Homec4science
Diffusion iCAPs public version 2 (master) Browse

iCAPs_v2/functions/04_Regression/evaluateSoftClusterThres_corrs.m
e5f5005bb2efmaster

Last Change
Raw File

evaluateSoftClusterThres_corrs.m
View Options

function evaluateSoftClusterThres_corrs(clusteringResults,TC,param,fid)
thresVals=param.softClusterThres;
AI=clusteringResults.AI';
AI_subject_labels=clusteringResults.AI_subject_labels;
subject_labels=clusteringResults.subject_labels;
time_labels=clusteringResults.time_labels;
dist_to_centroid=clusteringResults.dist_to_centroid;
iCAPs_z=ZScore_iCAPs(clusteringResults.iCAPs);
nThres=length(thresVals);
nSub=param.n_subjects;
nVox=size(AI,1);
nTP_all=size(AI,2);
nI=size(subject_labels,1);
%% correlation between measured and estimated TC amplitudes
for iT=1:nThres
[IDX_mat,dist_thres,clos_to_centroid]=getIDXmat(dist_to_centroid,thresVals(iT));
dist_mat_all(:,:,iT)=dist_thres;
IDX_mat_all(:,:,iT)=IDX_mat;
nClus_mat_all(:,iT)=sum(IDX_mat,2);
clus_weights(:,:,iT)=clos_to_centroid;
end
disp('Computing normalized innovations...')
Activity_inducing_norm=zeros(size(AI));
innovations_norm=zeros(size(AI));%zeros(size(AI,1),size(AI,2)-nSub);
for iS=1:nSub
% indices of time series from subject iS
vols_iS=AI_subject_labels==iS;
nTP_sub(iS)=nnz(AI_subject_labels==iS);
% divide by standard deviation
Activity_inducing_norm(:,vols_iS)=AI(:,vols_iS)./repmat(sqrt(mean(AI(:,vols_iS).^2,2)),[1,nTP_sub(iS)]);
% remove mean across voxels (global signal regression)
% Activity_inducing_norm(:,vols_iS)=Activity_inducing_norm(:,vols_iS) - repmat(mean(Activity_inducing_norm(:,vols_iS),1),n_vox,1);
Activity_inducing_norm(isnan(Activity_inducing_norm))=0;
innovations_norm(:,vols_iS)=[zeros(size(AI,1),1),diff(Activity_inducing_norm(:,vols_iS),1,2)];
end
% compute the innovation amplitudes in the clustering maps, threshold 1.5, 0.5 and 2.3
disp('Computing innovation amplitudes in the cluster maps...')
thres=1.5;
disp(['Spatial threshold (zscore): ' num2str(thres)]);
for iT=1:nThres
disp(['Soft cluster assignment factor: ' num2str(thresVals(iT))]);
for iI=1:nI
if ~mod(iI,1000)
fprintf([num2str(iI) ' ']);
end
iS=subject_labels(iI);
t=time_labels(iI);if t>nTP_sub(iS);t=t-nTP_sub(iS);end
clusterID=find(IDX_mat_all(iI,:,iT));
clusterWeights=clus_weights(iI,:,iT);
vols_iS=AI_subject_labels==iS;
% compute "innovation amplitude", i.e. the average innovation inside
% the mask of the iCAP this frame was assigned to
innovSub=innovations_norm(:,vols_iS);
if t==1;t=t+1;end
TC_change_tmp=TC{iT}{iS}(:,t)-TC{iT}{iS}(:,t-1);
% loop through all clusters
AI_change_measured(iI,iT)=0;
TC_change_estimated(iI,iT)=0;
for iC=clusterID
AI_change_measured(iI,iT)=AI_change_measured(iI,iT)+...
mean(innovSub(innovSub(:,t)~=0&iCAPs_z(iC,:)'>thres,t))*clusterWeights(iC);
TC_change_estimated(iI,iT)=TC_change_estimated(iI,iT)+TC_change_tmp(iC)*clusterWeights(iC);
end
end
end
AI_change_measured(isnan(AI_change_measured))=0;
TC_change_estimated(isnan(TC_change_estimated))=0;
corr_amps=diag(corr(AI_change_measured,TC_change_estimated));
[~,opt_id]=findpeaks(corr_amps);
opt_id=opt_id(1);
WriteInformation(fid,['The optimum correlation is at xi=' num2str(thresVals(opt_id))]);
figure('position',[440 565 270 233]);
plot(thresVals,corr_amps,'*');
hold on
plot(thresVals(opt_id),corr_amps(opt_id),'*r')
title('correlation measured vs. estimated transient amplitudes')
print(fullfile(param.outDir_reg,'evalCorr'),'-depsc2','-painters');
c4science · Help