Dear FSL Experts,
I have a group of 672 subjects with various age, from middle age to old age. These subjects will be used in a group ICA and then in a dual regression to get the default mode network in each subject. However, when running group ICA, it splits the DMN into two sub-networks, where a linear combination of the two components (e.g. an average) looks most similar to the DMN. Could it be that the DMN is not consistent within the group, given the age range of the subjects, and that this may result in the finding of two DMN components? Melodic was set to automatically estimate the dimensionality and this resulted in 20 components.
What kind of options do I have for running dual regression? Should I use the average of the two components or should I try to rerun the group ICA with less dimensionality in the hope of finding only one DMN component? Another option I can think of is to rerun the group ICA on a subset of subjects (e.g. by only using the youngest subjects) in order to make the group more homogeneous. However, then I’m unsure if it is OK to use the resulting DMN component for dual regression against the older subjects since they were not part of the construction of it?
Many thanks for your help!
Best regards,
Lars Forsberg
|