Connectivity image size is not matched with no of EEG channels

ddaschakladar · November 11, 2021, 6:02am

Hi sir,
I have created the connectivity matrix (correlation, coherence) based on 20 channels of EEG. So it is obvious that the resultant connectivity matrix size should be 20x20 (one pixel/connection), however after computing the connectivity matrix the image size is 873x710. How this can happen (20x20->873x710)? How is the image size big for 20 channels connectivity?
I have attached the image and .excel file of correlation connectivity.
Please help me out.THMS_MA_CorrMat_20Channels.csv (6.4 KB)

Francois · November 11, 2021, 9:46am

The screen capture of the figure that you posted shows a 20x20 connectivity matrix. Click anywhere on the figure to see the name of the row/column in the legent.
The CSV you posted is also 20x20.

Are you getting mixed up with with number of pixels in your screen capture?

ddaschakladar · November 11, 2021, 10:56am

Thanks for the update.
Actually I know its a 20x 20 connectivity matrix and each cell (row/column) represents the connectivity value between channels. However the figure size is 873x710 pixels. So my question is how the brainstorm map each connection value to pixels?
Actually I have tried to run a deep convolutional model (2DCNN) that enforces a connection from that image pixels size.

Francois · November 11, 2021, 11:34am

[quote="ddaschakladar, post:3, topic:31739"]

However the figure size is 873x710 pixels. So my question is how the brainstorm map each connection value to pixels?

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/gui/figure_image.m#L783


      
          Data = GlobalData.DataSet(iDS).Figure(iFig).Handles.Data(ind{1}, ind{2}, ind{3}, ind{4});

          % Permute to get the two dimensions of interest as the first ones

          iPerm = [iDims, setdiff([1 2 3 4], iDims)];

          Data = permute(Data, iPerm);

          % Get rid of the unused dimensions

          Data = Data(:,:,1,1);

          % Get labels: dimension 1

          Labels{1} = AllLabels{iDims(1)};

          if ~isempty(Labels{1}) && (max(ind{iDims(1)}) < length(Labels{1}))

              Labels{1} = Labels{1}(ind{iDims(1)});

          end

          % Get labels: dimension 1

          Labels{2} = AllLabels{iDims(2)};

          if ~isempty(Labels{2}) && (max(ind{iDims(2)}) < length(Labels{2}))

              Labels{2} = Labels{2}(ind{iDims(2)});

          end

          % If maximum is not defined yet: Update figure handles

          if isResetMax || isempty(GlobalData.DataSet(iDS).Figure(iFig).Handles.DataMinMax)

              GlobalData.DataSet(iDS).Figure(iFig).Handles.DataMinMax = [min(Data(:)), max(Data(:))];

          end

          DataMinMax = GlobalData.DataSet(iDS).Figure(iFig).Handles.DataMinMax;

Actually I have tried to run a deep convolutional model (2DCNN) that enforces a connection from that image pixels size.

You should never use the displayed values from the figure.
Use the real data, saved in the TF matrix of the file.

ddaschakladar · November 11, 2021, 11:37am

Thanks sir for the information.