 |
 |
Hi Joel, As Jesper says, it is always best to generate your own templates from your data. However, this can be extremely time/computationally intensive depending on how many subjects you want and the desired level of contrast in your final image, not to mention that you must have correct 'updating' steps for the template at each iteration. Alternatively you can make use of one or more of the age-specific templates that are now available. For instance, see the following publication: Unbiased average age-appropriate atlases for pediatric studies. Fonov et al Neuroimage 2011(This would probably be useful reading if you did want to go the route of making a new template anyway). I hope that's helpful! David On Wednesday, April 9, 2014, Jesper Andersson < [log in to unmask]> wrote: > Dear Joel, > > > We have a large set of child and adolescent brains that we'd like to get > into MNI 2mm atlas space. We've tried to follow steps we've used for > adults (with success) but are running into problems. The issue appears to > be that the lateral ventricles, around areas of the caudate, are so small > that they're only a voxel or two in thickness, in plane. FLIRT tends to > push the caudate into the space of the ventricles in MNI, FNIRT pushes them > further. Is there a way to force interpolation of CSF (enlarge the > ventricles) or to get the gray mater (caudate) to "fit" the corresponding > region in MNI? > > I would suggest considering using a/building your own child standard > brain. Forcing the brains of very young children into the standard brain is > likely to result in suboptimal registration. > > The way to do that would be to start by flirting a population of (similar > age) children into standard space and then iteratively fnirting then to the > average of that population, updating that average after each iteration. > > If you have a large material you might even want to create several > standard brains, say for infants, toddlers, children and adolescents and > then also create the mappings that take you from one of these to the next. > > Good luck Jesper > > > > > We've tried playing with parameters in the .cnf file for fnirt but it > tends to result in the computer locking and the process being killed. > > > > See attached image of an example brain (raw image on left, target slice > in MNI, and affine/nonlinear transforms on the right). > > <adolescent_fnirt_problem.jpg> >