%% EEG GO during Response task: TASK Post-processing: for Time-Frequency ananlysis
%%% DATA was Segmented from the GO
% Get the directories for pre-processed EEG file folder
clear; clc; close all;

cwd = pwd; 
addpath(cwd);

datalocation = [cwd '\ALL_DATA\ALL_Processed_Data\'];   % Data are here
savelocation = [cwd '\ALL_DATA\ALL_Processed_Data\'];   % Data save here
cd(datalocation);

[~, ~, SubjInfo] = xlsread([ cwd '\Patients_MS.xlsx'],'AllPedaling');


%%
for nn = 1: 2%size(SubjInfo,1)
        
    type = SubjInfo{nn,1};
    subj = SubjInfo{nn, 2};
    
    disp([type,num2str(subj),'_Ped_Processed.mat']);    
    load([type,num2str(subj),'_Ped_Processed.mat']);
    
    %Get Subj Info and remove the bad ICAs on the basis of VEOG correlation
    %Bad ICAS 1,2,3 are ADJUST, VEOG correlation, and Gaussian template.
    %ADJUST is a stand-alone program that isn't very good.
    %VEOG correlation is the most reliable and the gaussian template can be useful.  
    %APPLE VEOG corr is RESP about 85% of teh time.

   % % Get the data here and remove bad ICAs: VEOG and TEMPLATE
    temp1a = cell2mat(bad_ICAs(2));   % 2 = bad_VEOG_ICAs 
    temp1b = cell2mat(bad_ICAs(3));   % 3 = Bad TEMPLATE ICAs 
        temp1 = [temp1a temp1b];   
    tpidx = find (temp1==0);
    temp1(tpidx)=[];
    
    if isempty(temp1)
        temp1 = 'NaN';
    end

    if isnumeric(temp1)
      bad_ICAs_To_Remove=temp1;
    elseif strmatch('NaN',temp1)
      bad_ICAs_To_Remove=NaN;
    else
      bad_ICAs_To_Remove=str2num(temp1);
    end
    clear temp1a temp1b temp1 tpidx;

    %%% Remove the (presumptive) bad ICAs:
    if ~(isnan(bad_ICAs_To_Remove))
       EEG = pop_subcomp( EEG, bad_ICAs_To_Remove, 0);
    end
    clear bad_ICAs_To_Remove bad_chans bad_epochs bad_ICAs
                
     
    %% ###################################   TIME FREQUENCY ANALYSIS %% ########################################################
    
    % Set Times
    tx=EEG.times;
    b1=find(tx==-300);  b2=find(tx==-100);  %% Baseline
    t1=find(tx==-500);  t2=find(tx==2000);
    tx2disp=-500:2:2000;
    
    %%%%%%%%%%%%%%%%%%% Setup Wavelet Params %%%%%%%%%%%%%%%%%%%%%%%%%%%%
    min_freq = 1.0233;
    max_freq = 50;
    num_freqs=50;
  
    %%%%% logspace
    frex = logspace(log10(min_freq),log10(max_freq),num_freqs); 
    s=logspace(log10(3),log10(10),num_freqs)./(2*pi*frex);
    
    %%%% timelength of the wavelet
    t=-2:1/EEG.srate:2;
    
    % Definte Convolution Parameters
    n_wavelet = length(t); half_of_wavelet_size = (n_wavelet-1)/2; clear dims
    % -------- GO
    dims{1} = size(EEG.data);    n_data{1} = dims{1}(2)*dims{1}(3);  n_convolution{1} = n_wavelet+n_data{1}-1;  n_conv_pow2{1} = pow2(nextpow2(n_convolution{1}));
    
    %%% decide channels
    channels = EEG.nbchan;
    for  chani = 1:channels
        % get FFT of data
        EEG_fft{1} = fft(reshape(EEG.data   (chani,:,:),1,n_data{1}),n_conv_pow2{1});
       
        for fi=1:num_freqs
            wavelet{1} = fft( exp(2*1i*pi*frex(fi).*t) .* exp(-t.^2./(2*(s(fi)^2))) , n_conv_pow2{1} );
           
                % convolution for 1 data set
                for convo=1
                    EEG_conv = ifft(wavelet{convo}.*EEG_fft{convo});
                    EEG_conv = EEG_conv(1:n_convolution{convo});
                    EEG_conv = EEG_conv(half_of_wavelet_size+1:end-half_of_wavelet_size);
                    EEG_multi_conv{convo} = reshape(EEG_conv,dims{convo}(2),dims{convo}(3)); clear EEG_conv;
                    temp_POWER{convo} = abs(EEG_multi_conv{convo}(t1:t2,:)).^2;
                end

                 % Baseline {1}
                 BASE = mean(mean(abs( EEG_multi_conv{1}(b1:b2,:)).^2));
                 
                 temp_POWER_avg(:,1) = mean(temp_POWER{1}(:,:),2); % GO
                
                % dB correct power by base
                % WARNING CUE; GO/GO Cue; RESPONSE
                  POWER(chani,fi,:,1) = 10*( log10(temp_POWER_avg(:,1)) - log10(repmat(BASE,size(temp_POWER_avg(:,1),1),1)) ); % GO
                 clear temp* EEG_multi_conv wavelet BASE1 BASE2;
        end
        clear *_fft;
    end
    
    % plot figure
%     ChanCz = find(strcmpi('Cz',{EEG.chanlocs.labels}));
%     yfreq = 50;
%     
%     imagesc(tx2disp,[], squeeze( POWER(ChanCz,:,:,1))); axis xy; hold on; plot([0 0],[1 yfreq],'k:');
%     set(gca,'xlim',[tx2disp(1),tx2disp(end)],'ylim',[1 50],'clim',[-4 4],'YTick',1:4:length(frex),'YTickLabel',round(frex(1:4:end)));  title('Cz: POWER');
%     suptitle([type,num2str(subj)]);
%     saveas(gcf, [savelocation, type,num2str(subj),'_ANALYZED.png']);
        
    %% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% SAVE DATA %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
    chanlocs = EEG.chanlocs;
    MainOuts.POWER = POWER; 
    
    save([savelocation, type,num2str(subj),'_ANALYZED.mat'],'MainOuts','chanlocs' );
     
   %%  Clear everything except some variables to run loop
   close all;
   clearvars -except datalocation savelocation SubjInfo
   
end