Tensor2metric RD, ADC, AD not giving good values



Hello everyone,

I am using tensor2metric to generate RC, ADC, and AD maps. However, when I plot them on FSLeyes, the maps pretty much look like binary masks with different shades of gray for each of the values. However, the FA map looks good and how it is supposed to.

I know this is a similar post (Tensor2metric works not well to give negative RD, AD and RD inside the brain), but I wasn’t sure how to go about fixing this problem.

Thanks in advance for the help.

Tensor visualisation

This could be due simply to these maps containing both very small and very large values, maybe in non-brain regions or places where the fitting is poor. Depending on how FSLeyes figures out the intensity range to display, this might look like what you’re describing. If that’s the case, I’d expect to see more or less the same thing in mrview – what does that look like? In both cases, it would be a simple matter of adjusting the windowing (on mrview: right-click & drag, or use the View tool to set the min/max of the intensity range; not sure for FSLeyes).

If that’s not the issue, maybe you can post a screenshot to give us a clearer idea of what the problem is?

Dwi2tensor wrong values

You were right. It’s funny because I played with the intensity range, but for my case, it was very small numbers between -0.001 and 0.001.

Thanks for the help!


Yes, ADC values should be in units of mm²/s, since the b-value is provided in s/mm². So the range of intensities should be 0 - 0.002 or so. When the fitting goes wrong in non-brain regions, it can give rise to completely unrealistic values, way outside of the expected range. This leads to problem in the display since the intensity range is by default set to min/max, which is not great in this case.

I’m pretty sure the idea of using more sophisticated strategies to set the intensity range has been discussed before, using a histogram approach or using e.g. the 10-90 percentiles as min/max, which would go a long way to avoiding this type of issue (can’t find the thread though…). But as always, it’ll take effort, and is hard to prioritise given that it’s non-essential…