Plot actual values on 2-D disk

ebock · November 6, 2017, 10:42pm

Hi. Is it possible to plot a 2D disk with the actual values in a file? In this example, I have selected three channels that appear to have three different colors on the topo, but the values indicate I should have two red and one yellow. I guess there is some interpolation? Can I turn that off? It's a coherence result.

Francois · November 7, 2017, 8:01am

Hello,

There are 3 interpolations stacked in a typical “2D Disc” figure on MEG data:

The magnetic interpolation used to smooth the values based on a simple inverse/forward model. To turn this off, right-click on a file > MEG > No magnetic interpolations > 2D Disc. But this should not be applicable for connectivity or time-frequency values. Let me know if you have access to this option with your file, this is an error.
The interpolation of the 2D Disc: this is a stretched display, the values are interpolated from the sensors to a regular “disc” grid. This causes some smoothing. To turn this off: use “2D sensor cap” instead.
We have a discreet set of values, associated to the vertices of the grid in the “2D sensor cap”. The values are interpolated over the faces by Matlab (FaceColor=‘interp’). It is not possible to produce a 2D plot without any interpolation of this type, otherwise we cannot get a continuous surface, but simply discreet points.

Francois

ebock · November 7, 2017, 4:31pm

OK, Thanks.
1). I do not have access to this interpolation from the connectivity plot.
2). Yes, I understand
3). This makes sense if the values are on a continuum, however, they are phase values, that I have computed and I don’t want to have fake interpolated phases in between two real values. I am trying to replicate a plot from a publication. I will have to find a way to extend the actual values along the surface.
Thanks!

Francois · November 8, 2017, 9:40am

This is not possible with the 2D/3D display functions with Brainstorm: the color values are always set “per vertex”, which means that to paint a face, there is necessarily a color interpolation between the 3 vertices of this face. Setting the color per face instead of per vertex is not as easy as it sounds, I can’t think of anything that would do this easily.

Option #1: You create a patch object with a ‘FaceVertexCData’ matching the number of faces (but I don’t know how you can set the values of the faces…)
https://github.com/brainstorm-tools/brainstorm3/blob/master/toolbox/gui/figure_3d.m#L3430

Option #2: You let Matlab compute one value per face, and disables the smooth interpolation across faces, by setting the property ‘FaceColor’ to ‘flat’ instead of ‘interp’ (https://github.com/brainstorm-tools/brainstorm3/blob/master/toolbox/gui/figure_topo.m#L168)

Option #3: You use a very high resolution for the sensors patch (add many tess_refine here: https://github.com/brainstorm-tools/brainstorm3/blob/master/toolbox/gui/figure_topo.m#L453), then replace the interpolation function with a nearest neighbor search instead of a distance-weighted interpolation (https://github.com/brainstorm-tools/brainstorm3/blob/master/toolbox/gui/figure_topo.m#L617). You would still have transitions between colors, but you could hope that the transitions are shorter. But I’m not sure the final rendering would look good…

Option #4: You set your sensors to “EEG” in the channel file, keep only one location per sensor and delete the orientation field, and display it with the “3D electrodes” mode: http://neuroimage.usc.edu/brainstorm/Tutorials/Epilepsy#Averaging_spikes (not very convincing…)

Sorry for not being more helpful.

Georgios · April 24, 2020, 12:45am

Hello,

I would like to follow-up on this topic to ask about the options that we have with the EEG and for any new features introduced in brainstorm.
I have done power analysis on specific time segments of an EEG recording and I want to plot the results in 2D disc/cap.

What kind of interpolation is used when plotting power with the 2D disc/cap?
Is there a way of making it less smooth?

Thank you so much,
Georgios

Francois · April 24, 2020, 9:24am

What kind of interpolation is used when plotting power with the 2D disc/cap?

The smoothing over the EEG cap is computed here:

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/gui/figure_topo.m#L716


      
              iPreStim = find(TimeVector < 0);

              if (length(iPreStim) > 50) && ~any(iPreStim > size(F,2))

                  FpreStim = F(:, iPreStim);

              else

                  FpreStim = F;

              end

              % This function does everything that is needed here

              WExtrap = channel_extrapm('GetTopoInterp', ChannelFile, selChan, Vertices, Faces, bfs_center, bfs_radius, FpreStim);

          % No magenetic extrap: Compute interpolation function from sensors to patch surface (simple 3D interp)

          else

              % Detect the points at (0,0,0)

              precision = 1e6;

              iChanZero = find(all(abs(chan_loc) < 1/precision, 2));

              % Check if some sensors are located at the same position

              GoodChanLoc = chan_loc;

              GoodChanLoc(iChanZero,:) = [];

              uniqueChanLoc = unique(round(GoodChanLoc * precision)/precision,'rows');

              % If number of unique positions is different from total number of channels, there are multiple sensors at the same place

              if (size(uniqueChanLoc,1) ~= size(GoodChanLoc, 1)) && ~isequal(TopoInfo.Modality, 'NIRS')

                  bst_error('Two or more sensors are located at the same position. Please try to fix this problem.', 'Plot topography', 0);

              end

Is there a way of making it less smooth?

Yes, with the menu "No smoothing":