Background
http://brainvoyager.com/bv/doc/UsersGuide/StatisticalAnalysis/TimeCourseNormalization.html
For Group ICA analysis, fMRI data intensity needs to be scaled before PCA. In GIFT gICA toolbox, there are 4 ways to normalize intensity:
%% Data Pre-processing options
% 1 – Remove mean per time point
% 2 – Remove mean per voxel
% 3 – Intensity normalization
% 4 – Variance normalization
preproc_type = 4;
icatb_preproc_data.m ( icatb_runAnalysis.m >> icatb_dataReduction.m >> icatb_calculate_pca.m)
%% Intensity normalization
tmp = repmat(100./(mean(tmp) + eps), size(tmp, 1), 1) .* tmp;
%% Variance normalization
tmp = detrend(tmp);
tmp = tmp.*repmat(1./(std(tmp) + eps), size(tmp, 1), 1);
Explanation in GIFT Manual (p18):
’Select Type Of Data Pre-processing’ – Data is pre-processed prior to the first data reduction. Options are discussed below:
’Remove Mean Per Timepoint’ – At each time point, image mean is removed.
’Remove Mean Per Voxel’ – Timeseries mean is removed at each voxel.
’Intensity Normalization’ – At each voxel, time-series is scaled to have a mean of 100. When intensity normalization is selected as the pre-processing step, don’t use Z-scores or percent signal change for scaling components. (i.e. each voxel value divided by its mean across time, *100)
‘Variance Normalization’ – At each voxel, time-series is linearly detrended and converted to z-scores. (i.e. each voxel value divided by its std across time, *100)
p106:
Note: When using one common data reduction step on multiple subjects, data for each subject needs to be normalized using intensity normalization or variance normalization. Also, higher number of components needs to be estimated from the data when compared to two data reduction step.
References:
Click to access 2010_OHBM_Elena_ICAPrenormalization_submitted.pdf
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1263605
Group ICA workflow:
https://onlinelibrary.wiley.com/doi/full/10.1002/hbm.21170