How to run a PPI analysis in Feat

Instructions for implementing a PPI analysis with the Feat gui.

See also: Frequently Asked Questions (PPI)

1. Make some decisions

Because you start by choosing and ROI and task contrast, PPI is heavily hypothesis driven. If you don’t have a clear hypothesis about functional connectivity, you might consider using a model-free approach like MELODIC, which also gives you ‘network’ information of a different kind. On the other hand, you can also try running lots of different PPIs with different seed regions until you find something interesting….

Choose your ROI

For a PPI analysis, you must select a seed Region of Interest (ROI) – the point of the analysis is to look for areas which ‘interact’ with this seed region.

There are several types of ROI you might go for:

Choose your task contrast

PPI analysis always looks for voxels with increased functional connectivity to your seed ROI in one condition compared to another. Usually, the contrast will be between conditions within a single scanning session, and you need to decide whether this will be a contrast of one condition vs everything else (a 1 0 contrast) or a contrast between two conditions (a 1 -1 contrast). More details about this are in the FAQ.

However, you may also be interested in comparing between scanning sessions, for example before and after an intervention, such as a drug or brain stimulation - or even betweens groups (e.g. patients vs. controls). A discussion of between-subjects designs can be found here.

2. Prepare your regressors

You need to extract the timecourse from the seed ROI and put it into a format which Feat can read, before using Feat to make your PPI regressor.

Make masks for the seed ROI

You need to make a mask for your seed region in each individual subject’s functional native space (that is, the space of the fMRI image you want to analyse).

Anatomical region of interest

Your options are:

Functional region of interest

Your options are:

This may be a more successful strategy when the functional regions are anatomically heterogeneous but functionally well defined, e.g. in the parietal cortex.

Size of ROIs

If you are defining ROIs individually (as in the second functional-ROI case, or some of the anatomical cases) you analysis will likely work better (have higher signal-to-noise) if you keep the ROI small. This is because you are only taking one measurement from the whole ROI – so by enlarging the ROI to include voxels with a weaker effect you are actually ‘watering down’ the signal.

On the other hand, if you are using a standard-space mask (as in the first functional-ROI strategy), you will want to make sure your ROI is large enough to capture the individual activation peak for each subject, despite inter-individual variations.

Extract the time-course of the seed ROI

Do this for each subject separately, using the fslmeants command.

You should use the filtered func data from your initial analysis to extract the timecourse from - not the raw data as this will be noisy.

The output is a column vector giving a value of raw signal at each time-point; there is one time-point per volume. The time-course is saved under a filename specified by you, for use later on.

3. Set up your Feat design

In the stats tab

You will need the following regressors:

EV1 is your psychological regressor (PSY). This will simply be your task regressor, convolved with an HRF;

EV2 is your physiological regressor (PHYS). This will be the time-course of your seed ROI:

EV3 is PPI, which you generate here in the Feat GUI;

This completes the set-up for the first-level analysis. You can then compare between groups at the second level as normal.


PPIHowToRun (last edited 10:51:51 30-06-2014 by PaulMcCarthy)