%% Confusion matrices comparing voxel-based vs [aal3,schaefer,mental] - BRAIN & BEHAV SCORES

% 3 confusion matrices showing correlations between NORMALISED AVERAGED Br & Beh
% scores on modes of (all Pearson) voxel-wise & network, voxel-wise & AAL, 
% and voxel-wise & voxel-wise-mental analyses.



clear;clc

% After running all 4 models, you have to change the paths below accordingly.
basedir_main = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA\framework\spls_fair20-5_fair_modeA_pub_noAnom_new_common_bmi\res\';
basedir_aal3 = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA_aal3\framework\spls_fair20-5_fair_modeA_pub_aal3_new_noAnom_common_bmi\res\';
basedir_schaefer = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA_schaeferbuckchoi\framework\spls_fair20-5_fair_modeA_pub_schaeferbuckchoi_noAnom_common_bmi\res\';
basedir_mental = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA\framework\spls_fair20-5_fair_modeA_pub_noAnom_new_common_bmi_mental\res\';



cd(basedir_main)

P_mean_main = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_main.(['mode' num2str(i)]) = mean([P(:,1) P(:,2)],2);
cd(basedir_main)
end

cd(basedir_aal3)

P_mean_aal3 = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_aal3.(['mode' num2str(i)]) = mean([P(:,1) P(:,2)],2);
cd(basedir_aal3)
end


cd(basedir_schaefer)

P_mean_schaefer = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_schaefer.(['mode' num2str(i)]) = mean([P(:,1) P(:,2)],2);
cd(basedir_schaefer)
end


cd(basedir_mental)

P_mean_mental = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_mental.(['mode' num2str(i)]) = mean([P(:,1) P(:,2)],2);
cd(basedir_mental)
end


cd(basedir_main)


for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_aal3,2)

main_aal3_correls(i,j) = corr(P_mean_main{:,i},P_mean_aal3{:,j});
end
end
    
    
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_schaefer,2)

main_schaefer_correls(i,j) = corr(P_mean_main{:,i},P_mean_schaefer{:,j});
end
end


for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_mental,2)

main_mental_correls(i,j) = corr(P_mean_main{:,i},P_mean_mental{:,j});
end
end


figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6'};
h = heatmap(xvalues,yvalues,main_aal3_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Brain & Psychosocial score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = 'Anatomical parcellation analysis';
h.CellLabelFormat = '%.2f';

set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_aal3_brain_behav.png']);



figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6'};
h = heatmap(xvalues,yvalues,main_schaefer_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Brain & Psychosocial score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = 'Network-based parcellation analysis';
h.CellLabelFormat = '%.2f';

set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_schaefer_brain_behav.png']);


% Dirty hack color for proper color-scaling, add a 7th dummy mode to adjust
% color-scaling, and then discard it with manual image processing.
P_mean_main.mode7 = P_mean_mental.mode2;
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_mental,2)
main_mental_correls(i,j) = corr(P_mean_main{:,i},P_mean_mental{:,j});
end
end

figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6', 'Mode 7'};
h = heatmap(xvalues,yvalues,main_mental_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Brain & Psychosocial score correlations';
h.YLabel = '"Main" Voxel-wise analysis';
h.XLabel = '"Mental" Voxel-wise analysis';
h.CellLabelFormat = '%.2f';

set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_mental_brain_behav.png']);



%% Confusion matrices comparing voxel-based vs [aal3,schaefer,mental] - BRAIN SCORES


clear;clc

% After running all 4 models, you have to change the paths below accordingly.
basedir_main = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA\framework\spls_fair20-5_fair_modeA_pub_noAnom_new_common_bmi\res\';
basedir_aal3 = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA_aal3\framework\spls_fair20-5_fair_modeA_pub_aal3_new_noAnom_common_bmi\res\';
basedir_schaefer = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA_schaeferbuckchoi\framework\spls_fair20-5_fair_modeA_pub_schaeferbuckchoi_noAnom_common_bmi\res\';
basedir_mental = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA\framework\spls_fair20-5_fair_modeA_pub_noAnom_new_common_bmi_mental\res\';


cd(basedir_main)

P_mean_main = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_main.(['mode' num2str(i)]) = P(:,1);
cd(basedir_main)
end


cd(basedir_aal3)

P_mean_aal3 = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_aal3.(['mode' num2str(i)]) = P(:,1);
cd(basedir_aal3)
end


cd(basedir_schaefer)

P_mean_schaefer = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_schaefer.(['mode' num2str(i)]) = P(:,1);
cd(basedir_schaefer)
end


cd(basedir_mental)

P_mean_mental = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_mental.(['mode' num2str(i)]) = P(:,1);
cd(basedir_mental)
end


cd(basedir_main)

for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_aal3,2)

main_aal3_correls(i,j) = corr(P_mean_main{:,i},P_mean_aal3{:,j});
end
end
    
    
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_schaefer,2)

main_schaefer_correls(i,j) = corr(P_mean_main{:,i},P_mean_schaefer{:,j});
end
end


for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_mental,2)

main_mental_correls(i,j) = corr(P_mean_main{:,i},P_mean_mental{:,j});
end
end


figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6'};
h = heatmap(xvalues,yvalues,main_aal3_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Brain score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = 'Anatomical parcellation analysis';
h.CellLabelFormat = '%.2f';
set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_aal3_brain.png']);


figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6'};
h = heatmap(xvalues,yvalues,main_schaefer_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Brain score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = 'Network-based parcellation analysis';
h.CellLabelFormat = '%.2f';
set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_schaefer_brain.png']);

% Dirty hack color for proper color-scaling, add a 7th dummy mode to adjust
% color-scaling, and then discard it with manual image processing.
P_mean_main.mode7 = -P_mean_main.mode5;
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_mental,2)
    
main_mental_correls(i,j) = corr(P_mean_main{:,i},P_mean_mental{:,j});
end
end

figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6','Mode 7'};
h = heatmap(xvalues,yvalues,main_mental_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Brain score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = '"Mental" Voxel-wise analysis';
h.CellLabelFormat = '%.2f';
set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_mental_brain.png']);


%% Confusion matrices comparing voxel-based vs [aal3,schaefer] - BEHAV SCORES

clear;clc

% After running all 4 models, you have to change the paths below accordingly.
basedir_main = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA\framework\spls_fair20-5_fair_modeA_pub_noAnom_new_common_bmi\res\';
basedir_aal3 = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA_aal3\framework\spls_fair20-5_fair_modeA_pub_aal3_new_noAnom_common_bmi\res\';
basedir_schaefer = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA_schaeferbuckchoi\framework\spls_fair20-5_fair_modeA_pub_schaeferbuckchoi_noAnom_common_bmi\res\';
basedir_mental = 'C:\Users\Konstantinos Tsirlis\Dropbox\My Desktop\Mourao - Miranda\ABCD results\abcdAll_noA\framework\spls_fair20-5_fair_modeA_pub_noAnom_new_common_bmi_mental\res\';


cd(basedir_main)

P_mean_main = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_main.(['mode' num2str(i)]) = P(:,2);
cd(basedir_main)
end


cd(basedir_aal3)

P_mean_aal3 = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_aal3.(['mode' num2str(i)]) = P(:,2);
cd(basedir_aal3)
end


cd(basedir_schaefer)

P_mean_schaefer = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_schaefer.(['mode' num2str(i)]) = P(:,2);
cd(basedir_schaefer)
end


cd(basedir_mental)

P_mean_mental = table();
for i = 1:(length(dir('level*')))
cd(['level' num2str(i)])
load('P.mat')
P = reshape(P,[11288,2]);
P = (P-mean(P))./std(P);
P_mean_mental.(['mode' num2str(i)]) = P(:,2);
cd(basedir_mental)
end

cd(basedir_main)

for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_aal3,2)

main_aal3_correls(i,j) = corr(P_mean_main{:,i},P_mean_aal3{:,j});
end
end
    
    
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_schaefer,2)

main_schaefer_correls(i,j) = corr(P_mean_main{:,i},P_mean_schaefer{:,j});
end
end
    
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_mental,2)

main_mental_correls(i,j) = corr(P_mean_main{:,i},P_mean_mental{:,j});
end
end

figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6'};
h = heatmap(xvalues,yvalues,main_aal3_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Psychosocial score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = 'Anatomical parcellation analysis';
h.CellLabelFormat = '%.2f';
set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_aal3_behav.png']);


figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6'};
h = heatmap(xvalues,yvalues,main_schaefer_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Psychosocial score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = 'Network-based parcellation analysis';
h.CellLabelFormat = '%.2f';
set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_schaefer_behav.png']);

% Dirty hack color for proper color-scaling, add a 7th dummy mode to adjust
% color-scaling, and then discard it with manual image processing.
P_mean_main.mode7 = -P_mean_main.mode2;
for i = 1:size(P_mean_main,2)
for j = 1:size(P_mean_mental,2)
    
main_mental_correls(i,j) = corr(P_mean_main{:,i},P_mean_mental{:,j});
end
end

figure('Units', 'Normalized', 'OuterPosition', [0.4 0 0.5 0.78]);
pos = [0.22 0.2 0.6 0.56];
axes('Position', pos,'linewidth',2)

% You need to change the legends of the modes according to the new ones, in case they are different.
xvalues = {'Mode 1','Mode 2','Mode 3'};
yvalues = {'Mode 1','Mode 2','Mode 3', 'Mode 4','Mode 5','Mode 6','Mode 7'};
h = heatmap(xvalues,yvalues,main_mental_correls,'ColorMap',redbluecmap,'gridvisible',0);

h.Title = 'Psychosocial score correlations';
h.YLabel = 'Voxel-wise analysis';
h.XLabel = '"Mental" Voxel-wise analysis';
h.CellLabelFormat = '%.2f';
set(gca, 'FontSize', 30, 'FontName', 'Arial');

saveas(gcf, ['heatmap_main_vs_mental_behav.png']);