correlation_analysis_in_fmri_data.md

title	tags
Correlation analysis of fMRI data	example fmri raw freq coherence

Correlation analysis of fMRI data

This script demonstrates how FieldTrip can be used for time-series analysis of fMRI data.

Get the data in a "raw" data structure, similar to preprocessed MEG data

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% contruct the projection matrix for the ROIs
load aal_new_mat
aal = ft_read_mri('aal_new.hdr')
project = zeros(length(ROI), prod(aal.dim));
for i=1:length(ROI)
  disp(i);
  project(i,:) = double(aal.anatomy(:)==ROI(i).ID);
  project(i,:) = project(i,:)./sum(project(i,:));
end
project = sparse(project);

mri = ft_read_mri('bwamarspr_rs_sess01_vol0001.nii');

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% read the data and put into a FT structure
session = {
'sess01'
'sess02'
'sess03'
'sess04'
'sess05'
'sess06'
'sess07'
'sess08'
'sess09'
'sess10'
'sess11'
'sess12'
};

raw = [];
raw.fsample  = 1/1.35;
raw.label    = {ROI.Nom_L}';

for s=1:length(session)
path = session{s}
filename = dir(path);
filename = {filename(~[filename.isdir]).name}';
for i=1:length(filename)
  filename{i} = fullfile(path, filename{i});
end

dat = zeros(length(ROI), length(filename));
for i=1:length(filename)
  disp(i);
  tmp = ft_read_mri(filename{i});
  dat(:,i) = project * tmp.anatomy(:);
end

raw.time{s}  = (0:(size(dat,2)-1)) ./ raw.fsample;
raw.trial{s} = dat;
end % for session

save project project -v6
save mri mri -v6
save raw raw -v6

Apply frequency analysis, compute frequency specific correlation/coherence

cfg = [];
cfg.blc = 'yes';
cfg.keeptrials = 'yes';
timelock = ft_timelockanalysis(cfg, raw);

winsize = 64;
fr = (256/winsize)./timelock.time(end);

cfg = [];
cfg.method     = 'mtmwelch';
cfg.output     = 'powandcsd';
cfg.taper      = 'hanning';
cfg.channel    = 'all';
cfg.channelcmb = {'all', 'all'};
cfg.foi        = 0:fr:(raw.fsample/2)
cfg.toi        = raw.time{1}((winsize/2):winsize:256);
cfg.tapsmofrq  = nan * ones(size(cfg.foi));
cfg.t_ftimwin  = winsize/raw.fsample * ones(size(cfg.foi));
cfg.keeptrials = 'no';
cfg.keeptapers = 'no';
cfg.pad        = 'maxperlen';
freq = ft_freqanalysis(cfg, timelock);

cfg = [];
% cfg.complex = 'abs';
cfg.complex = 'real';
fd = ft_freqdescriptives(cfg, freq);

cfg = [];
cfg.layout = 'ordered';
lay = ft_prepare_layout(cfg, fd);

cfg = [];
cfg.layout = lay;
cfg.interactive = 'yes';
cfg.showlabels = 'yes';
figure; ft_multiplotER(cfg, fd);

cfg = [];
cfg.layout = lay;
cfg.interactive = 'yes';
cfg.showlabels = 'yes';
cfg.zparam = 'cohspctrm';
cfg.zlim = [0 1];
cfg.cohrefchannel = fd.label{1};
figure; ft_multiplotER(cfg, fd);