 |
 |
Dear Gustaf, > I posted this question during the summer and was met by a compact > silence. I'm goint o give it another > try, so let me know if you have any ideas, or if my question is too > obscure. > > I have a question concerning the choice of the best target in the > registration step of tbss (tbss_2_reg). > I was having some trouble with the registration of a set of subjects > and in an attempt to bone out the > kink, I put together a matlab script that performs the individual > registrations between subjects using > fsl_reg (I have not rewritten the algorithms, just executed them in > the terminal using matlab), in > accordance with what is stipulated in tbss_2_reg. From the resulting > warp (.msf) files, I extracted > a best target. We evetually got tbss working and performed the > registration with the -n option and > to my surprise obtained a different target. > > I have attached two plots with individual comparisons of 'all > comparisons' for the matlab (blue) and > tbss (red), as well as the averaged coefficients for each subject. > > What is being performed differently when I loop in matlab instead of > a shell script? If anyone has > some input on what I have missed, I would be very grateful. I am afraid that from this description it is nearly impossible to say what might have happened. There might be a bug in our TBSS scripts, or there might be a bug in your Matlab script. Since the TBSS scripts are widely used and we haven't had any reports of something like this before I am afraid I am inclined to believe the latter. To debug it I would suggest hacking both scripts to calculate the warp coefficient for a single pair (one that gives significantly different results in your previous run) and pin point what the difference is there. Hopefully that will give you enough information to solve the problem, or alternatively to return to the mailbase with a more specific question. Sorry I couldn't be of more help. Jesper > > Best regards, > > Gustaf Mårtensson > > Astrid Lindgren Children's Hospital > [log in to unmask] > <individualmatlabandfsl.png><allcomparisonsmatlabandfsl.png>