I want to know if there is a way to know what the value of the proportionality constant mu (as mentioned in the papers) is after running SIFT/SIFT2
If you’ve run tcksift
, this will be reported as a key-value entry “sift_mu
” in the track file’s header, visible through tckinfo
. For tcksift2
, you don’t get this since no track file is generated (in the coming update there’s an additional command-line option to write this value to a text file, which can come in handy). The value does however get reported if you run either command with the -info
flag (both before and after filtering in the case of tcksift
, and just once for tcksift2
since it remains fixed).
So what I’ve done in the past is run tcksift
with the -info
and -nofilter
flags: That’ll redo the FOD segmentation and track mapping, print the value of the proportionality coefficient, but won’t actually remove any streamlines.
I worked my way backward, using the TD
and FD
from after_tdi.mif
and after_target.mif
to compute mu
. But this value always turns out to be 1. Why is it so?
Because after_tdi.mif
is pre-scaled by u
.
Is “Track density scaled
” nothing but mu.TD
where TD
is the same as “Track density unscaled
”? If so, I should be able to get at mu
just by a simple operation.
Correct - if you’ve generated that file already.
If this is indeed the case, the mu
that I get from after_scatterplot.csv
should match with what I get from after_diff.mif
. But this doesn’t happen. Am I missing something here?
Suspect this is a continuation of the point above?
Also, what does the field “Weight
” correspond to?
This was expressed as PM
in the manuscript. I called it a “processing mask” early on and stuck with it, but in retrospect it’s a bad name because the influence of these data is non-binary. Different fixels contribute more or less toward the model on a continuous scale depending on the (estimated from 5TT image) partial volume contamination.