Blood Oxygen Level Dependent functional magnetic resonance imaging (BOLD fMRI) is a brain mapping technique using deoxyhemoglobin contained in the blood vessels as an endogenous contrast agent to produce functional activation maps[1
Neural activation induces a transient increase in regional oxygen extraction from the blood that is coupled with a much larger increase in cerebral blood flow (CBF) and cerebral blood volume (CBV). This influx of oxygenated hemoglobin results in a net decrease in regional deoxyhemoglobin concentration. This drop in paramagnetic deoxyhemoglobin concentration leads to an increase in the magnetic relaxation times T2 and T2*. The mapping of eloquent areas is thus obtained by acquiring T2 or T2*-weighted images consecutively while the subject is at rest or performing a task, and detecting the signal increase related to the local reduction of deoxyhemoglobin concentration accompanying the functional activation relative to baseline.
fMRI studies are carried out using a gradient echo sequence with echo planar readout[2
]. This sequence is very sensitive to the static local magnetic field inhomogeneities and therefore is suitable for detection of T2* related signal changes, and at the same time allows scanning of the whole brain within one repetition time (TR) of 2-3 s. The subject performs a cognitive, sensorimotor or visual paradigm inside the bore of the MRI scanner during image acquisition. In a typical clinical fMRI paradigm, a block design is used consisting of 20-30 s blocks (a.k.a., epochs) of rest or control tasks alternating with blocks of active tasks. In this way the subject’s brain activity is usually monitored for a total duration of 200-300 s by acquiring whole brain MR images with a good spatial (typically 3 mm × 3 mm × 3 mm) and temporal resolution (e.g. 2 s). Four dimensional (3D+time) datasets are created where a signal time series is recorded and stored for each voxel. Functional activation maps are obtained by detecting areas of the brain showing a statistically significant signal increase due to the executed active task relative to the resting or control task.
To achieve this goal, a series of processing steps needs to be carried out on the fMRI dataset[3
]. First the raw images are initially time shifted so that all the slices in each whole brain acquisition (volume) that occur in one TR result as if they were acquired at the same time. Then all the volumes are registered to a reference volume to correct for small head motions. A further preprocessing step consists of spatially smoothing each voxel signal time series in order to reduce low and high frequency noise.
At this point a statistical analysis is carried out that aims to determine, voxel by voxel, the “goodness of the fit” of the signal time series to a theoretical hemodynamic response function obtained by convolving the paradigm timing with an impulse response function (Figure ). The “goodness of the fit” can be expressed through several statistical parameters, such as P-value, Z or t score or cross correlation coefficient.
Figure 1 Magnetic resonance signal time series (black line) in an activated voxel. Subject was engaged in a language phonemic task over a time period of 260 s. The red line curve represents the ideal hemodynamic response function the voxel time series was fitted (more ...)
Activation maps are generated choosing a threshold (significance) on the statistical score. The suprathreshold regions are visualized as “hot spots,” often overlaid in color and coregistered on a higher resolution anatomical MR image (Figure ).
Figure 2 Activation (colored voxels) in the Broca’s and Wernicke’s areas in a normal subject performing a phonemic fluency task. The t-score map was thresholded at 4.0 value (P < 0.001) and superimposed on a high resolution T1 weighted (more ...)
The neurobehavioral paradigms used for functional MRI studies can be divided into two categories, block-design and event-related design. Block design paradigms, which are more commonly used clinically, utilize consistent and repeated blocks of stimuli (active task) and rest (control task), often of the same duration.
In the event-related paradigm design single events are used as stimuli instead of epochs of consecutively administered multiple stimuli. Each trial is considered separately as being time locked to the beginning of the stimulus, and signal changes are explored in relation to the onset of the event generated by the trial.
Block design paradigms provide higher sensitivity for detection of statistically significant signal changes between the control and active conditions as well as allow for better patient compliance, and for these reasons are generally preferred for clinical fMRI studies[4
Introduced in the early nineties by Ogawa et al[5
], BOLD fMRI has become an extensively used imaging technique in the neuroscience community.
BOLD fMRI has been applied to the study of a broad spectrum of brain functions from simple motor and visual activities to complex language, memory and emotion and even higher-level reasoning tasks such as abstract mathematical reasoning[6
In the last decade it has made the transition from a purely research imaging technique to a viable clinical technique used primarily for presurgical planning in patients with brain tumors and other resectable brain lesions.