Biomedical Engineering Reference
In-Depth Information
single-cell electrical recordings, electroencephalogram (EEG),
magnetoencephalogram (MEG), positron emission tomography
(PET), and most recently, functional magnetic resonance imag-
ing (fMRI). To date, however, these techniques are either invasive
or limited in their ability to accurately localize neural activities in
space or in time. Even when the latest multimodal neuroimag-
ing methodology is used, there remain fundamental limitations
that introduce sources of error and interpretative difficulties. For
example, the simple combination of EEG (which has a high tem-
poral resolution) and fMRI (which has a high spatial resolution)
does not generally provide an unambiguous delineation of the
spatiotemporal sequence of functional brain activity.
Among all the neuroimaging methods currently available,
fMRI
(1-4)
has experienced a particularly explosive growth in
recent years. The noninvasive nature of magnetic resonance imag-
ing (MRI), along with its high apparent spatial resolution and
moderate temporal resolution, rapidly engendered its emergence
as one of the dominant techniques in functional brain research.
Just about all of the fMRI contrast mechanisms developed thus
far, however, have relied on indirect measures of neuronal activity.
For example, the widely used blood oxygenation level dependent
(BOLD) contrast relies on relative oxygenation changes, perfu-
sion contrast relies on cerebral blood flow (CBF) changes, and
cerebral blood volume (CBV) contrast relies on task-induced ves-
sel expansion. Yet, none of the activities that these techniques
measure are themselves thought to mediate information process-
ing in the brain, but rather are hemodynamic ramifications of the
neuronal activity itself. The hemodynamic modulations inevitably
disperse the observed functional signal change both spatially and
temporally, despite the high apparent spatiotemporal resolution of
these techniques
(5, 6)
. Ideally, one would want to bypass these
indirect markers and measure the activity of neurons directly.
Motivated by the many advantages of MRI ideal for investigat-
ing brain function, researchers in the MR community have thus
begun exploring and extending its use with the goal of being able
to directly image transient neuronal activities.
Thus far, however, this type of direct MRI technique, albeit
theoretically conceivable and probably the most intriguing, has
still remained largely unattainable. The goal of noninvasive detec-
tion of neuroelectric activity with high spatiotemporal resolution
has been extraordinarily challenging in the context of neuroimag-
ing, as the electrical activity of neuronal tissue is extremely weak
and the imaging voxels quite small. Indeed, even direct intracra-
nial recordings of evoked local field potentials, which collectively
acquire electrical signals from an area on the order of 1 cm
2
,
only measure potentials on the order of a few hundred microvolts
and require hundreds of time-locked signal averages. Further,
the electrical activities are also temporally transient and spatially
