BOLD signal arises from changes in deoxyhemoglobin (dHb) concentrations

dHb changes due to changes in oxidative metabolism, blood flow and blood volume

Calibrated fMRI can be used to extract oxygen metabolism information from the BOLD signal

Calibrated fMRI requires a blood-gas manipulation and measurements of cerebral blood flow and BOLD signal

Based on a family of biophysical models of dHb dilution

Can be used to measure task-evoked and resting metabolism in absolute units of ml O₂/100g tissue/min

BOLD signal arises from the fact that oxygenated and deoxygenated hemoglobin have different magnetic properties. While oxyhemoglobin is diamagnetic, deoxyhemoglobin is paramagnetic. The paramagnetic nature of dHb means that blood that is not fully oxygenated causes an attenuation of the T2* signal. Since arterial blood is almost fully oxygenated in healthy humans, venous blood is darker than arterial blood (1).

During neuronal activity, oxygen consumption increases. This creates an increase in local dHb concentration and therefore a decrease in BOLD signal. However, at the same time, vasodilation occurs to bring fresh blood into the area. Since this influx of blood is larger than the increase in oxygen consumption, this creates a net decrease in dHb concentration and therefore an increase in BOLD signal. This is why the BOLD signal we measure following neuronal activity is an increased signal. This vasodilation also results in an increased local blood volume (2). This increase in blood volume spans both the arterial and venous component. Increased venous blood volume is associated with an increase in local dHb concentration, which decreases the BOLD signal. Therefore, the BOLD signal we measure is a combination of decreased BOLD signal due to increased oxygen consumption and increased venous blood volume, and a larger fractional change in blood flow which brings fully oxygenated blood to the area and causes an increased BOLD signal.

Because of these opposing contributions, the BOLD signal is not only and indirect measure of neuronal activity, but also a an ambiguous one. When we compare a BOLD signal between different people or different conditions, we assume that the BOLD signal means the same thing in terms of neuronal activity. However, if this balance between the different sub-components is changed, the BOLD signal is no longer linked in the same way to neuronal activity and BOLD signal changes can no longer be compared directly. This is thought to be particularly problematic in studies of aging and diseases, such as cardio- or neurovascular diseases, dementia, etc (3-6). Other techniques such as Arterial Spin Labeling (ASL) to measure flow (7), VAscular Space Occupancy (VASO) to measure blood volume (8) and calibrated fMRI to measure oxidative metabolism (9, 10) can however be used to get around some of these issues.

Calibrated fMRI techniques have been developed to extract the oxidative metabolism component from the BOLD signal measured in response to a task. Oxidative metabolism is isolated by first estimating the vascular component of the BOLD response through a calibration manipulation. Once estimated, this vascular contribution can be used to estimate the signal attenuation from the metabolic increase in deoxyhemoglobin (dHb) during performance of a task. Such a determination requires expressing the activation-induced BOLD signal as a fraction of the total attenuation of T2*-weighted signal attributable to dHb at baseline. The calibration procedure is typically done by measuring the BOLD and CBF responses during a blood-gas manipulation. Using a biophysical model of the BOLD signal, we can extrapolate from these measurements to an estimate of the calibration parameter M, which corresponds to the maximal possible BOLD signal change that would occur upon removal of all dHb. Once the M parameter has been measured, it can be used to estimate the cerebral metabolic rate of O₂ consumption in absolute units of ml O₂/100g tissue/min. This is done by combining CBF and BOLD responses measured during a task with the M parameter using the model proposed by Davis (9).

Various calibration methods have been proposed in which the maximal BOLD increase M is estimated through extrapolation of BOLD signal increases observed during mild hypercapnia (9), hyperoxia (11), or a combination of the two (12). Current quantitative BOLD approaches thus rely on extrapolative blood-gas manipulation techniques.

In the hypercapnia method, small amounts of carbon dioxide (typically 5–10% CO₂ by volume) are added to the air breathed by subjects during acquisition of BOLD and CBF image series. The vasodilatory properties of CO₂ lead to increases in cerebral blood flow and tissue oxygen delivery, producing BOLD signal increases throughout gray matter as well as in large veins. The maximal BOLD signal M at a given location can then be extrapolated using the Davis model of BOLD as a function of CBF, assuming a constant arterial saturation of 100% and unchanged metabolism during hypercapnia. This is the original calibration method, introduced by Davis et al. and since used in a number of studies (13-15).

In the hyperoxia method, subjects inhale high levels of oxygen (typically 50–100% O₂ with balance nitrogen where applicable) during acquisition of BOLD image series and recordings of expired oxygen concentration (a surrogate for the arterial partial pressure of O₂). The enriched O₂ raises the total oxygen content of the arterial blood, leading to increases in venous hemoglobin saturation SVO₂) and hence increases in BOLD signal. The maximal BOLD signal M can then be extrapolated using the hyperoxia calibration model of BOLD as a function of SVO₂, with an approximated correction for the small decreases in CBF known to occur during hyperoxia (11, 16, 17). Because of the difficulty of measuring small CBF changes with the somewhat noisy ASL method, a fixed value for the hyperoxia-induced CBF decrease is generally assumed.

The Generalized Calibration Model (GCM) is based on an extension of previous models and enables biophysical modeling of BOLD MRI, perfusion MRI, and end-tidal O₂ (ETO₂) responses to arbitrary combinations of hyperoxia and hypercapnia (18). This allows the use of carbogen gas (5-10% CO₂, balance O₂) for measurement of the M parameter, which has been shown result in more accurate calibration (18).

Recently, an alternative method has been proposed that does not require the use of a breathing manipulation for calibration (19). This method uses the information in a spin echo and a gradient echo sequence to estimate R2’ and thereby obtain the M parameter. R2’, also called the reversible transverse relaxation rate, is a component of the R2* signal used to obtain the BOLD contrast and is thought to be dependent on venous blood volume and dHb concentration (19, 20).

One new class of MRI methods utilizes respiratory calibration with multiple gas challenges to quantify tissue Oxygen Extraction Fraction (OEF) and resting CMRO₂ in the brain. Traditional calibrated BOLD techniques have relied on a single isometabolic gas challenge, such as hypercapnia or hyperoxia, to measure relative changes in oxygen metabolism during a functional task. These approaches typically assume a baseline OEF value and estimate the M parameter. Through use of multiple gas manipulations and the GCM, local baseline values of M and OEF are available per tissue voxel (21-23). Several variants of this respiratory calibration approach have been implemented with pure hyperoxia and hypercapnia (22), or multiple combinations of gases with different combinations of O₂ and CO₂ concentrations (21, 23).