Neurovascular Coupling & the BOLD Signal

Active neurons call for a blood delivery

Neurons have almost no local fuel reserve — they rely on a continuous supply of glucose and oxygen from the blood. When a patch of cortex becomes active, its energy demand rises within milliseconds, and a coordinated signal recruits more blood to that exact patch within a couple of seconds: nearby arterioles dilate, local blood flow rises, and blood volume in the area increases. This chain of events, from firing neuron to expanded blood supply, is neurovascular coupling, and it is mediated by a "neurovascular unit" of neurons, astrocytes and the blood vessel wall, with nitric oxide as one of several key vasodilating signals released as part of the response.

Why the signal overshoots

Here is the counterintuitive part that makes fMRI possible at all. Active neurons consume more oxygen, which by itself would increase local deoxyhemoglobin — and deoxyhemoglobin is paramagnetic, distorting the local magnetic field and darkening the MRI signal. But the blood flow increase triggered by neurovascular coupling delivers oxygenated blood in far greater excess than what the tissue actually consumes. The net result is that the concentration of deoxyhemoglobin in the activated region actually drops, and the MRI signal in that region brightens. This blood-oxygen-level-dependent brightening is the BOLD signal that essentially all task-based fMRI studies measure.

live demo · blood flow rising and relaxing after a burst of activity● LIVE

The balloon model

The standard quantitative model of this process, introduced by Buxton and colleagues in 1998 and extended by Friston and colleagues into the widely used balloon-Windkessel model, treats the local venous compartment as an elastic balloon inflated by inflow and drained by outflow that depends nonlinearly on how full the balloon already is:

dv/dt = ( f_in(t) - f_out(v) ) / tau         // blood volume balloon

f_out(v) = v^(1/alpha)                        // outflow grows with fullness

dq/dt = ( f_in(t) * E(f_in)/E0 - f_out(v) * q/v ) / tau   // deoxyhemoglobin

BOLD(t) ≈ V0 * ( k1*(1 - q) + k2*(1 - q/v) + k3*(1 - v) )

Volume v inflates when inflow f_in exceeds outflow, and outflow grows faster than linearly once the balloon is already fairly full — a viscoelastic property borrowed from the physical Windkessel model of arterial compliance. Deoxyhemoglobin content q rises with oxygen extraction but is washed out by outflow, and the final BOLD signal is a weighted combination of how much blood volume and deoxyhemoglobin have changed from baseline, with weights (k1, k2, k3) that depend on the scanner's field strength and pulse sequence.

Why the signal lags and overshoots

Because the model has two coupled state variables (volume and deoxyhemoglobin) with different time constants, the BOLD response to a brief burst of neural activity is a slow, smooth bump — the canonical hemodynamic response function — that peaks roughly 4 to 6 seconds after the neural event and does not return to baseline for 15 to 20 seconds, vastly slower than the millisecond timescale of the underlying spikes. A well-documented post-stimulus undershoot — the signal dipping slightly below baseline after the main peak — falls directly out of the model too: blood volume relaxes back to baseline more slowly than blood flow does, so for a few seconds deoxyhemoglobin is elevated relative to the still-inflated balloon, darkening the signal briefly before everything settles.

What this means for reading an fMRI map

Because BOLD is a hemodynamic proxy rather than a direct electrical measurement, its temporal resolution is fundamentally limited by the multi-second timescale of the vascular response, no matter how fast the scanner itself can sample. It also means BOLD amplitude reflects a mix of neural activity and vascular reactivity, which can differ between brain regions, individuals, and health conditions — a caveat every fMRI study has to account for when comparing signal strength across sites or populations.

Frequently asked questions

Does fMRI directly measure neural activity?

No. Standard fMRI measures the BOLD signal, which reflects local changes in blood oxygenation caused by neurovascular coupling — the brain's response to increased neural activity, not the neural activity itself. It is an indirect proxy, delayed by several seconds and blurred in space by the vascular response, in contrast to methods like EEG or intracranial electrodes that record electrical activity directly.

Why does BOLD signal increase with neural activity if deoxyhemoglobin also increases?

Because the vascular response overcompensates. Active neurons consume more oxygen, which would increase deoxyhemoglobin and darken the BOLD signal, but the accompanying increase in blood flow delivers oxygenated blood in far greater excess than what is consumed, diluting deoxyhemoglobin concentration. The net effect is a paradoxical increase in the BOLD signal despite higher local oxygen consumption.

What causes the initial dip and post-stimulus undershoot sometimes seen in BOLD signals?

The initial dip, seen in some high-field studies, is attributed to a brief rise in deoxyhemoglobin before blood flow has had time to increase and dilute it. The post-stimulus undershoot, a well-replicated dip below baseline after the response ends, arises in the balloon model from blood volume relaxing back to baseline more slowly than blood flow does, leaving deoxyhemoglobin transiently elevated relative to volume for several seconds.

Try it live

Everything above runs in your browser — open Neurovascular Coupling and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.

▶ Open Neurovascular Coupling simulation