Inverse Problem : Computing Kernel*DataFile

unes1111 · July 26, 2018, 8:43pm

Hello,
The code bellow is from process_inverse_2018 .mat file.
I originally thought once inverse model is computed, the result of the kernel is sent to this function where in here source is computed by multipliying the kernel by the data at each point. However, I feel like this isn't the code that is computing because when I changed ResultsMat.ImageGridAmp output, the output of Display on cortex didnt change. I even added a line of a=3 on matlab to see if I can see the output on command line but I didnt.

Can you please tell me what is wrong and direct me to the correct code where Kernel*Measurment occures?

Also , what if instead of raw data we are looking at average spike, is there another code where kernel is multiplied by average spike ?

thank you for your helps

Francois · July 27, 2018, 12:21pm

The piece of code you copied here is executed only when the option ComputeKernel is set to 1, which corresponds to the advanced minimum norm option “Full results”.

In most cases, ImagingKernel is multiplied with the data when the source file is loaded for display or computation.

Search the code for “ImagingKernel *” to see where else this variable is used (the tool “Find files” from Matlab - CTRL+SHIFT+F, with the option “include subfolders” is a good tool for this prupose).
You’ll find this: bst_memory.m (line 1957), in_bst.m (117) and in_bst_results (line 275)

tengchaolin · July 30, 2018, 1:57am

@Francois
Hello, I have consructed head model. I also obtained forward and inverse matrix,that is ImagingKernel and Gain, from OpenMEEG BEM and MN: EEG(Unconstr) in corresponding file, respectively. Now, I want to compute the activity on the cortext. Is ImagingKernel * EEGdata correct way to acqurie cortical electrical activity ? Moreover, If I want to compute scalp potential, Is ImagingKernel * EEGdata *Gian correct way to acqurie ? If it is correct, but the results is not equal to EEGdata ? Can you tell me the correct way if I am wrong?

Thank you for you helps.

Marc.Lalancette · July 31, 2018, 2:05pm

I believe what you want is: Gain * ImagingKernel * EEGdata. However, this will not give exactly the EEGdata back, but hopefully something close if the data is clean.

Francois · August 1, 2018, 4:37pm

Is ImagingKernel * EEGdata correct way to acqurie cortical electrical activity ?

Yes, see the code of the functions I indicated previously.

Is ImagingKernel * EEGdata *Gian correct way to acqurie ? If it is correct, but the results is not equal to EEGdata ?

The model does not explain all your EEG data. You can see the modeled data by right-clicking on the source file > Model evaluation > Simulate recordings.
The corresponding code:

github.com

brainstorm-tools/brainstorm3/blob/master/toolbox/math/bst_simulation.m#L163


      
          end

          % Display mesasge to illustrate the 

          disp(['BST> Simulation: Using head model file "' HeadModelFile '"']);

          % Constrain head model to match the source model (normal to the cortex or mixed)

          switch (nComponents)

              case 0,   HeadModelMat.Gain = bst_gain_orient(HeadModelMat.Gain, HeadModelMat.GridOrient, GridAtlas);

              case 1,   HeadModelMat.Gain = bst_gain_orient(HeadModelMat.Gain, HeadModelMat.GridOrient);

          end      

          

          

          %% ===== SIMULATION LOOP =====

          % Get time vector

          TimeVector = bst_memory('GetTimeVector', iDS, iResult, 'UserTimeWindow');

          % Maximum number of time points to process at once

          blockSize = 100;

          % Compute number of blocks

          nBlocks = ceil(length(TimeVector) / blockSize);

          % Initialize simulated matrix

          nTime = length(TimeVector);

          F = zeros(size(HeadModelMat.Gain,1), nTime);

          % Progress bar

tengchaolin · August 5, 2018, 4:33am

Thanks very much. And I also appreciate the detailed code you provided.

Kirill · January 13, 2019, 8:14pm

Hello!
Could you explain or give articles why do we multiply by raw data when we make an invesre solution (MNE or sLORETA)?
How does the notion Regularization work?

Francois · January 14, 2019, 8:02am

Minimum norm: Baillet S, Mosher JC, Leahy RM
Electromagnetic brain mapping, IEEE SP MAG 2001.

dSPM: Dale AM, Liu AK, Fischl BR, Buckner RL, Belliveau JW, Lewine JD, Halgren E
Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity. Neuron 2000 Apr, 26(1):55-67

ncbi.nlm.nih.gov

Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity.

AM Dale, AK Liu, BR Fischl, RL Buckner, JW Belliveau, JD Lewine and E Halgren, Neuron, Apr 2000

Functional magnetic resonance imaging (fMRI) can provide maps of brain activation with millimeter spatial resolution but is limited in its temporal resolution to the order of seconds. Here, we describe a technique that combines structural and functional MRI with magnetoencephalography (MEG) to obtain spatiotemporal maps of human brain activity with millisecond temporal resolution. This new technique was used to obtain dynamic statistical parametric maps of cortical activity during semantic processing of visually presented words. An initial wave of activity was found to spread rapidly from occipital visual cortex to temporal, parietal, and frontal areas within 185 ms, with a high degree of temporal overlap between different areas. Repetition effects were observed in many of the same areas following this initial wave of activation, providing evidence for the involvement of feedback mechanisms in repetition priming.

sLORETA: Pascual-Marqui RD
Standardized low-resolution brain electromagnetic tomography (sLORETA): technical details, Methods Find Exp Clin Pharmacol 2002, 24 Suppl D:5-12

ncbi.nlm.nih.gov

Standardized low-resolution brain electromagnetic tomography (sLORETA): technical details.

RD Pascual-Marqui, Methods and findings in experimental and clinical pharmacology, 2002

Scalp electric potentials (electroencephalograms) and extracranial magnetic fields (magnetoencephalograms) are due to the primary (impressed) current density distribution that arises from neuronal postsynaptic processes. A solution to the inverse problem--the computation of images of electric neuronal activity based on extracranial measurements--would provide important information on the time-course and localization of brain function. In general, there is no unique solution to this problem. In particular, an instantaneous, distributed, discrete, linear solution capable of exact localization of point sources is of great interest, since the principles of linearity and superposition would guarantee its trustworthiness as a functional imaging method, given that brain activity occurs in the form of a finite number of distributed hot spots. Despite all previous efforts, linear solutions, at best, produced images with systematic nonzero localization errors. A solution reported here yields images of standardized current density with zero localization error. The purpose of this paper is to present the technical details of the method, allowing researchers to test, check, reproduce and validate the new method.