F. Di Salle, T. Scarabino, F. ESPOSITO, A. Aragri, O. Santopaolo, A. Elefante, M. Cirillo, S. Cirillo, R. Elefante
Since its introduction in 1992, fMRI has proved so sensitive to the detection of functional phenomena, flexible in use and reliable in its results, and has evolved so rapidly in such a short period of time, that it has become by far the most widely used imaging technique for studying brain function in humans.
Among the main reasons for the astonishing success fMRI has enjoyed so far, a pivotal role has certainly been played by its high sensitivity in analysing brain functional phenomena and its lack of biological inva-siveness, resulting in an unprecedented and unparalleled flexibility of use . Within the few years since its introduction, fMRI has produced a huge amount of information on brain functional anatomy, contributing to the unravelling of many basic mechanisms of brain physiology, the clarification of details of the hierarchical organization of functional areas and the unveiling of the dynamics of parallel processing of information in the living brain.
In brief, the introduction of fMRI has had a revolutionary effect on neuroscience research, since most of the knowledge we have on the functional organization of the brain has been rewritten, modified or at least complemented by new concepts and innovative information coming from fMRI.
As an example, the detailed information we have concerning the functional organization of the human visual cortex has been largely gathered from fMRI experiments, which have enabled the occipital cortex to be divided into a complex constellation of functionally different activities segregated into multiple neighbouring visual areas .
Fundamental to the success of fMRI have been its high spatial and temporal resolution, which are clearly favourable compared to the other methods of functional neuroimaging. In fact, brain functional phenomena are analysed with higher spatial and temporal resolution by fMRI than by the neuroimaging methods using radioactive tracers [1, 2], and with a far higher spatial resolution, compared to electrophysiological measures.
Due to these advantages, fMRI can produce more localized information, and allows a better understanding of the neural dynamics at the level of single functional areas and of the neural constituents of functional patterns.
The spatial resolution of fMRI, which normally ranges within a few millimetres, can be increased to further improve its analytical potential. This is very simple to achieve, since it requires only a modification of the combination of field of view (FOV) and acquisition matrix in the direction of an increased matrix or a reduced FOV. Nonetheless, these changes as a side effect produce a linear decrease in the signal to noise ratio (SNR) paralleling the reduction of voxel dimension (and, thus, the quantity of protons per voxel), which in turn can greatly impairs image quality, practically nullifying the advantages of raising spatial resolutions above a reasonable compromise level.
While this practical limitation in the possibility of increasing spatial resolution is true for any application of MRI, it is especially evident and particularly detrimental for fMRI experiments. Here, the decrease in voxel SNR cannot be compensated for by increasing the number of signal averages, as is generally achieved in structural MR imaging, since this would produce a dramatic reduction in temporal resolution.
Furthermore, fMRI images are not considered in isolation, as is the case for structural imaging, but contribute collectively to a post-processing statistical analysis, which can be considerably impaired by a high level of noise. Since a low SNR thwarts the statistical analysis to such an extent that no reliable results are obtained at the commonly used statistical thresholds, a pre-processing step of spatial smoothing is usually applied in the presence of noisy data. Smoothing the data, in turn, practically implies a loss of the original spatial resolution.
Hence, especially in fMRI applications, the only real way to take advantage of the increase in spatial resolution is by the concurrent use of acquisition strategies able to improve SNR.
Besides the possibility of improving the intrinsic pulse sequence SNR, and the use of high SNR coils, the most successful solution for gaining SNR consists of using high magnetic field intensity fMRI units.
It can be claimed that the spread of high-field MRI units for human studies throughout the world has been appreciably stimulated, if not in fact determined, by the need for an acceptable SNR in functional studies at high spatial resolution.
Furthermore, a number of ancillary reasons suggest the use of high-field units: many fMRI applications have been proposed that may lead to interesting and insightful perspectives in understanding fine neural mechanisms at the cost of decreasing the statistical power of the fMRI studies. It is the case, for instance, for single-trial event-related studies or other demanding fMRI paradigms, which can be beyond the reach of fMRI studies at lower field intensity due to the insufficient statistical power when carried out in unfavourable conditions of SNR .
An important reason for pursuing higher spatial resolution fMRI examinations also comes from neuro-physiological considerations. In order to gain a comprehensive understanding of brain functions, the observation of the behaviour of single neurons, as permitted by the recording of the electrical activity of single neuronal units, is not enough. Nor can it be fully satisfactory to study the functional properties of large macroscopic structures such as whole cytoarchitectonic areas. The study of the primary sensory areas of primates has taught us that many fundamental neural dynamics take place at a dimensional scale intermediate between the level of single neurons and the level of functional areas, dwelling in specific laminar, columnar and mul-ticolumnar domains. Hence, much key information about brain functions is contained in the spatial dimensions ranging beyond the level of a single neuron, where the dynamics are fragmented and incomplete, and below the level of Brodmann's areas, where many neural dynamics can be blurred and confused.
fMRI has a very high potential for analysing brain functional phenomena at this intermediate level, which are thoroughly accessible only by increasing spatial resolution up to and beyond the millimetre range.
Nevertheless, as with any imaging method, fMRI has intrinsic limitations in spatial and temporal resolution, which are due to different factors but which are intimately interdependent.
Was this article helpful?