If tractograms are to be provided with just streamline endpoints, the .tck format is highly sub-optimal: it stores a NaN triplet as a delimiter between streamlines, so with exactly two vertices per streamline it increases storage by 50%.
Quantitative values per streamline are currently stored in ASCII form. Using binary floating-point (with a guaranteed endianness) would require less space. For fixel data we actually just use .mif images that are 1D, and that permits flexible endianness and memory-mapping; we could do that here, or something different. But either way, it may require some development at the MRtrix3 end to allow using those data as input to tck2connectome.
If tractograms are to be provided with just streamline endpoints, the
.tck
format is highly sub-optimal: it stores a NaN triplet as a delimiter between streamlines, so with exactly two vertices per streamline it increases storage by 50%.Quantitative values per streamline are currently stored in ASCII form. Using binary floating-point (with a guaranteed endianness) would require less space. For fixel data we actually just use
.mif
images that are 1D, and that permits flexible endianness and memory-mapping; we could do that here, or something different. But either way, it may require some development at the MRtrix3 end to allow using those data as input totck2connectome
.