What is fMRT?
Functional magnetic resonance imaging – a simplified presentation of its basic principles
Since the early nineties, a new method measuring functional parameter of the cortex has been developed besides already established procedures (PET, SPECT, MEG, and EEG).Functional magnetic resonance imaging (fMRI) enables us to measure cortical reactions on external stimuli with a spatial resolution superior to the previous procedures. Applying suitable techniques it is furthermore possible to map the identified activated areas to the respective anatomic structures. While early studies used contrast agents, this has become unnecessary. Instead, local blood oxygenation is the crucial parameter which impacts the intensity of the signal.
This phenomenon, also known as BOLD (blood-oxygen-level dependent) contrast, was described by Ogawa et al. in 1990.This procedure uses the change in magnetization between oxygen-rich and oxygen-poor blood as its basic measure. This change is due to the iron bound to the hemoglobin. In the oxygenated form, the magnetic properties of the iron are masked, that is, it is diamagnetic. In deoxygenated hemoglobin (dHb), on the other side, the iron atoms have their full magnetic properties, that is, they are paramagnetic. This paramagnetic dHb causes inhomogeneities in the local magnetic field which can be measured.
Stimulating areas of the Cortex leads to an increased metabolism rate, and the activated brain area shows a higher blood flow. Subsequent to activation, more oxygen is transported into the activated area than can be consumed by the neural activity. As a consequence, the oxygen level in the capillary bed increases and the proportion between oxygen-rich and oxygen-poor hemoglobin changes. The reduced proportion of paramagnetic deoxygenated hemoglobin reduces also the magnetic inhomogeneities in the respective area. This again leads to an increased signal in the T2* weighted magnetic gradient image in the activated area of the brain.
The BOLD response to a stimulus develops in time, the hemodynamic response function. This function depends strongly on the conditions of the simulation and must be taken into account when planning the measurement. Shortly after the onset of the stimulation there is a short decrease in the function which is probably due to a local oxygen depletion. In the following, the increased blood flow induced by the stimulation causes normally an increase of the signal starting about 1-2 seconds after the onset of brain activity. It may well take up to 10 seconds until the maximum of the signal is reached. The subsequent decrease of the signal is also strong and usually undershoots the original signal level for a few seconds.
When planning fMRI experiments it is important to consider that stimulus presentation and response execution are difficult to disentangle due to the sluggishness of the BOLD response. Thus, the stimulation frequency has to be low enough to discern single stimulation events. A source of artifact are also inflow effects. These inflow effects can overlay the actual activation and can thus distort the measurement.
In an fMRI experiment often periods in which areas of the brain are activated using suitable paradigms (e.g., finger-tapping) alternate with resting periods. The differences of the signal in these two periods in the individual pixels of the image can be used to create an activation map of the cortex.
With a field intensity of 1.5 Tesla, these differences of the signal have a size of about 3-5%. For such small differences, statistical evaluation procedures are suitable which are based on differences between activation and resting images (Z-score, student t-test) or which correlate the timecourses of the measured activation with an appropriate model function of the activation (e.g., rectangle or sine). Partitioning the paradigm into alternating activity and recovery stages facilitates the elimination of random artifacts which normally do not have such a periodicity.
The subsequent image shows an example for a simple block-design paradigm with four activity and five recovery blocks. Each block takes 40 seconds giving a total time of measuring of 360 sec. During this time, 45 scans of the whole brain are made.
For the analysis of fMRI data, also more complicated mathematical procedures are used, e.g. fast Fourier transformation, factor analysis, or neural networks. The activation effects measured by fMRI are small and overlaid by artifacts. Therefore, stimulus repetitions are necessary to detect a response statistically. To increase sensitivity for complex cognitive stimuli given the limited time of measurement, an enhancement of the signal-noise ratio is desirable. Noise can be due to technical or physiological reasons as well as to movements of the subject during the measurement (including physiological movements like swallowing). Therefore, the first step of the data analysis is movement correction for all images following the first image.
For identifying the activated brain areas anatomically, a high-resolution anatomical image is overlaid with the functional MR image. It is also possible to transfer the data into a common coordinate system which facilitates the anatomical evaluation significantly. For functional investigations of the brain, the brain mapped by Talairach & Tournoux (1988) has proven very useful. For evaluating patient data, however, this transformation should be applied with caution because of distortions of the functional images (overregularization). In these cases, manual procedures applied by fMRI experts can be superior.
Compared with other methods for functional investigation of the brain like PET or SPECT, fMRI has many advantages. It is non-invasive, there is no radiation exposure, and the spatial and time resolution is significantly higher.
Research using fMRT thus is not only able to confirm earlier results obtained with other methods but also yielded signficant new insights into the working of the brain.