average_response/FBA questions

Hi guys,

I am trying to compare DTI data from one single case (lesion) with data from a group of controls using FBA.
I reached step 5 of the protocol, normalising the intensity for all cases.
I calculated the response function for every case (tournier) but i have problems averaging the responses getting this error message
(multi-shell response functions must have the same number of coefficients per b-value (line).
The b-values of the control group and the lesion case are different, is that a problem?
Any suggestion of how i should compare the data?

Thank you


:scream: yes, that’s a big problem! I can’t see how you could do such a comparison, no matter which analysis you decide to use. The b-value is the fundamental contrast determinant, if you were to find differences between the two groups, there’s no way you could rule out the difference in acquisition as being the source of the difference. It might be OK if you have a mix of both b-values across both groups, in roughly equal proportion - but if they’re different across groups, I really can’t see how such a comparison would stand up to any form of scrutiny.

So I was struggling to understand how this could be the case given that the ‘tournier’ approach produces single-shell responses - but given the above, it’s clear that the different b-values have the potential to introduce differences in the spherical harmonic order fitted. Either the number of directions is different (?), or the DW gradient table in one case is suboptimal, leading the algorithm to drop to a lower order to avoid fitting instabilities. Either way, it’s bad in that it indicates inconsistencies in the acquisition between subjects within your groups, which is not good…

That said, I do think we should probably alter that particular message to better reflect the fact that the number of coefficients is not consistent across subjects here, not across b-values.

Thank you for your help