Functional Magnetic Resonance Imaging (fMRI) does not measure neurons; it measures plumbing. Specifically, the Blood Oxygen Level Dependent (BOLD) signal tracks changes in blood oxygenation that follow neural activity. While this proxy has revolutionized neuroscience, it faces a fundamental physical limit: blood flows.
When neurons fire, they trigger a localized hemodynamic response, but the resulting deoxygenated blood doesn't stay put. It drains into a hierarchical network of vessels, moving from tiny capillaries in deep cortical layers up through "diving" venules to large pial veins resting on the brain's surface. This drainage creates a "macrovascular bias" that can displace, smear, or inflate the perceived site of brain activity.
The Geometry of Signal Smearing
The primary challenge in high-resolution fMRI is distinguishing between the site of oxygen consumption (the neurons) and the site of the signal (the veins). As Wouter Schellekens and colleagues recently asked in Neuronal or vascular receptive fields?, are we mapping the brain’s functional logic or its vascular architecture?
Because blood drains toward the surface, the BOLD signal accumulates as it moves upward. This creates a spatial gradient where the "loudest" signals often appear in the superficial layers of the cortex, regardless of where the actual computation happened. Li M. Chen and John C. Gore demonstrated that even at ultra-high field strengths (9.4T), proximity to large draining vessels can introduce submillimeter offsets that displace the perceived center of activation.
The Laminar Conflict
This vascular bias is particularly problematic for laminar fMRI, which attempts to map activity across the ~2.5mm thickness of the human cortex. The cortex is organized into layers with distinct roles: deep layers often handle input, while superficial layers handle complex transformations.
Researchers like Michelle Moerel and Federico De Martino have used 7T fMRI to show that the auditory cortex performs more complex processing in its superficial layers Processing complexity increases in superficial layers of human primary auditory cortex. However, they must constantly contend with the fact that these superficial layers are also where pial veins are most concentrated. This "vascular gain" acts as a multiplicative confound; as Jan W. Kurzawski and Kendrick N. Kay discovered, non-neural factors like cortical thickness and vessel distribution can account for up to 26% of the variance in BOLD signal magnitudes Non-Neural Factors Influencing BOLD Response Magnitudes within Individual Subjects.
Strategies for De-veining the Brain
To achieve true mesoscopic precision, researchers are developing "de-veining" techniques to isolate the microvascular signal from the macrovascular noise.
- Phase-based Suppression: Patricia Pais‐Roldán and N. Jon Shah have optimized pre-processing strategies that use the phase component of the MRI signal to identify and filter out the magnetic signatures of large blood vessels Pre-processing of Sub-millimeter GE-BOLD fMRI Data for Laminar Applications.
- Alternative Contrasts (VASO): Instead of measuring oxygen (BOLD), which is prone to downstream smearing, researchers like Laurentius Huber and Elisa Zamboni use VASCular Occupancy (VASO) to measure changes in cerebral blood volume. This technique is inherently more localized to the capillaries near active neurons Mapping curvature domains in human V4 using CBV-sensitive layer-fMRI at 3T.
- Source Separation: In optical imaging, Holger Schoner and Martin Stetter applied Blind Source Separation to "unmix" neural signals from the occluding patterns of surface blood vessels Application of Blind Separation of Sources to Optical Recording of Brain Activity.
- Venous Mapping: New methods like DiSpect, developed by Ekin Karasan and Chunlei Liu, map "venous territories" by tracing blood flow in reverse—from the large pial veins back to their tissue origins—allowing researchers to pinpoint exactly which upstream region fueled a specific signal MR perfusion source mapping depicts venous territories.
By accounting for the brain's "plumbing," these researchers are moving fMRI from a blurry map of blood flow toward a high-fidelity record of the cortical computations that define human thought.