Problems with amp2sh

Hi everyone!

I would like to view a amp2sh file I just created, but I am having problems with it.

The problem is that everything seems rotated of 90 degrees.

Here is what I obtain when I open the file:

And this is what I should get:

Does someone know why I am getting this strange output and how to correct it?

Thank you,

Hi @Carlotta_Fabris,

There’s nothing wrong with the top one: those are SH representations of the DWI signal (not FODs), which has large magnitudes perpendicular to the main fibre orientation(s).

As the bottom one is in fact labelled “FOD.jpg”, it’s the bottom one that has a severe mistake in it: note the orientation of those (supposedly intended to be FODs) is wrong in e.g. the corpus callosum and the corticospinal tract, indicating a problem with the gradient table (a combination of flips of axes and permutations of columns is typically the cause of such a problem).


Thank you a lot for your answer Thijs!

So my output is not wrong? Can I continue like this?

Also the image I put as correct one was taken from the internet, just to show you what I thought I had to obtain, so probably I took a wrong image!

Thank you,

It’s not wrong, as in: that is definitely the correct output of amp2sh. The question though is what you’re ultimately after (i.e. how you see yourself continuing from this output onwards)… Were you genuinely after an SH representation of the DWI signal? Or are you after FODs? (because the signal is something very different from FODs)

If you’re confused about this, I’d definitely encourage you to read up a bit on the literature (e.g. read Donald’s spherical deconvolution and CSD papers), before proceeding…

What I am after is interpolating the data we have. We started form DSI and used Dsi Studio to interpolate them.

Now we would like to select just the points we want and where we want them. To do this, we passed to SH and then we will choose the points and go back.

Thank you,

Sounds alright then, as long as you take into account that it’s an SH fit of course. Definitely make sure you’re providing amp2sh with a dense enough sampling, and potentially choose the lmax wisely.