Event-Related Averaging Specification

ERA Spec

Event-related averaging is an important tool in order to obtain the fMRI response profile of different conditions in a single voxel or a larger region-of-interest (ROI). The mean time course for a condition is simply obtained by averaging all peri-stimulus time course segments belonging to the same condition. For block designs and slow event-related designs, event-related averaging is an ideal tool because individual condition-related time course segments are temporally separated. For rapid event-related designs, event-related averaging is not appropriate because condition responses overlap in time due to close spacing of events. In the latter case, deconvolution analysis can be used to disentangle overlapping responses.

BrainVoyager QX provides the “Event-Related Averaging Specification” dialog to define the details on how to perform event-related averaging for a given experiment. Here I consider one relevant parameter for averaging, namely the definition of the size of the time course segment used for averaging and display. Peri-event time course epochs are defined with two values in BV, one for the number of time points before (“Pre” value) and one for the number of time points after (“Post” value) the onset of an event. If these values are too small, the obtained response time courses might not be visible completely. If these values are too large, the resulting averaging plots might include time segments at the beginning and end of the plot, which are a mixture of other events occurring before or after the target event. In previous versions of BV, it was often necessary to repeatedly enter the dialog and to adjust the “Pre” and “Post” values until the right size for the peri-stimulus segment was found.

For the upcoming BV QX 1.7, I have added a simple tool, the “Expected response plot”, which makes the selection of the right window size easy and intuitive. When you click on the name of a condition, an expected ideal response profile is calculated immediately and displayed in the “Expected response plot” panel. The performed calculation is based on the duration of the included conditions and the TR value. This information is retrieved from the functional data files included in the averaging process and the stimulation protocols linked by these files. A standard two-gamma function is used to calculate gradual time courses reflecting the hemodynamic response characteristics. For each selected condition, a different time course is plotted, each in the condition’s color as defined in the referenced stimulation protocol files. With the expected response plot, it is easy to find the right “Pre” and “Post” values since the plot is immediately updated when the respective spin boxes are used.

For visual purposes, conditions are shown with different amplitudes, which are generated randomly in an interval from 0.2 - 1.0. These randomly selected amplitudes can be interpreted as a simulation of different voxels with different response strengths for the selected conditions. If you want to change the selected amplitudes, simply click somewhere in the expected response plot panel. I interpret this as looking at a different voxel’s response profile.

After specifying the “Pre” and “Post” values and optionally other parameters in the dialog, you may generate an AVG file as usual by clicking the “Create AVG” button. Then simply select the saved AVG file from a “ROI Signal Time Course” window to get true event-related averaging plots shown in “Event-Related Averaging Plot” windows. It might be instructive to compare the predicted with the calculated averaging plots. For a future release, I plan to add the possibility to show predicted response plots and calculated average plots together within “Event-Related Averaging Plot” windows.

As a closing remark, I would like to point out that the “Event-Related Averaging Specification” dialog has been improved also in BV QX 1.6. In that release I added a little feature, which automatically adds functional files to the “Functional data files” list box. The program tries first to add all functional files, which are referenced in a currently available (computed or loaded) GLM. This is a reasonable approach since most people want to run event-related averaging over the same data, which is also used for GLM analyses. If you have computed a GLM with many subjects and runs, you will probably like this feature because it saves you from adding all runs manually, which not only takes some time but is also error-prone. If the program does not find a GLM, it puts the currently linked functional data file, if available, in the list box. The automatic adding tool as well as the expected response plot tool work with all three functional data types, FMR, VTC and MTC. Event-related averages are also calculated much faster than in previous versions due to some code optimizations.