero, the growth of blood vessels can mimic or mask actual neurological maturation. Taymourtash et al. (2022) emphasize that the precise developmental coupling between vascular growth and brain function remains unclear, potentially confounding observations of emerging thalamocortical pathways. To address this, Karolis et al. (2023) introduced "maturational networks" to model connectivity as an emerging process, though they reinforce that isolating neural circuit maturation from hemodynamic development remains a persistent challenge.

[Neural Maturation / Synaptogenesis] ---\
                                         +--> [Altered BOLD Signal]
[Vascular Growth / Angiogenesis] -------/

At the other end of the lifespan, normal aging causes blood vessels to lose elasticity and stiffen. When resting-state fMRI reveals a decline in functional network intensity within white matter, Muwei Li et al. (2023) explain that BOLD metrics struggle to differentiate between a genuine drop in neural or glial metabolic demand and age-related changes in the cerebral vasculature. This ambiguity limits our ability to use hemodynamic measures as precise proxies for neural health in older adults. Similarly, when utilizing novel metrics like Functional Correlation Tensors to map direction-dependent signals in white matter, Lyuan Xu et al. (2024) note that it remains difficult to determine whether observed functional variations stem from neuronal signaling degradation or age-related vascular decline.

Pathological Confounders in Clinical Networks

In clinical settings, diseases frequently disrupt neural tissue and blood flow simultaneously, confounding diagnostic interpretations.

  • Cerebrovascular Burden: In patients with cognitive decline, distinguishing between primary neurodegenerative effects and secondary vascular influences is a major hurdle. Vipin et al. (2018) demonstrated that concomitant cerebrovascular disease inflicts early structural damage and shifts functional patterns within the Executive Control Network, confounding the memory-centric network degradation typical of Alzheimer's disease.
  • Perfusion Artifacts: Localized pathologies can cause "action at a distance" artifacts. Frontal lobe gliomas alter local cerebral blood flow and perfusion, weakening functional connectivity measurements in the distant posterior Default Mode Network (Zhang et al. (2016)). Without direct measurements of vascular dynamics, it is impossible to know if the tumor has disrupted the neural network or merely altered the hemodynamic response due to localized metabolic demands.
  • Signal Timing Delays: After a stroke, brain activity often exhibits a "synchronization lag." While this lag correlates with structural damage to white matter pathways, Wang et al. (2019) note that the influence of altered, localized blood flow on signal timing cannot be entirely ruled out, leaving the exact cause of the delay ambiguous.

When Rhythms Decouple

A foundational assumption of resting-state fMRI is that low-frequency fluctuations in the BOLD signal directly track spontaneous neural communication. However, evidence suggests that the vascular system can march to its own beat.

Ming Li et al. (2023) provided a striking demonstration of this by using a chemical blocker to disrupt neurovascular communication in mice. They discovered that spontaneous brain oscillations arise from at least two distinct sources—one neural and one vascular—rather than the blood simply mirroring the neurons. If hemodynamic and neural signals can decouple naturally outside of experimental settings, using blood-flow-based metrics to diagnose neurological diseases could lead to inaccurate conclusions about functional connectivity.

Normal State:      [Neural Rhythms] ====(Coupled)====> [Vascular Oscillations]
Disrupted State:   [Neural Rhythms]                    [Vascular Oscillations] (Independent)

This independent vascular rhythm complicates the interpretation of global brain states. Daniel Gutierrez‐Barragan (2024) discovered that mammalian brains exhibit evolutionarily conserved "coactivation modes" that cycle at infraslow frequencies. However, it remains unclear to what extent these patterns represent purely neural choreography versus non-neural physiological processes, such as systemic vascular rhythms or cerebrospinal fluid flow.

Similarly, large-scale synchronization disruptions observed in conditions like schizophrenia—where the brain struggles to reach highly polarized activity states—could be heavily influenced by vascular or hemodynamic factors rather than purely neural network connectivity (Miller et al. (2019)).

Conversely, some low-frequency oscillations (LFOs) are tightly bound to brain viability. Schulthess et al. (2024) demonstrated that specific LFOs tracked via fNIRS are a direct byproduct of active neurovascular coupling. While systemic signals like heartbeats persist after cerebral death, these neurogenic vascular oscillations completely vanish, providing a potential bedside biomarker for brain viability.

Spatial Contamination and Structural Biases

The vascular architecture of the brain is not uniformly distributed; it follows physical anatomical constraints that bias imaging data independent of neural activity.

Kurzawski et al. (2022) demonstrated that non-neural factors—such as cortical thickness, curvature, and macrovascular distribution—account for up to 26% of BOLD variance. In fact, these structural factors explain away half of the long-observed signal boost along the brain's representation of the horizontal visual meridian, proving that local anatomy can easily be mistaken for functional neural specialization.

This spatial contamination is particularly challenging when imaging across cortical depth (laminar fMRI). Roefs et al. (2024) showed that the BOLD signal's laminar profile is heavily confounded by macrovascular blood pooling in the superficial layers due to venous drainage architecture. To resolve fine-grained spatial patterns of neural activity, researchers must explicitly normalize data to correct for these venous blood volume biases.

Grounding the Signal

To resolve the vascular-neural ambiguity, neuroscientists use direct electrophysiological validation, genetic models, and advanced signal decomposition.

On a cellular level, researchers have sought to identify the exact cellular messengers responsible for functional hyperemia. By utilizing a genetic knockout model, Huard et al. (2000) proved that functional hyperemia is not a generic property of all neurons; instead, specific interneurons, such as stellate cells in the cerebellum, act as the essential links between synaptic activity and blood vessel dilation.

To validate macroscopic imaging, Krumin et al. (2021) performed simultaneous high-density "Neuropixels" recordings and functional ultrasound imaging (fUSI) in awake mice. They established a quantitative, linear link showing that slow blood volume changes faithfully reflect slow neural firing rates (<0.3 Hz), particularly from inhibitory neurons, providing a solid biological justification for using vascular imaging in systems neuroscience.

[Neuropixels Recording]  <-- (Direct Spiking) -- [Inhibitory Neurons]
                                                        |
                                               (Neurovascular Coupling)
                                                        v
[Functional Ultrasound]  <-- (Linear Proxy) ---- [Blood Volume (<0.3 Hz)]

Finally, mathematical unmixing techniques help isolate neural signals from aggregate data. Vince D. Calhoun (2009) argues that while BOLD signals lack direct electrophysiological validation for low-frequency fluctuations, spatial Independent Component Analysis (sICA) can successfully identify robust functional networks. As Xu et al. (2013) demonstrate, sICA can "unmix" overlapping neural and vascular components within a single voxel. This reveals complex functional networks that are completely hidden from traditional analysis due to signal cancellation, helping map decomposed imaging components to verifiable neurobiological mechanisms.

Key Foundational Work

Linear Systems Analysis of Functional Magnetic Resonance Imaging in Human V1 (1996) — contributed 1 concept elements to this direction.
Stellate Neurons Mediate Functional Hyperemia in the Cerebellar Molecular Layer (2000) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_vascular_neural_signal_ambiguity_2b095a55,
  title = {vascular-neural signal ambiguity},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/vascular-neural-signal-ambiguity/rdr_2b095a55ccae464093796fc7dfa92a19},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_vascular_neural_signal_ambiguity_2b095a55_v1,
  title = {vascular-neural signal ambiguity},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/vascular-neural-signal-ambiguity/rdr_2b095a55ccae464093796fc7dfa92a19/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake