PPI - What is it?

From Rebecca:

Another thing, regarding the more conceptual idea of a PPI, is the fact that the stimulus is already being essentially represented in the brain area you extracted as the way we perform a random effect analysis is by seeking the regions which matches our task best. The only way I could think that adding the stimulus again would help is by analogising it to a multiple regression.

Psychophysiological interaction is used to measure the response of one area to another during a task or stimulus, rather than the interaction of these regions with a common task or stimulus (Friston et al., 1997). This is performed by extracting the timetrend during the stimulus from a region of interest and multiplying this extracted vector by the original stimulus input heamodynamic response curve. By doing so, the multiplied curve produced will have features of the region selected and of the stimulus itself. From this, an analysis resembling resting state connectivity is performed, by which brain regions with a similar response to the stimulus to our multiplied curve (‘seed’ region) will be sought after. This produces a functional connectivity which reflect both the stimulus and the nature of BOLD signal change in a region of the brain, adding psychophysiological data to the original stimulus.

This form of analysis has similarities to a multiple regression, as it allows us to investigate all important factors in one model, leading to a (hopefully) more precise connectivity analysis. In terms of a multiple regression the x variables are the extracted timetrend from the seed region and the stimulus/task heamodynamic response curve and the y variable being the functionally connected brain region that we seek.