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

Scripts_Sharing/ISFC toolbox/Utilities/ISFC_computations/Perform_Bootstrapping.m
1c5bab09dbdamaster

Last Change
Raw File

Perform_Bootstrapping.m
View Options

%% This function performs the ISFC bootstrapping scheme on a specific type
% of session for a group of subjects
%
% Inputs:
%
% - TC contains the time courses of all sessions to jointly compute ISFC on
% in a cell array format (each cell of size n_ROI x n_timepoints)
% - Session_Label contains the index of the session of each recording; this
% is used to include an equal number of samples from each session in the
% reference group of the bootstrapping scheme
% - W is the window length (in TR)
% - S is the window step size (in TR)
% - n_boot is the number of bootstrapping folds to perform
% - n_refsubj is the number of subjects to use in the reference group at
% each fold
function [ISFC] = Perform_Bootstrapping(TC,Session_Label,W,S,n_boot,n_refsubj)
% Number of sliding window estimates that we will have according to W
% (window length) and S (step size)
n_swtp = floor((size(TC{1},2)-W)/S)+1;
n_ROI = size(TC{1},1);
n_conn = n_ROI*(n_ROI-1)/2;
% Session indices present in this call
sesvals = unique(Session_Label);
% ISFC will contain the final ISFC estimates for all sessions
% (one per cell)
ISFC = cell(1,length(Session_Label));
[ISFC{:}] = deal(zeros(n_conn,n_swtp));
% Denom_factor will highlight, for each session, how many times ISFC
% was computed on it (i.e., how many times it was not part of the
% reference group out of the n_boot folds)
Denom_factor = cell(1,length(Session_Label));
[Denom_factor{:}] = deal(0);
% tmp_denom will temporarily contain the information for a given fold,
% with 1 highlighting sessions on which ISFC was computed
tmp_denom = cell(1,length(Session_Label));
[tmp_denom{:}] = deal(0);
% idx_allsessions contains the indices of all included sessions, while
% idx_sepsessions will contain all indices of session s in its cell s
idx_allsessions = 1:length(Session_Label);
idx_sepsessions = cell(1,length(sesvals));
% n_recperses is the number of recordings that should be sampled, at
% each fold, for each session index; it can be a non-integer value at
% this point (e.g., 1.5 if we want to have 6 reference samples out of 4
% different session subtypes)
n_recperses = n_refsubj/length(sesvals);
% We iterate over folds...
for boot = 1:n_boot
if boot == 1
tic;
end
idx_inref = [];
tmp_rand = 0;
for s = 1:length(sesvals)
% For each session subtype s, samples only the related indices
% and permutes them into a random order
idx_sepsessions{s} = idx_allsessions(Session_Label == sesvals(s));
idx_sepsessions{s} = idx_sepsessions{s}(randperm(length(idx_sepsessions{s})));
% For each session subtype s, picks how many samples should be
% incorporated in the reference group
if s == length(sesvals)
tmp_rand(s) = n_refsubj-sum(tmp_rand);
else
tmp_rand(s) = randi([floor(n_recperses),ceil(n_recperses)]);
end
% Adds the sample indices from session subtype s to idx_inref
idx_inref = [idx_inref,idx_sepsessions{s}(1:tmp_rand(s))];
end
clear tmp_rand
% Sorts the selected indices in ascending order
idx_inref = sort(idx_inref,'ascend');
% idx_notinref contains all the indices of the sessions on which
% ISFC will be computed at this fold (not in the reference)
idx_notinref = 1:length(TC);
idx_notinref(idx_inref) = [];
% Selects the sessions in the reference group, and the ones not in
% there
tmp_ref = TC(idx_inref);
tmp_notref = TC(idx_notinref);
% Computes ISFC for tmp_notref, using tmp_ref as a reference
% subgroup, for the current fold
ISFC_onefold = ISFC_ToReferenceGroup(tmp_notref,tmp_ref,W,S);
% We save the current fold output in tmp_cell, and sum it to the
% ISFC variable (which keeps track of the sum of ISFC through all
% folds)
tmp_cell = cell(1,length(Session_Label));
[tmp_cell{:}] = deal(zeros(n_conn,n_swtp));
tmp_cell(idx_notinref) = ISFC_onefold;
ISFC = cellfun(@(x,y) x+y,tmp_cell,ISFC,'UniformOutput',0);
clear tmp_cell
% At the same time, we also keep track of how many times each
% subject belonged to the non-reference side in Denom_factor, which
% is incremented as needed here
tmp_denom(idx_notinref) = {1};
Denom_factor = cellfun(@(x,y) x+y,tmp_denom,Denom_factor,'UniformOutput',0);
[tmp_denom{:}] = deal(0);
if boot == 1
time_onefold = toc;
disp(['Estimated total run time: ',num2str((n_boot-1)*time_onefold),' seconds...']);
end
end
% Averages the results; we divide by the number of times a given subject has
% been included
ISFC = cellfun(@(x,y) x/y,ISFC,Denom_factor,'UniformOutput',0);
end
c4science · Help