Thanks you for answering me. With these output files I have not enough information because I can't check outliers before correction and outliers after correction. I can not perform a comparison.

I have been designing a Matlab script to get outliers from a DWI volume. Using fslstats with -S and -M options I get volume standard desviation and mean. I consider a voxel as outlier if that voxel has a value higher than mean+2*std or less than mean-2*std. Dou you consider that method is appropiate? Should I change the range [mean-2*std,mean+2*std]?

when it comes to detecting outliers in diffusion images you should always use some sort of model based approach (even if that “model” happens to be just a Gaussian process). The difference in intensity, for a given voxel, between two different volumes with different diffusion gradients can be very substantial. Especially for high b-value data. So comparing it to the mean will never do a good job. 

Take a look for example at RESTORE, which compares the observed intensity to a prediction from a tensor model

or the detection in eddy that compares it to the prediction from a Gaussian Process.

Jesper

Thanks you very much. I appreciate your opinion.

Best,

M del Mar