Can I combine across different acquisition protocols in one TBSS/FA/etc analysis?
In general it is dangerous to combine different protocols (e.g., different dMRI resolutions, b-values, number of directions, etc.). DTI metrics such as FA are generally not "biologically quantitative", and so most protocol changes are likely to change measures such as FA. However, if you combine different protocols that are not too different, and if your design of interest (e.g., patients vs. controls) is not highly correlated with the different protocols, then you may be OK. This means that the design matrix EVs of interest (e.g. one for patients, one for controls) should not be highly correlated with confound EVs (e.g., one for one acquisition protocol, and one for the other).
I hear that TBSS v2 is in the works...?
We have been working to see if we can replace the skeleton projection step with a more accurate (whilst also being regularised), fine-detailed nonlinear registration. It looks like this is working (see [de Groot 2013]), so we hope to have an all-new version in 2015.