% Script to generate correlations between volumes from segmentations on % both photo reconstructions and the MRI used as ground truth. The script % relies upon saved .mat files from the script UWphoto_script_extractStats. clearvars %% load data PHOTO_RECON_HOME=getenv('PHOTO_RECON_HOME'); figuresDir = fullfile(PHOTO_RECON_HOME,'figures'); load(fullfile(figuresDir,'AdjustedCaseStats.mat'),'segVolumeInfo') strfields = fieldnames(segVolumeInfo); %% sort out cases to remove removalLevel = 2; removalFlag = (0<[segVolumeInfo.removaltype]) &... ([segVolumeInfo.removaltype] <=removalLevel); segVolumeInfo_kept=segVolumeInfo(~removalFlag); corltnsDir = fullfile(figuresDir,'correlations',... ['removalLevel_',num2str(removalLevel)]); if ~exist(corltnsDir,'dir') mkdir(corltnsDir) end softdir = fullfile(corltnsDir,'soft'); if ~exist(softdir,'dir') mkdir(softdir) end harddir = fullfile(corltnsDir,'hard'); if ~exist(harddir,'dir') mkdir(harddir) end crctdir = fullfile(corltnsDir,'hardCorrected'); if ~exist(crctdir,'dir') mkdir(crctdir) end %% setup variables mriFlag = ismember({segVolumeInfo_kept.segtype},'MRI'); N_mri = sum(mriFlag); hrdFlag = ismember({segVolumeInfo_kept.segtype},'Hard'); N_hrd = sum(hrdFlag); sftFlag = ismember({segVolumeInfo_kept.segtype},'Soft'); N_sft = sum(sftFlag); CorrelationStruct(length(strfields)-2) = struct(); %% iterate through labels for il=5:length(strfields) strctI = il-4; MRIsplot = [segVolumeInfo_kept(mriFlag).(strfields{il})]; hardplot = [segVolumeInfo_kept(hrdFlag).(strfields{il})]; hardCrct = [segVolumeInfo_kept(hrdFlag).(strfields{il})]... .*[segVolumeInfo_kept(hrdFlag).volumeAdjustmentFactor]; softplot = [segVolumeInfo_kept(sftFlag).(strfields{il})]; CorrelationStruct(strctI).Label = strfields{il}; CorrelationStruct(strctI).MRI.raw = MRIsplot; %% soft statistics CorrelationStruct(strctI).soft.raw = softplot; [rho,p,rhoLower,rhoUpper]=corrcoef(MRIsplot,softplot); CorrelationStruct(strctI).soft.corrcoef.rho = rho(1,2); CorrelationStruct(strctI).soft.corrcoef.p = p(1,2); CorrelationStruct(strctI).soft.corrcoef.rhoLower = rhoLower(1,2); CorrelationStruct(strctI).soft.corrcoef.rhoUpper = rhoUpper(1,2); [b,fitstats] = robustfit(MRIsplot,softplot); CorrelationStruct(strctI).soft.robustfit.b = b; CorrelationStruct(strctI).soft.robustfit.stats = fitstats; %% plot figures fig_soft=figure; plot(MRIsplot,softplot,'*') hold on plot(MRIsplot,b(1)+b(2)*MRIsplot,'r','LineWidth',2) pltTitle = strrep(['Soft: ', strfields{il}],'_',' '); title(pltTitle,'Interpreter','none') str_r=[' r= ',num2str(rho(1,2)),', p= ',num2str(p(1,2))]; T = text(min(get(gca, 'xlim')), max(get(gca, 'ylim')), str_r); set(T, 'fontsize', 12, 'verticalalignment', 'top', 'horizontalalignment', 'left'); xlabel('MRI label volume') ylabel('Photo label volume') %% hard statistics CorrelationStruct(strctI).hard.raw = hardplot; [rho,p,rhoLower,rhoUpper]=corrcoef(MRIsplot,hardplot); CorrelationStruct(strctI).hard.corrcoef.rho = rho(1,2); CorrelationStruct(strctI).hard.corrcoef.p = p(1,2); CorrelationStruct(strctI).hard.corrcoef.rhoLower = rhoLower(1,2); CorrelationStruct(strctI).hard.corrcoef.rhoUpper = rhoUpper(1,2); [b,fitstats] = robustfit(MRIsplot,hardplot); CorrelationStruct(strctI).hard.robustfit.b = b; CorrelationStruct(strctI).hard.robustfit.stats = fitstats; %% plot figures fig_hard=figure; plot(MRIsplot,hardplot,'*') hold on plot(MRIsplot,b(1)+b(2)*MRIsplot,'r','LineWidth',2) pltTitle = strrep(['Hard: ', strfields{il}],'_',' '); title(pltTitle,'Interpreter','none') str_r=[' r= ',num2str(rho(1,2)),', p= ',num2str(p(1,2))]; T = text(min(get(gca, 'xlim')), max(get(gca, 'ylim')), str_r); set(T, 'fontsize', 12, 'verticalalignment', 'top', 'horizontalalignment', 'left'); xlabel('MRI label volume') ylabel('Photo label volume') %% volume corrected hard statistics CorrelationStruct(strctI).hardCrctd.raw = hardCrct; [rho,p,rhoLower,rhoUpper]=corrcoef(MRIsplot,hardCrct); CorrelationStruct(strctI).hardCrctd.corrcoef.rho = rho(1,2); CorrelationStruct(strctI).hardCrctd.corrcoef.p = p(1,2); CorrelationStruct(strctI).hardCrctd.corrcoef.rhoLower = rhoLower(1,2); CorrelationStruct(strctI).hardCrctd.corrcoef.rhoUpper = rhoUpper(1,2); [b,fitstats] = robustfit(MRIsplot,hardCrct); CorrelationStruct(strctI).hardCrctd.robustfit.b = b; CorrelationStruct(strctI).hardCrctd.robustfit.stats = fitstats; %% plot figures fig_crct=figure; plot(MRIsplot,hardCrct,'*') hold on plot(MRIsplot,b(1)+b(2)*MRIsplot,'r','LineWidth',2) pltTitle = strrep(['Hard (VC): ', strfields{il}],'_',' '); title(pltTitle,'Interpreter','none') str_r=[' r= ',num2str(rho(1,2)),', p= ',num2str(p(1,2))]; T = text(min(get(gca, 'xlim')), max(get(gca, 'ylim')), str_r); set(T, 'fontsize', 12, 'verticalalignment', 'top', 'horizontalalignment', 'left'); xlabel('MRI label volume') ylabel('Photo label volume') %% save figures saveas(fig_soft,fullfile(softdir,strfields{il})) saveas(fig_soft,fullfile(softdir,[strfields{il},'.tiff'])) saveas(fig_hard,fullfile(harddir,strfields{il})) saveas(fig_hard,fullfile(harddir,[strfields{il},'.tiff'])) saveas(fig_crct,fullfile(crctdir,strfields{il})) saveas(fig_crct,fullfile(crctdir,[strfields{il},'.tiff'])) close all end %% handle ventricle outlier case LatVentIndices = [34,36,56]; for il=1:length(LatVentIndices) Icase = LatVentIndices(il); newloc = length(CorrelationStruct)+1; MRIsplot = CorrelationStruct(Icase).MRI.raw; hardCrct = CorrelationStruct(Icase).hardCrctd.raw; hardplot = CorrelationStruct(Icase).hard.raw; softplot = CorrelationStruct(Icase).soft.raw; %% find outlier [mVal,mInd]=max([MRIsplot;hardCrct;hardplot;softplot],[],2); IQR_foroutlier = iqr(MRIsplot); upperQrtl = prctile(MRIsplot,75); if mVal(1) < upperQrtl + 1.5*IQR_foroutlier warning('found value not an outlier') continue end if any(mInd~=mInd(1)) warning('Not located outlier'); continue end MRIsplot(mInd(1)) = []; hardCrct(mInd(1)) = []; hardplot(mInd(1)) = []; softplot(mInd(1)) = []; CorrelationStruct(newloc).Label = [CorrelationStruct(Icase).Label,'_pruned']; CorrelationStruct(newloc).MRI.raw = MRIsplot; CorrelationStruct(newloc).removedIndex = mInd(1); %% soft statistics CorrelationStruct(newloc).soft.raw = softplot; [rho,p,rhoLower,rhoUpper]=corrcoef(MRIsplot,softplot); CorrelationStruct(newloc).soft.corrcoef.rho = rho(1,2); CorrelationStruct(newloc).soft.corrcoef.p = p(1,2); CorrelationStruct(newloc).soft.corrcoef.rhoLower = rhoLower(1,2); CorrelationStruct(newloc).soft.corrcoef.rhoUpper = rhoUpper(1,2); [b,fitstats] = robustfit(MRIsplot,softplot); CorrelationStruct(newloc).soft.robustfit.b = b; CorrelationStruct(newloc).soft.robustfit.stats = fitstats; %% plot figures fig_soft=figure; plot(MRIsplot,softplot,'*') hold on plot(MRIsplot,b(1)+b(2)*MRIsplot,'r','LineWidth',2) pltTitle = strrep(['Soft: ', CorrelationStruct(newloc).Label],'_',' '); title(pltTitle,'Interpreter','none') str_r=[' r= ',num2str(rho(1,2)),', p= ',num2str(p(1,2))]; T = text(min(get(gca, 'xlim')), max(get(gca, 'ylim')), str_r); set(T, 'fontsize', 12, 'verticalalignment', 'top', 'horizontalalignment', 'left'); xlabel('MRI label volume') ylabel('Photo label volume') %% hard statistics CorrelationStruct(newloc).hard.raw = hardplot; [rho,p,rhoLower,rhoUpper]=corrcoef(MRIsplot,hardplot); CorrelationStruct(newloc).hard.corrcoef.rho = rho(1,2); CorrelationStruct(newloc).hard.corrcoef.p = p(1,2); CorrelationStruct(newloc).hard.corrcoef.rhoLower = rhoLower(1,2); CorrelationStruct(newloc).hard.corrcoef.rhoUpper = rhoUpper(1,2); [b,fitstats] = robustfit(MRIsplot,hardplot); CorrelationStruct(newloc).hard.robustfit.b = b; CorrelationStruct(newloc).hard.robustfit.stats = fitstats; %% plot figures fig_hard=figure; plot(MRIsplot,hardplot,'*') hold on plot(MRIsplot,b(1)+b(2)*MRIsplot,'r','LineWidth',2) pltTitle = strrep(['Hard: ', CorrelationStruct(newloc).Label],'_',' '); title(pltTitle,'Interpreter','none') str_r=[' r= ',num2str(rho(1,2)),', p= ',num2str(p(1,2))]; T = text(min(get(gca, 'xlim')), max(get(gca, 'ylim')), str_r); set(T, 'fontsize', 12, 'verticalalignment', 'top', 'horizontalalignment', 'left'); xlabel('MRI label volume') ylabel('Photo label volume') %% volume corrected hard statistics CorrelationStruct(newloc).hardCrctd.raw = hardCrct; [rho,p,rhoLower,rhoUpper]=corrcoef(MRIsplot,hardCrct); CorrelationStruct(newloc).hardCrctd.corrcoef.rho = rho(1,2); CorrelationStruct(newloc).hardCrctd.corrcoef.p = p(1,2); CorrelationStruct(newloc).hardCrctd.corrcoef.rhoLower = rhoLower(1,2); CorrelationStruct(newloc).hardCrctd.corrcoef.rhoUpper = rhoUpper(1,2); [b,fitstats] = robustfit(MRIsplot,hardCrct); CorrelationStruct(newloc).hardCrctd.robustfit.b = b; CorrelationStruct(newloc).hardCrctd.robustfit.stats = fitstats; %% plot figures fig_crct=figure; plot(MRIsplot,hardCrct,'*') hold on plot(MRIsplot,b(1)+b(2)*MRIsplot,'r','LineWidth',2) pltTitle = strrep(['Hard (VC): ', CorrelationStruct(newloc).Label],'_',' '); title(pltTitle,'Interpreter','none') str_r=[' r= ',num2str(rho(1,2)),', p= ',num2str(p(1,2))]; T = text(min(get(gca, 'xlim')), max(get(gca, 'ylim')), str_r); set(T, 'fontsize', 12, 'verticalalignment', 'top', 'horizontalalignment', 'left'); xlabel('MRI label volume') ylabel('Photo label volume') %% save figures saveas(fig_soft,fullfile(softdir,CorrelationStruct(newloc).Label)) saveas(fig_soft,fullfile(softdir,[CorrelationStruct(newloc).Label,'.tiff'])) saveas(fig_hard,fullfile(harddir,CorrelationStruct(newloc).Label)) saveas(fig_hard,fullfile(harddir,[CorrelationStruct(newloc).Label,'.tiff'])) saveas(fig_crct,fullfile(crctdir,CorrelationStruct(newloc).Label)) saveas(fig_crct,fullfile(crctdir,[CorrelationStruct(newloc).Label,'.tiff'])) close all end %% save correlations statistics save(fullfile(corltnsDir,'CorrelationStructure'),'CorrelationStruct')