I'm currently working with a set of high-resolution (spatial and angular) diffusion images from paraformaldehyde-fixed post-mortem brains. I learned, from past discussions on this list about Probtrackx, that the recommended step length for fiber tracking is 1/4 of the voxel size. However, tractography runs using this step length and a very large number of steps (I've tried up to 500000), compared with those computed using the voxel size as the step length, yield very poor results: while the larger step length (= voxel size) gives nice, long tracts that enter subcortical structures and the brainstem in a way that makes anatomical sense, the smaller step length (= 1/4 * voxel size) only produces rather short tracts are significantly less extensive. Increasing the number of steps and/or the maximum number of steps do not improve the results, as I used pretty big values for both parameters to begin with. At this point, I'm very tempted to go ahead and use the larger step length.

So my question is, are there any parameters I can change to improve my results at the step length of 1/4*voxel size? If not, what is the maximum step length I could use without significantly sacrificing the accuracy of my tracts?

Thanks in advance.

-Thomas