fMRI Tutorial #4: Preprocessing
Many of the examples are run from the
Flanker/sub-08 directory; I recommend navigating to that directory with your Terminal before reading the rest of the chapter.
Now that we know where our data is and what it looks like, we will do the first step of fMRI analysis: Preprocessing.
Think of preprocessing as cleaning up the images. When you take a photo with a camera, for example, there are several things you can do to make the image look better:
Remove red eye;
Increase color saturation;
Similarly, when we preprocess fMRI data we are cleaning up the three-dimensional images that we acquire every TR. An fMRI volume contains not only the signal that we are interested in - changes in oxygenated blood - but also fluctuations that we are not interested in, such as head motion, random drifts, breathing, and heartbeats. We call these other fluctuations noise, since we want to separate them from the signal that we are interested in. Some of these can be regressed out of the data by modeling them (which is discussed in the chapter on modeling fitting), and others can be reduced or removed by preprocessing.
To begin preprocessing sub-08’s data, read the following descriptions of each step.
- Chapter 1: Brain Extraction (also known as “skullstripping”)
- Chapter 2: The FEAT GUI and loading the functional data
- Chapter 3: Motion Correction
- Chapter 4: Slice-Timing Correction
- Chapter 5: Smoothing
- Chapter 6: Registration and Normalization
- Chapter 7: Checking your Preprocessed Data
- Checkpoint: Preprocessing
When you have finished all of the chapters, click here for a playlist containing all of the videos used to demonstrate the preprocessing steps.
Different software packages will do these steps in slightly different order - for example, FSL will normalize the statistical maps after model fitting. There are also analyses which omit certain steps - for example, some people who do multi-voxel pattern analyses don’t smooth their data. In any case, the list above represents the most common steps that are performed on a typical dataset.