functional Magnetic Resonance Imaging
Functional Magnetic Resonance Imaging (fMRI) is a tool to measure brain activity non-invasively via hemodynamic changes in the brain. Whenever neurons fire, they require oxygen, which results in an initial decrease of oxygenation followed by an oversupply of oxygenated blood to those regions. Differences in these alterations are measurable by MRI scanners as the so-called blood-oxygenated-level dependent (BOLD) signal. Unfortunatly, the BOLD signal differs from the underlying neural activity, as it is much slower and blood volume dependent. To derive the neural activity from the BOLD signal a linear Hemodynamic Response Function (HRF) is often being applied. This model is only an approximation of the relationship between the BOLD signal and the neural activity. In fact, how those two are related is still up for debate. fMRI displays the deviations of the BOLD signal as image intensities in cubic pixel space. One pixel is called voxel and its size defines the spatial resolution of the image. During fMRI multiple images or scans are taken to capture the whole brain with low spatial but high temporal resolution in comparison to structural MRI images. This means, that the images turn out blurry. However, the acquisition time, also called repetition time (TR), is thereby highly reduced, making it possible to trace the activity of one single brain over the time of a whole experiment. Unfortunately, fMRI does not only capture the BOLD signal but also noise and nuisance parameters. Those can arise for example from measurement errors or thermal fluctuations of the participant. Differentiating between the BOLD signal and the noise is not an easy task to do. Fortunately, filters or models of nuisance signals can help to decode the signal. This is one reason why the captured images need to be preprocessed (more in Preprocessing) before statistical analysis. During statistical analysis a General Linear Model (GLM) can be constructed in order to derive the task-dependent neural activity from the BOLD signal. Thereby, different conditions are modelled to cause different signal fluctuations. The impact of each condition, can be contrasted using different test statistics in order to create so-called contrast images. These contrast images are then used to answer hypotheses, either on single subject level or summarizing the results across a group of participants. Often statistical analysis is also restricted to a specific area of the brain, Regions Of Interest (ROI). This can reduce the problem of multiple comparison.
If you are new to fMRI Data Analysis in general, please have a look at these sources to get a deeper insight:
Soares JM, Magalhães R, Moreira PS, Sousa A, Ganz E, Sampaio A, Alves V, Marques P, Sousa N. A Hitchhiker's Guide to Functional Magnetic Resonance Imaging. Front Neurosci. 2016 Nov 10;10:515. doi: 10.3389/fnins.2016.00515. PMID: 27891073; PMCID: PMC5102908.
Chen, J.E., Glover, G.H. Functional Magnetic Resonance Imaging Methods. Neuropsychol Rev 25, 289–313 (2015). https://doi.org/10.1007/s11065-015-9294-9