fMRI Primer

10 Preparing to run FEAT

Pre-FEAT tips

For most individual subjects, you will conduct at LEAST four FEAT analyses, and more likely dozens. First you will preprocess the subject’s data. Second, you will analyze each experiment they completed. Third, you will often perform a within-subject analysis, as when you compare results from two runs at different times (eg before and after treatment, mood manipulation, learning). Fourth, you will perform a group analysis across all subjects in a group.
In reality, you will use FEAT many more times. After an initial analysis you will likely want to experiment with changing a regressor, changing a smoothing kernel, changing a statistical cutoff threshold, and a thousand other possibilities. In short, you will use your results to guide subsequent analysis. In this way using FEAT to mine signal from your data is an iterative process. Moreover, because group runs using FLAME take an especially long time, you will often run FEAT twice, once using an ordinary least squares (ols) analysis, and then, once this ‘works,’ a FLAME analysis.
1. The cornerstone of a FEAT analysis is the creation of an .fsf file. This is created using a FEAT GUI (graphical user interface). This requires two types of work: the grunt-work involved in finding the files to enter in the various fields – a process made much easier if you use excellent record keeping and data structure, as described above – and extremely subtle conceptual work, in which you determine which regressors to apply to the matrices.
2. Please note that the first time you run a FEAT do it as a fast-and-dirty analysis – as an OLS ordinary least squares analysis. In 20 minutes you will know if your program is going to work or not.
Please note you will need your fMRI plan sheet, to specify the number of volumes, the TR, and that sort of thing.
Creating a Log Book

The most important investment you can make in your FEAT process is, from the start, to keep a simple log of what you did when. This should be computerized in an excel document, and should be constructed like this:

Subject Run Type of run Explanation file name date complete error comments
203 R1 Individual Pre-sadness ES r1NANANANA2EV3C20060126A.feat 2006.04.04
203 R4 Individual 25m post-sadness ES r4ANANANAN2EV3C20060126A.feat 2006.04.04
203 R1-R4 Within-Subject Group Within-subject subtraction wsgFEr4vr120060201.gfeat 2006.04.04
210 R1 Individual Pre-sadness ES r1NANANANA2EV3C20060126A.feat 2006.04.05
210 R4 Individual 25m post-sadness ES r4ANANANAN2EV3C20060126A.feat 2006.04.06
210 R1-R4 Within-Subject Group Within-subject subtraction wsgFEr4vr120060201.gfeat 2006.04.07
203,210 Group R1-R4 Group test of group run tempr1r4203210 worked!!!

By constructing this table in Excel, you can sort the lists in various ways to see which subjects have been run, which programs have been run; there is automatic fill-in; you can create pivot tables; add regressors in; add columns ad infinitum, and many other advantages to using Excel over Word.
Top Ten FEAT Practical Tips:
1. Always open FSL from the folder on which you are going to do a run. EG, navigate to subject 203 and then open FSL. This will save you lots of typing down the road.
2. One of – if not the – primary advantages FSL has over SPM is this incredible time-saving feature: A copy of the .fsf file for each FEAT run, called design.fsf, is automatically saved in the .feat file produced by the run. This means that while in your master models folder you may save a generic copy of a design.fsf file, to speed processing of numerous subjects, if an individual run doesn’t work, DON’T load the generic again and tailor it again to the individual; go to the individual’s folder, and load their individualized design.fsf from there. This is very easily done by opening FSL from within the folder in which the .fsf is.
3. If FSL returns the error message: “FEAT setup file is too old to load, sorry!” when loading a design file, this is because you tried to load a ‘.feat’ directory and not the design file (.fsf) contained within. Try again.
4. On the DATA tab, remember that getting the name right is crucial. This can hardly be over-emphasized. Your two worst-case scenarios are this: 1) you send your results to a bizarre file that you never remember the name of 2) you name two files the same thing (eg, forget to change the subject number when copying subject A’s design.fsf for subject B.) In the first case you are out of luck; in the second, it will create a file with the same name but a ‘+’ in front of it. Just rename and move tothe appropriate folder. The design.fsf file in the + folder will reveal, upon being opened, just whose data the analysis was run on.
5. If you ever get a wierd glitch in FSL or FEAT, just close the gui you are working in, and then reopen it. This resolves most problems.
6. When entering numbers – especially in the regressor and contrast fields under STATS, don’t ever try to use the numeric keypad. Just use the numbers on the keyboard. Moreover, FSL is ‘wierd’ about the number 1, and writing the number 10 is hard, for example, so you may have to write the second digit first and then the first digit second. It is wierd about decimals too. Enter the number first, then the decimal, and lastly the negative sign where appropriate
Opening FEAT
Opening FSL on a Sun:
1. Open a terminal
2. FSL &

Opening FSL on a mac:
1. open xterm, type ssh –Y pfreed@nimbus
2. Then log in with password fmri759
3. FSL &

Double click on FEAT. You should get the GUI (graphical user interface) below.
{Pete and Ted – you need a screenshot here!}

Here is what a total experiment, in which each subject has two runs that need to be compared in some way, looks like from start to group analysis finish.

FEAT housekeeping flowsheet


Leave a Comment »

No comments yet.

RSS feed for comments on this post. TrackBack URI

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

Blog at

%d bloggers like this: