Error using SVD when computing the source

Discussions
1

Hello, when I computed the source [2018], I faced the following problem:

--------------------------content in the command window-------------------
BST_INVERSE > Depth weighting is only available for minnorm, ignoring option UseDepth=1
BST_INVERSE > Rank of the 'EEG' channels, keeping 59 noise eigenvalues out of 60 original set
BST_INVERSE > Using the 'median' method of covariance regularization.
BST_INVERSE > Covariance regularized by flattening tail of eigenvalues spectrum to the median value of 1.0e-05
BST_INVERSE > Using 'free' surface orientations
Error using svd
Input to SVD must not contain NaN or Inf.

Error in bst_inverse_linear_2018 (line 729)
[Ua,Sa,Va] =
svd(L{kk}(:,ndx),'econ');

Error in process_inverse_2018>Compute (line
645)
[Results, OPTIONS] =
bst_inverse_linear_2018(HeadModel,
OPTIONS);

Error in process_inverse_2018 (line 24)
eval(macro_method);

Error in panel_protocols>TreeInverse (line
1224)
[OutputFiles, errMessage] =
process_inverse_2018('Compute',
iStudies, iDatas);

Error in panel_protocols (line 44)
eval(macro_method);

Error in
tree_callbacks>@(h,ev)panel_protocols('TreeInverse',bstNodes,'2018')
(line 2436)
gui_component('MenuItem', jPopup, ,
'Compute sources [2018]',
IconLoader.ICON_RESULTS, ,
@(h,ev)panel_protocols('TreeInverse',
bstNodes, '2018'));
--------------------------content in the command window-------------------
PS: I used the individual MRI and the view seemed ok.
I used the neuroscan quickcap 64 channel electrode template (M1 M2 CB1 CB2 were deleted resulting 60 channels remained.) and used the openMEEG BEM model.
For source computing, since I was analyzing the resting EEG data, I just used the Identity matrix as the noise covariance matrix and made all other settings default.
(Under this condition, I think the source computing is not related to the EEG signal if I didn't calculate something like the data covariance matrix. Am i right? I am not very sure.)

At the same time I also checked the following issues to see whether it was caused by the existence of zero or nan in headmodel.Gain:

-------check part-----------
Data exported as "HeadmodelMat"

size(HeadmodelMat.Gain)

ans =

      64       45009

numel(find(HeadmodelMat.Gain==0))

ans =

 0

numel(find(HeadmodelMat.Gain==nan))

ans =

 0

----------------check part---------------------
image

image

So may I know what I should do next and anything I should check?

Looking forward to your kind reply and thanks in advance.

2

I guess you have a problem in the forward model.

To test correctly the Gain matrix for head model, you cannot use "==nan" because "NaN==NaN" returns 0 in Matlab, these are not numbers and are not equal to anything.

Instead, use:

nnz(isnan(HeadmodelMat.Gain))
nnz(isinf(HeadmodelMat.Gain))
3

Thanks so much for your kind reply.
I may have found the source of the problem. The electrodes of the original template did not touch the skin of the head model. So I did the MRI registration and selected 'project electrodes on surface' as well as 'refine registration using head points' operations. It seems all electrodes have touched the scalp and the source can be calculated properly. However, I am not very sure whether these operations were safe or any potential risk since I am not very confident about the automatic electrodes registration.

image
image698×603 67.3 KB

Looking forward to your reply and thanks again!

4

If you are using default electrodes positions from one of the templates in Brainstorm, do not use the function "Refined registration using head points": this is only for the electrodes positions have been digitized on the head of the subject.

The correct procedure:

  • Link your recordings to the database
  • Right-click on the channel file > Add EEG positions > ...
  • Right-click on the channel file > MRI registration > Edit.
  • Adjust the electrodes on the head of the subject: rotations, translations, rescaling, until it looks like the real positions of the EEG cap during acquisition (a photo of the participant with the cap on would help)
  • Then finally click on the button "Project electrodes on surface", to make sure the electrodes are in contact with the skin and not inside the head (which causes issues in the forward modeling)
5

Thanks so much and I will have a try following your instructions.