Tensor2dwi to compute sse map

Is there a way to convert back tensors to dwi to be able to compute sse (sum of squared errors) ?

Thank you

Hi Arnaud,

I can’t think of an existing way in which this could be achieved without writing some code. It wouldn’t be a particularly difficult program to write though. If you have some coding ability and would like to give it a go, I’d be happy to point you in the right direction for making use of the MRtrix libraries. Otherwise, we can add an issue to the GitHub tracker as a requested feature.


Hi Rob,

It’s more a question of time these days. I think it would be better to add an issue, when I’ll have time I’ll look into it.


Hi @arnaudbore ,

I’ve just added an option to the existing dwi2tensor to output the predicted dwi signal from the fit. If you update your MRtrix (git pull and ./build), you can now use the -predicted_signal option of dwi2tensor.

Note that, with dwi2tensor, you can also fit the kurtosis tensor (-dkt) if you have multi-shell data (b=0 and 2 or more other b-values). The -predicted_signal quite nicely shows how you really need to go beyond DTI for those kinds of data. :slight_smile:


A post was split to a new topic: Synthesis of DWI volumes from tensor