 |
 |
I agree with John. Surface Laplacian could give you more interpretable sensor data. However, with 19 electrodes it's also not ideal. Vladimir On Mon, Oct 30, 2017 at 3:20 PM, nasibeh talebi <[log in to unmask]> wrote: > > > Dear Vladimir > > Thanks a lot for your help. > I will try your comments. > > Best regatds. > > Nasibeh > > On Mon, Oct 30, 2017 at 6:01 PM, Vladimir Litvak > <[log in to unmask]> wrote: > Dear Nasibeh, > > Connectivity analyses are challenging even with a large number of > channels. Projecting your data to source space does not get rid of volume > conduction. I would recommend you to use a minimum norm solution and > perhaps carefully choose your connectivity measure to, for instance, only > use the imaginary part of the cross-spectrum. > > Best, > > Vladimir > > On Mon, Oct 30, 2017 at 2:21 PM, nasibeh talebi <[log in to unmask]> > wrote: > > Dear Vladimir > > Thanks a lot for your helpful advice. Actually, I am going to perform a > connectivity study with this dataset, and as the connectivity measures are > affected by the "volume conduction" effect, I want to extract the > underlying sources of the scalp potentials. > I am wondering if I can have your comments. > > Best, > > Nasibeh > > > > > On Monday, October 30, 2017, 5:39:21 PM GMT+3:30, Vladimir Litvak < > [log in to unmask]> wrote: > > > Dear Nasibeh, > > 19 electrodes is indeed a pretty low number. If the spatial solution > cannot be well estimated then also the time courses will not be separated > well. If you are just interested in the temporal dynamics it might be > better to stay at the sensor level. > > I don't think using a coarse mesh can change anything in this case, but > perhaps using IID or COH solution (the latter similar to LORETA) will be > more robust with a small electrode number than MSP. > > Best, > > Vladimir > > On Mon, Oct 30, 2017 at 1:53 PM, nasibeh talebi <00000a9d20e1b6de-dmarc- > [log in to unmask] <[log in to unmask]>> > wrote: > > Dear Eugenio > > Thanks a lot for your great help. Would you please help me with these two > question: > > 1. The low accuracy problem of the source reconstruction (due to the > small number of EEG electrodes ) is only related to the spatial > specification of the sources, or their temporal activity is also poorly > estimated? > 2. for a small number of electrodes, is it preferable to use "*coarse *cortical > mesh" instead of *normal*? > > Best regards > Nasibeh > > > > > On Monday, October 30, 2017, 11:56:14 AM GMT+3:30, Eugenio Abela < > [log in to unmask]> wrote: > > > Source recon from 19 electrode EEG is tricky because of low spatial > sampling; be careful when interpreting the data. > > That being said, your time series looks ok to me - SPM applies a Hanning > window per default, so the data will have highest amplitude in the middle > of the trial and be damped towards beginning and end. This can be switched > off; have a look at the EEG section of the SPM manual > > Hope that helps > - Eu > > Am 30.10.2017 um 08:17 schrieb nasibeh talebi <00000a9d20e1b6de-dmarc- > [log in to unmask] <[log in to unmask]>>: > > > Dear SPM experts > > > > Hi > > > > I have a 19-channel *resting state* EEG dataset. The EEG is recorded for > about 120s at the sampling rate of 128Hz. > > I want to perform the 3D source reconstruction and extract the temporal > activity of the sources. I used the *MSP* algorithm in SPM, but the > results are strange (attached file). > > Does SPM have an appropriate source reconstruction algorithm for resting > state EEG? or should I choose an alternative toolbox? > > > > Best regards > > Nasibeh > > <td5_sourceSample_MSP.tif> > > > >