Interpretation of fiber density values

Dear experts,

I have been performing fixel based analysis for a bit now, and even though the results always make sense and look reasonable, I still have some fundamental questions about the meaning of fiber density.

After reading the paper “Quantitative streamlines tractography: methods and
inter-subject normalisation” (DOI: 10.52294/ApertureNeuro.2022.2.NEOD9565) written by some of the MRtrix3 pioneers, I interpret the meaning of fiber density as follows: within a voxel, a certain proportion of the volume of that voxel is filled with white matter bundles. Multiple white matter bundles can pass through a voxel, which is why these bundles are assigned to different fixels within a voxel. So each fixel has a fiber density value based on the estimated volume that is taken up by the white matter bundle belonging to that fixel.

In the images I’m working with, I have fixels that have a FD higher than 1 (e.g. FD = 1.3). The resolution of my images are 2.0 x 2.0 x 2.0 mm, but during preprocessing the images are upsampled to 1.25 x 1.25 x 1.25 mm. Intuitively I would think FD values are between 0 and 1 if it represents the proportion of a voxel volume filled with white matter bundle, but that doesn’t show in reality. Does the FD value depend on the resolution of your images and how does upsampling influence this?

Theoretically, what is the maximum FD value in the case of 2.0 x 2.0 x 2.0 mm, would that be a FD of 8 mm³?

Thank you!


1 Like

Until someone more knowledgeable comes around, allow me to directly quote from Dhollander et al, 2021, DOI 10.1016/j.neuroimage.2021.118417 :

While apparent FD is approximately (linearly) proportional to intraaxonal volume, it doesn’t provide a direct absolute or standardized volume measurement of it. CSD techniques in this context are applied to the dMRI signal without voxel-wise normalisation by the b=0 image (Raffelt et al., 2012b), unlike most other dMRI modelling techniques. Not only is apparent FD expressed in arbitrary units, it also requires correction for spatial intensity inhomogeneities (bias fields) of the dMRI data as well as some form of global intensity normalization to render it comparable between different subjects within a study (Raffelt et al., 2012b, 2017; Dhollander et al., 2021).

The entire article is very much worth reading for anyone getting involved with FBA.