main.m

%% load and prepare all single tiffs
%importing
[data_g, data_r] = CellSortLoadSingleFiles('ChanA_*_*.tif', 'ChanB_*_*.tif');
%% matVis
matVis(data_r);
%% get ref frame for shift correction
interval = 1:20;
refFrame = cast(mean(data_r(:,:,interval), 3), class(data_r));
matVis(refFrame);
%% shiftCorrection
%generates shift corrected files for the green and red channels, needs
%DIPlib (available for free)
[data_g, data_r] = CellSortShiftCorrection(data_g, data_r, refFrame);
%temporal binning
binningFactor = [1 1 2]; %x, y, t
[data_g, data_r] = CellSortTemporalBinning(data_g, data_r, binningFactor);

fn = 'processedData.mat';
save(fn, 'data_g', 'data_r', '-v7.3');
%% load files
%if you haven't used the shiftCorrection you have to load the already processed files
fn= 'processedData.mat';
load(fn);
%% 1. PCA
%calculate PCAs
fn_green = 'processedData.mat';
nPCs = 100;
flims = [];
dSamp = [1 1]; %temporal, spatial downsampling
[mixedsig, mixedfilters, CovEvals, covtrace, movm, movtm] = ...
    CellsortPCA(fn_green, flims, nPCs, dSamp, [], [], data_g);

%% 2a. Choose PCs
%choose PCAs, type f/b in command line to go back or forward. type two number
%to select the range "1 ENTER 25 ENTER" would select the PCAs from 1 to 25
[PCuse] = CellsortChoosePCs(fn_green, mixedfilters, data_g);

%% 2b. Plot PC spectrum
%optionally, you can display the PC spectrum..
CellsortPlotPCspectrum(fn_green, CovEvals, PCuse, data_g)

%% 3a. ICA
%calculate ICAs, set mu to a value between 0 and 1, where 1 means pure
%temporal identification and 0 pure spatial. A value of 0.1-0.2 seems to be
%best.. make sure that the algorithm converges (see command line). If not
%run it again
nIC = length(PCuse);
mu = 0.2;

[ica_sig, ica_filters, ica_A, numiter] = CellsortICA(mixedsig, mixedfilters, CovEvals, PCuse, mu, nIC);

%% 3b. Plot ICs
%optionally, plot ICs (you have to use at least 20.. or change the number
%on the buttom
tlims = [];
dt = 0.1;

figure(2)
CellsortICAplot('series', ica_filters, ica_sig, movm, tlims, dt, [], [], [1:size(ica_sig,1)]);

%% 4a. Segment contiguous regions within ICs
%search for indepent segments, you have to adjust these values well:
%smwidth: width of a smoothing filter (1 should be okay)
%thresh: threshold for normalized standard deviation in ICs (verey
%        sensitive, standard: 2)
%arealims: limit for the detected areas in pixeks: [min max] or [min]
%plotting: 1 = yes or 0 = no.. shall a plot be shown or not?

smwidth = 1;
thresh = 2;
arealims = [250];
plotting = 1;

[ica_segments, segmentlabel, segcentroid] = CellsortSegmentation(ica_filters, smwidth, thresh, arealims, plotting);

%% 4b. CellsortApplyFilter
%filter signal and extract regions signals
subtractmean = 1;

cell_sig = CellsortApplyFilter(fn_green, ica_segments, flims, movm, subtractmean, data_g);

%% 5. CellsortFindspikes
%optionally, find spikes in signal
deconvtau = 0;
spike_thresh = 2;
normalization = 1;

[spmat, spt, spc] = CellsortFindspikes(ica_sig, spike_thresh, dt, deconvtau, normalization);

%% Show results
%optionally, plot signals
figure(2)
CellsortICAplot('series', ica_filters, ica_sig, movm, tlims, dt, 1, 2, [1:size(ica_segments,1)], spt, spc);
%% show signal of cells (cell_sig)
%caluclate correlation matrix for cell_sig
[ica_segments, cluster, subPlotCols, subPlotWidth, sortedIdx] = calcCorrEff(cell_sig, ica_segments);
alternativePlot(ica_segments, cluster, cell_sig, subPlotCols, sortedIdx, data_r);

%% let user deselect and merge rois
%merge -> change cluster ->recereate plots
merge = {[1,2,3], [5,6]};
delete = [8,4,11];
figure;
cluster_new = cluster;
%delete
for i = 1:size(delete,2)
    cluster_new{delete(i)} = [];
end
%merge
for i = 1:size(merge, 2)
    for j = 2:size(merge{i},2)
        cluster_new{merge{i}(1)} = [cluster_new{merge{i}(1)}, cluster_new{merge{i}(j)}];
        cluster_new{merge{i}(j)} = [];
    end
end
cluster_new = cluster_new(~cellfun(@isempty, cluster_new));
subPlotCols = size(cluster_new, 2);

%redraw
alternativePlot(ica_segments, cluster_new, cell_sig, subPlotCols, sortedIdx, data_r);