although the mechanism(s) responsible for this is not fully understood. One hypothesis regarding a principal signaling route is that a feedforward pathway involving neuronal-glial interactions after neurotransmitter release stimulates regional CBF. Astrocytes play a crucial role in neurotransmitter recycling, using energy reliant on glycolysis (nonoxidative glucose metabolism) to clear extracellular glutamate and convert it to glutamine after neuronal firing. Increased glycolysis in astrocytes is suspected to trigger intracellular events that couple glutamate cycling rate to the production of vasoactive agents, including nitric oxide and eicosanoids. Therefore, according to this view, neurovascular coupling is mediated by neuronal signaling mechanisms via glial pathways, as opposed to signaling mechanisms of an energy deficit in neurons per se. This view supports, in part, the notion that glycolysis is relevant to the detection of BOLD activity changes and in explaining the apparent mismatch between CBF, CMRglu, and CMRO2 during evoked brain activity (Raichle and Mintun 2006).
Detailed biophysical models have also been proposed to explain the complex shape of the hemodynamic response observed in BOLD fMRI studies, accounting for the changes in CBF, CBV, and CMRO2 that accompany increased neuronal activity. Most prominent is the "balloon model'' of Buxton and colleagues. According to this work, the apparent discrepancy between CBF and CMRO2 results from how oxygen is supplied to neurons related to its poor diffusion in brain tissue. That is, blood flow must increase more than oxygen consumption to maintain tissue-oxygen gradients supporting oxygen delivery to tissue because its extraction (by passive diffusion) from blood is less efficient at higher flow rates (Buxton 2002). Evidence favoring this model versus the former hypothesis (and vice versa) can be found in expanded form in Buxton et al. (2004) and Raichle and Mintun (2006), respectively.
Regardless of the precise cause(s) of the physiological changes that give rise to the BOLD signal, evidence has been marshaled in support of a close relationship between evoked hemodynamic and neuronal activity changes. Notably, in the work of Logothetis et al. (2001), which compared BOLD fMRI and intracranial electrophysiolog-ical measurements recorded simultaneously in monkeys, BOLD signal was found to be spatially well localized and scaled with neuronal activity. Specifically, these authors reported that the amplitude of the BOLD signal was better correlated with recordings of local field potentials rather than multiunit activity (Logothetis et al. 2001). That is, BOLD signal better reflects the weighted average of synchronized activity of the input signals into a neuronal ensemble than their spiking (action potential) activities. This suggests that BOLD signal changes, primarily reflect input and integrative processes rather than output (communicative) activity. However, there remains some debate about the contributions of different types of neuronal activity to the BOLD signal (local field potentials vs. spiking activity), as the former will be correlated with the latter in many instances (Raichle and Mintun 2006).
In a conventional fMRI experiment, ► time-series of T2*-weighted images are acquired while subjects are exposed to a specific stimulus or set of stimuli ("task-on") that is systematically varied with respect to a "control-off" condition, typically in the context of a serial or ► cognitive subtraction or factorial design. The goal of this approach is to evoke significant changes in blood flow and oxygenation within a given region or network associated with the "task-on" state that will modulate BOLD signal intensity about its mean value. The duration of these stimulus presentations or epochs must be tailored to the dynamics of the hemodynamic response, and will be repeated multiple times to establish sufficient contrast and functional signal to noise ratios for the mapping of "activation" responses. In practice, the magnitude of task-related changes in fMRI studies is small (up to ± 5% but usually less) in comparison to the total image intensity and variability across time due to various sources of physical (MR system) and physiological noise. Careful experimental design and the use of post-processing methods for maximizing the detection of activation in the BOLD time series is therefore a critical feature of fMRI studies (Chapter 8-13, Huettel et al. 2004).
One common approach is to take advantage of the summed signal as a way of minimizing the influence of noise in fMRI experiments (Fig. 4). The idea here is that the BOLD response summed over several trials will reduce the influence of random noise sources as a result of averaging (Brown et al 2007). Blocked designs that present the same class of stimuli on multiple occasions seek to capitalize on this strategy. In turn, this involves selecting the correct number and timing of stimuli to occur within a block, the duration of the block itself and its number of repetitions, as well as the number of different block types to be included in a single acquisition for later comparison. Overall, block designs are powerful in terms of detecting significant sustained (steady-state) activation in fMRI studies but are generally poor estimators of the time course of the regional hemodynamic response to neuronal events because of their reliance on linear summation of individual responses.
Event-related designs are a second common approach in fMRI experiments and involve the presentation of specific stimuli as short duration events in order to detect transient associated changes in neuronal activity. With this approach, each event is separated temporally by an interval ranging from a few to tens of seconds and typically in a random order of predefined range. Investigators typically assume a canonical shape to the hemodynamic response to each stimulus presented and model it as a weighted sum to consecutive stimuli - although this linearity assumption may not hold, especially for the early phase of the hemody-namic response. Compared to blocked designs, event-related designs are superior in investigating the shape of regional hemodynamic responses and to compare features such as amplitude or relative timing differences between events. Event-related designs also allow for the investigation of BOLD responses sorted by response types, for instance comparing correct versus incorrect or fast versus slow responses. By comparison, their detection power is relatively poor with respect to blocked designs due to the fewer number of events that can be presented and averaged in a single experimental run.
It was previously stated that the BOLD fMRI time series is influenced by a number of sources of physical and physiological noise. In the former case, this includes system noise that causes fluctuations in MR signal (e.g., signal drift) due to magnetic field inhomogeneities and other factors. In the latter case, this includes gross head motion artifacts, motion related to the cardiac (beat-to-beat) and respiratory (breath-to-breath) cycles, as well as slow variations in respiratory rate and volume, which change the pressure of arterial CO2 - a potent vasodilator. Awareness of these various noise sources in BOLD fMRI studies has led to a range of methods to reduce or mitigate their influence, which continue to be improved upon and refined.
Magnetic Resonance Imaging (Functional). Fig. 4. Linearity of the hemodynamic response: events leading from the presentation of two stimuli to the generation of a summed BOLD signal. To a certain degree, the BOLD response to successive neural events can be predicted from the summed responses (superposition) to single neural events given an appropriate time delay between them. The figure assumes that stimulus 1 is presented 1 s before stimulus 2 and each stimulus evokes a response that alters CMRO2, cerebral blood flow, and CBV. The net effect of these physiological changes leads to the BOLD response for individual stimuli (a and b). The observed BOLD response (c) is a composite of the unobserved BOLD responses to the single stimuli. (Reproduced with permission from Brown et al. 2007. © 2007 Springer Netherlands.)
Was this article helpful?