The information about neuronal phenomena that fMRI extracts from the BOLD effect represents the convolution of the temporal envelope of local brain activity with the so-called haemodynamic input function .
The general effect of this convolution consists in the blurring of the functional information in space and in its dispersion in time. Thus, as is the case for the spatial resolution, the limit of temporal resolution is generally set by factors intrinsic to the local brain haemodyna-mics, and specifically by its time constants, rather than by technical factors. So, even if a theoretical time resolution of a few tens of milliseconds is easy to reach with modern fMRI gradient systems, the dispersion in time of functional information operated by the local haemo-dynamics does not permit the discrimination of two successive co-localized brain phenomena occurring at time intervals separated by less than 3-4 s . This does not hold true in the case of the initial dip: paradoxically, this early functional phenomenon takes advantage of the haemodynamic delay in order to be separable from the large positive BOLD signal change. The initial dip has been interpreted as being due to an „inverse uncoupling" between metabolism (regional cerebral metabolic rate for oxygen - rCMRO2) and flow (regional cerebral blood flow - rCBF). This early uncoupling, caused by an increase in rCMRO2 in the presence of an unchanged rCBF, can be considered the exact opposite of the later, better known, uncoupling which produces the common BOLD signal through an rCBF increase prevailing over a much less pronounced rCMRO2 change. As a result, the early uncoupling increases the regional deoxygenated haemoglobin and decreases the BOLD signal, while the later uncoupling produces opposite effects giving rise to the positive BOLD response .
As discussed above, the examination of the transient increase in deoxyhaemoglobin producing the negative signal change of the initial dip is controversial, even if it is very interesting since it overcomes some of the intrinsic limitations of the spatiotemporal resolution of the BOLD effect. The controversy, involving both fMRI and optical imaging [19, 20], originated from the failure of several laboratories to detect the effect in studies carried out at 1.5 T, probably determined by both the intrinsically reduced sensitivity of lower field strength units, and by a low contrast-to-noise ratio due to the reduced extent of the temporal window where the effect is visible . Even if the initial dip poses considerable problems in its use as a general approach for high-resolution fMRI, it appears clear that the dip is the most promising way actually available for this purpose. In fact, the dip corresponds much better than the positive BOLD effect to the localization of neural phenomena as described by the gold standard of electrical activity recording . If these concepts are accepted, then high magnetic field strength units are to be pursued as the only way of detecting the „elusive" initial dip.
With regard to the „positive" (conventional) haemo-dynamic response, advanced post-processing tech niques such as „deconvolution"  or temporal independent component analysis (ICA)  are able to extract temporally more sophisticated functional information from high temporal resolution fMRI time-series. Provided that a reasonable signal-to-noise ratio in the functional time-series is present, particularly with the use of high magnetic field strength, a decomposition in the time domain can produce interesting neuro-physiological information. In the auditory domain, an ICA decomposition in time and space has been successfully used for the qualitative and quantitative spatiotemporal analysis of BOLD dynamics in the neural processing of sound information [24, 25].
In „true" event-related studies conducted at high temporal resolution, the responses to each repetition of the task generate an fMRI signal which can be retrospectively reduced to a small set of fitted haemodyna-mic parameters like the onset or duration of the responses. These parameters, when correlated with variations in the experimental performances of the subjects [3, 26, 27] or compared across different areas for single  or averaged trial responses , enable the investigation of BOLD temporal dynamics on the temporal scale of hundreds of milliseconds. Also in this application, the essential requisite is to operate at a sufficient signal-to-noise ratio, as is usually permitted only by the use of high-field units. This is particularly true if the analysis is aimed at producing superior neurophysio-logical results, which generally require the simultaneous achievement of high spatial and temporal resolution.
A major technological improvement in MR hardware is represented by the use of multiple radiofre-quency receiver coils to independently encode spatial information in parallel from multiple regions (parallel imaging [30, 31]). These techniques are capable of accelerating data acquisition, allowing high temporal resolution functional imaging. Theoretical considerations have shown  that the maximum acceleration factor consistent with an acceptable SNR increases linearly with main magnetic field strength.
Was this article helpful?