I'm looking into setting up a MVPA analysis pipeline for my fMRI studies, and am weighing up options. So far I've found:
PyMVPA: http://www.pymvpa.org/
The Princeton MVPA toolbox: https://code.google.com/p/princeton-mvpa-toolbox/
Or PRoNTo: http://www.mlnl.cs.ucl.ac.uk/pronto/
I do most of my analysis with FSL, so would probably be leaning more towards the first two (PRoNTo seems designed to work primarily with SPM). Am reasonably comfortable with either Matlab or Python.
Does anyone have experience with any of these, and have any feedback, in terms of user-friendliness, features, or any other factor? Or are there any other similar toolboxes available that I should consider?
Many thanks!
Hi Matthew,
you may also consider the nicely set-up toolbox by Martin Hebart and Kai Görgens, which you can donload from sourceforge (after contacting the authors). Check his website http://martin-hebart.de/webpages/code/toolbox.html . It works nicely and gives you control over many features if you "dig" a bit. Very clever guys. It should be said that is written to work with SPM, I think. Not a problem for me, since I use SPM mainly.
And of course there is scikits-learn (http://scikit-learn.org/stable/) which offers a HUGE variety of algorithms and stuff. I think they improved the data supply, that is how to feed brain imaging data into the pipeline. If I am not mistaken the thing to look for is nipy (http://nipy.org/) or nipype (http://www.mit.edu/~satra/nipype-nightly/index.html). These tools are very powerful and most useful for somebody who is experienced in programming. You are in perfect control of what is happening. Very active community and clever people behind this project (e.g. INRIA).
We tried Pronto but were very frustrated since it remained quite obscure. It works but you do not really know what it's doing.
No experience with PyMVPA.
Hope this helps.
André
Hi Matthew,
you may also consider the nicely set-up toolbox by Martin Hebart and Kai Görgens, which you can donload from sourceforge (after contacting the authors). Check his website http://martin-hebart.de/webpages/code/toolbox.html . It works nicely and gives you control over many features if you "dig" a bit. Very clever guys. It should be said that is written to work with SPM, I think. Not a problem for me, since I use SPM mainly.
And of course there is scikits-learn (http://scikit-learn.org/stable/) which offers a HUGE variety of algorithms and stuff. I think they improved the data supply, that is how to feed brain imaging data into the pipeline. If I am not mistaken the thing to look for is nipy (http://nipy.org/) or nipype (http://www.mit.edu/~satra/nipype-nightly/index.html). These tools are very powerful and most useful for somebody who is experienced in programming. You are in perfect control of what is happening. Very active community and clever people behind this project (e.g. INRIA).
We tried Pronto but were very frustrated since it remained quite obscure. It works but you do not really know what it's doing.
No experience with PyMVPA.
Hope this helps.
André
I suggest you Brain Voyager, it's easy to use and using matlab you can solve many future problems.
I am in your same situation and, although I use FSL, I was told I could use PRonTo anyway. Is this correct? I thought I would be able to use the nifti files created as part of my filtered_func_data.nii.gz.
Isabelle, you should be able to use the toolbox, if you unzip the nifti (so the extension is just nii).
You may also find success with Tor Wager's CANLab tools (specifically the predict.m function), if you are looking for a more general machine learning tool that retains spatial information.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python