Thursday, August 1, 2013

AFNI Start to Finish Playlist

Although the title may sound a little risqué, what it actually refers to is a compilation of tutorials - twenty-one in all - that cover the analysis of a single subject from data import to viewing the results and making publication-quality pictures. I closely follow a script already up on the AFNI website written by Rick Reynolds, and I've included links to the relevant step of the script. The idea here was to actually show what each step looks like, and to provide some additional commentary. Half of the commentary is wrong, though; the only problem is, I'm not sure which half.

A link to the script can be found here; a link to the tutorials can be found here. Particularly important is understanding the design matrix, and gaining an intuition for how it is applied at each voxel; I will be going into more detail about that in the future.


  1. Hi Andrew,
    I am very new in fMRI and in my research I am trying to find high performance solutions for fMRI problems using parallel computation techniques. I am wondering can I use synthetic fMRI datasets in my research rather than actual datasets? since I need to test my approaches on multiple fMRI datasets with varying size of voxels and multiple number of subjects (e.g. 5000 subjects). I am generating random float numbers between -6 to 6 as intensity of each voxel. Is it a good idea? if it is not is there any source for already preprocessed fMRI datasets for ver large number of subjects?
    Thank you

    1. Hi there,

      Yes, you can use synthetic datasets; several researchers use them to test whether certain tools can distinguish between signal and noise. However, I wouldn't base all of your research on that; it would be hard to say much about the brain with only simulated datasets.

      I haven't created fMRI datasets, so I don't know what a good range of values is. If you want to be in the range of most published fMRI studies, I would generate values between -2 and 2.

      Good luck!