Homec4science
Diffusion Research Scripts (Thomas Bolton) (master) Browse

Scripts_Sharing/ISFC toolbox/Utilities/ISFC_thresholding/JOVE_FindExcursions.m
1c5bab09dbdamaster

Last Change
Raw File

JOVE_FindExcursions.m
View Options

%% This function computes a distribution of values for the considered
% null data (for each connection pair), and determines (in
% two-tailed fashion) the thresholds past which an excursion of
% connectivity in the data is significant. Then, those moments are returned
%
% Inputs:
%
% - Data and Null1 are the actual and null data (cell arrays n_subject
% with size n_pairs x n_sliding_window_timepoints) acquired during the
% stimulus sessions
% - Null2 contains the purely resting-state recordings, as opposed to the
% resting-state recordings that were acquired within a stimulus-including
% session
% - percentile (in percent) is the level of significance
%
% Outputs:
%
% - Out is a 3D matrix of size n_conn x n_slidingwindow_timepoint x
% subjects, filled with -1 (significant negative ISFC change), 0 (no
% significant change), or +1 (significant positive ISFC change)
function [Out,PCT] = JOVE_FindExcursions(Data,Null,percentile,ThreshType)
if(size(Data,1) ~= size(Null,1))
errordlg('The null data and the actual data do not have the same dimensions !!');
end
% Number of region pairs
n_pairs = size(Data{1},1);
n_TP = size(Data{1},1);
n_subjects1 = length(Data);
n_subjects2 = length(Null);
% Aggregates the null (i.e., resting-state) data
Null_marginalized = [];
for s = 1:n_subjects2
Null_marginalized = [Null_marginalized,Null{s}];
end
switch ThreshType
case 'Normal'
% Does the percentile computations
for s = 1:n_subjects1
PCT{s} = prctile(Null_marginalized,[100-percentile, percentile],2);
end
case 'Gaussian'
% Alternative: fit a Gaussian distribution to the marginalised null and
% sample at a better alpha-value using it
[Mu_hat,Sigma_hat] = normfit(Null_marginalized');
for s = 1:n_subjects1
PCT{s} = [((Mu_hat + repmat(erfinv(percentile/100),1,n_pairs))./Sigma_hat)',((Mu_hat + repmat(erfinv(1-percentile/100),1,n_pairs))./Sigma_hat)'];
end
otherwise
errordlg('problem...');
end
% Creates and fills the output, thresholded matrix Out with +-1 at the
% moments when there was a significant ISFC change, for a given subject
% and connection
Out = zeros(n_pairs,size(Data{s},2),n_subjects1);
for s = 1:n_subjects1
for i = 1:n_pairs
if sum(Out(i,Data{s}(i,:)>PCT{s}(i,1),s)) > 0
disp('!!!!!!');
end
Out(i,Data{s}(i,:)>PCT{s}(i,1),s) = 1;
Out(i,Data{s}(i,:)<PCT{s}(i,2),s) = -1;
end
end
end
c4science · Help