|
Notes:
Minor bugfix and feature release
Name change! Package/scripts now simply GTG (METAlab_ was
removed)
Changes:
Name change for all scripts (i.e., removed METAlab_)!
All changes were to Stage 1 (preprocessing path)
New features:
1) Added option to specify the minimum # of voxels required to be
in an ROI (if >= that #, the timeseries is replaced with NaNs),
previously this was hard coded
2) Added option to specify the minimum % of voxels retained after
masking by the functional mask required for an ROI (if >= that
%, the timeseries is replaced with NaNs), previously this was hard
coded
3) Added option to use existing nii files (with correct name)
instead of recomputing every file every time
4) Added safeguard whereby matlab will test whether text files
(i.e., containing white matter, ventricular, or global signal or
DVARS) already exist and, if so, add a number to the filename of
newly created files in order to distinguish them and avoid
confusion
Bugfixes:
1) Fixed bug whereby the threshold to assign NaNs during ROI
extraction was >90% of ROI removed during masking instead
of >50% (but logfile still said 50%)
2) Fixed a bug whereby the mean was used to extract white
matter/ventricular signal after the second round of processing
when scrubbing (i.e., after removing bad time points) even if
PCA was selected (PCA was used during the first round)
Other changes:
1) Updated to use the preferred version of matlab's pca script
2) Allowed PCA extraction of ventricular signal when using
local white matter processing
3) Changed assignment of sign/direction of signal when
using PCA to extract each ROI's timeseries;
previously the sign/direction was determined by PCA
alone; after conducting simulations, it was
determined that PCA accurately reflected the
sign/direction of underlying signal 90-95% of the time (SVD
was accurate only ~50% of the time), whereas the
mean always accurately captured the sign (PCA
more accurately captured the underlying
signal variance); therefore, the script now checks the
correlation between the mean and largest principal component
and, if it is -1, the PC is multiplied by -1
|
