Analysis_Ceff_lang.m

clear all;
addpath('/Users/cbc/Documents/sanz/utils')
addpath(genpath('/Users/cbc/Documents/matlab_stuffs/'));
 Isubdiag = find(tril(ones(100),-1));

NSUB=49;
for sub=1:NSUB
    load(sprintf('model_vortex_partition_lang_%03d.mat',sub));
    Ceff_task1_sub(sub,:,:)=Ceff_task1;
    Ceff_task2_sub(sub,:,:)=Ceff_task2;    
    FCv_task1_sub(sub,:,:)=FCvortex_emp_task1;
    FCv_task2_sub(sub,:,:)=FCvortex_emp_task2;
end

NP=100;

Ceff_task1 = squeeze(mean(Ceff_task1_sub));
Ceff_task2 = squeeze(mean(Ceff_task2_sub));


Ceff_task1_gbc=squeeze(mean(Ceff_task1,2));

Ceff_task2_gbc=squeeze(mean(Ceff_task2,2));



gbc_ceff_task2_sub=squeeze(mean(Ceff_task2_sub,2));
gbc_ceff_task1_sub=squeeze(mean(Ceff_task1_sub,2));



nn=1;
for i=1:NP
    a=gbc_ceff_task1_sub(:,i)';
    b=gbc_ceff_task2_sub(:,i)';
    pp(nn)=ranksum(a,b);
    nn=nn+1;
end
nsigMask=FDR_benjHoch(pp,0.05);




%%%  Classification

nsig=1:10000;

% Ceff
gbc_ceff_task1_sub=Ceff_task1_sub(:,:);
gbc_ceff_task2_sub=Ceff_task2_sub(:,:);


% vortex FC
% gbc_ceff_task1_sub=FCv_task1_sub(:,:);
% gbc_ceff_task2_sub=FCv_task2_sub(:,:);





NTRAIN=40;
kfold=1000;
cl=1:2;
pc=zeros(2,2);
for nfold=1:kfold
    shuffling=randperm(NSUB);
    TrainData1=[];
    for sub=shuffling(1:NTRAIN)
        TS=gbc_ceff_task1_sub(sub,nsig);
        TrainData1=vertcat(TrainData1,TS);
    end
    XValidation1=[];
    for sub=shuffling(NTRAIN+1:end)
        TS=gbc_ceff_task1_sub(sub,nsig);
        XValidation1=vertcat(XValidation1,TS);
    end
    Responses1=categorical(ones(size(TrainData1,1),1),cl);
    YValidation1=categorical(ones(size(XValidation1,1),1),cl);

    TrainData2=[];
    for sub=shuffling(1:NTRAIN)
        TS=gbc_ceff_task2_sub(sub,nsig);
        TrainData2=vertcat(TrainData2,TS);
    end
    XValidation2=[];
    for sub=shuffling(NTRAIN+1:end)
        TS=gbc_ceff_task2_sub(sub,nsig);
        XValidation2=vertcat(XValidation2,TS);
    end
    Responses2=categorical(2*ones(size(TrainData2,1),1),cl);
    YValidation2=categorical(2*ones(size(XValidation2,1),1),cl);

    TrainData=vertcat(TrainData1,TrainData2);
    XValidation=vertcat(XValidation1,XValidation2);
    Responses=vertcat(Responses1,Responses2);
    YValidation=vertcat(YValidation1,YValidation2);

    t = templateSVM('Standardize',true,'KernelFunction','linear','KernelScale','auto');
%    t = templateSVM('KernelFunction','linear','KernelScale','auto');
    svmmodel=fitcecoc(TrainData,Responses,'Learners',t);

    % compute
    valno1=size(XValidation1,1);
    con=zeros(2,2);
    test1=predict(svmmodel,XValidation1);
    for i=1:valno1
        winclass=test1(i);
        con(1,winclass)=con(1,winclass)+1;
    end
    valno2=size(XValidation2,1);
    test2=predict(svmmodel,XValidation2);
    for i=1:valno2
        winclass=test2(i);
        con(2,winclass)=con(2,winclass)+1;
    end
    con(1,:)=con(1,:)/valno1;
    con(2,:)=con(2,:)/valno2;
    accdist(nfold)=sum(diag(con))/2;
    pc=pc+con;
end
pc=pc/kfold
acc=sum(diag(pc))/2

%save results_Ceff_classifier_lan1lan2.mat accdist pc Ceff_task1 Ceff_task2 

load results_Ceff_classifier_lan1lan2.mat
for i=1:50
    acc_avg(i)=mean(accdist((i-1)*20+1:(i*20)));
end

figure;

xSize = 5; ySize = 10;
xLeft = (21-xSize)/2; yTop = (30-ySize)/2;
set(gcf,'PaperUnits','centimeters')
set(gcf,'PaperPosition',[xLeft yTop xSize ySize])
set(gcf,'Position',[50 50 xSize*50 ySize*50],'Color','w')
% xSize = 5; ySize = 15;
%swarm({acc_avg},{'Gec'},'tlt','Accuracy');ylim([0.45 1]);xlim([0.90 1.1])
violinplot(acc_avg)
ylim([0.45 1]);
hline(0.5,'r--')


data2mse =  [acc_avg];

group_names = {'corr Rest'};
condition_names= {'Temporal'};
c =  [0.45, 0.80, 0.69];
group_inx = [ones(1,50)];

figure
xSize = 5; ySize = 10;
xLeft = (21-xSize)/2; yTop = (30-ySize)/2;
set(gcf,'PaperUnits','centimeters')
set(gcf,'PaperPosition',[xLeft yTop xSize ySize])
set(gcf,'Position',[50 50 xSize*50 ySize*50],'Color','w')
% different color scheme, different position of boxplots and scatter
h = daviolinplot(acc_avg,'groups',group_inx,'color',c,'outsymbol','k+',...
    'xtlabels', condition_names(1:1),'scatter',2,'scattersize',10,'jitter',1,...
    'box',1,'boxcolors','same','scattercolors','same',...
    'boxspacing',0.2,'boxwidth',2)%,'legend',group_names(1:4));
ylabel('MSE');
ylim([0.45 1]);
hline(0.5,'r--')
xl = xlim; xlim([xl(1)-0.1, xl(2)]); % make more space for the legend
set(gca,'FontSize',10); grid off



figure
confusionchart(round(pc*100),'DiagonalColor','k','OffDiagonalColor','w')



% brain renders
% trampa para plotear en 1000 los datos en sch 100
% 
% load('schaefer1000to100.mat')
% tmp=zeros(1000,1);
% tmp2=zeros(1000,1);
% Mask=zeros(1000,1);
% 
% Mask2=zeros(100,1);
% Mask2(nsigMask)=1;
% 
% for i=1:100
%     indx=find(i==schaefer1000to100);
%     tmp(indx)= Ceff_task1_gbc(i);
%     tmp2(indx)= Ceff_task2_gbc(i);
%     Mask(indx)= Mask2(i);
% 
% end
% 
% 
% tmp=tmp.*Mask;
% tmp2=tmp2.*Mask;
% 
% 
% rendersurface_schaefer1000(tmp,0,0.025, 0,'BuDRd_12',2)
% rendersurface_schaefer1000(tmp2,0,0.025, 0,'BuDRd_12',2)


Mask2=zeros(100,1);
Mask2(nsigMask)=1;
tmp=Ceff_task1_gbc;%.*Mask2;
tmp2=Ceff_task2_gbc;%.*Mask2;



rendersurface_schaefer100(tmp,0,0.025, 0,'BuDRd_12',2)
rendersurface_schaefer100(tmp2,0,0.025, 0,'BuDRd_12',2)

rendersurface_schaefer100(tmp2-tmp,-max(tmp2-tmp),max(tmp2-tmp), 1,'RdYlBu10',2)