Yes, 30 DW directions at b=1000 is very much at the lower end of what would be acceptable for CSD. That said, it really depends on your SNR. You might find that if you have good images, CSD does a decent job. I’ve certainly managed to run it on 12 DW direction, b=1000 data, and while the output wasn’t great, it still gave remarkably decent tractography, considering the data. Not that I’d necessarily recommend using this kind of data, but if you absolutely have use these data, I think it can be done relatively well.
Looking at your data though, there’s two issues that immediately come up: one is your gradient table is not right: it looks like the y and z axes have been switched. Your corticospinal tract shows anterior-posterior directions (green) rather than the expected inferior-superior (blue), and vice-versa for the superior longitudinal fasciculus. You’ll want to double-check where this information came from and fix it… If the whole analysis was performed entirely within MRtrix3, using DICOM data as input, then there might be a bug in our DICOM handing (although that’s been very stable for a long time now). Let me know if you think there might be an issue here that we need to fix.
The other issue is that I’m not seeing fibres crossing in the expected places. This will probably be due to the response function being too fat - it certainly looks a bit too ‘blobby’ in your screenshot. I think @thijs has already suggested a few things in that respect on the old mailing list. We have indeed found that
dwi2response has not performed up to expectations, and have put a bit of effort into fixing it recently. These changes are pending in the
updated_syntax branch on GitHub - unless you feel confident switching branches on git and amending your installation (things have changed a fair bit on that branch…), I’d recommend you hang on for a week or two while we tidy everything up and update the main version.