Hi! Thank you for the useful videos, I am a fan. At 9:54 when you talk about low spatial frequency attenuation by Laplacian, why does the Power at low spatial frequency higher than that at a high spatial frequency ?. It seems counter-intuitive. What am I missing?
Hi Radha. That's an empirical observation in neuroscience -- lower spatial frequency activity tends to have larger power. It's probably a combination of brain organization plus the intrinsic biophysics of EEG: small-scale activity is too small to be picked up from outside the head, while larger-scale activity produces fields that are powerful enough to be measured through tissue types and at a distance.
Hello Mike! Thanks for this new version of your course. I have been watching your videos before the release of the new version, they are both great! I have couple of questions about Surface Laplacian: 1) Apart from functional connectivity, how wise it is to apply SL on ERP data? Or for time-frequency analysis? 2) In your old version of the course, you mentioned that after SL, the unit changes to uV per unit area square (mostly cm^2). Does this change in unit change anything about our interpretation of the EEG data? 3) You also mentioned in the old version to pay attention to the unit issue in case we are going to do TF analysis after SL. In other words, no normalize the data before TF analysis. Can we for example do: 10*log10(channel_data_in_uV_per_cm2)? Apologies if any of my questions doesn't make sense
Hi Haneen. 1) Yes, it's a good idea. SL improves SNR overall and also attenuates some artifacts like muscle activity. You can look up a special issue on the SL in International Journal of Psychophysiology from several years ago. 2) Nope, not really. It's still changes in voltage over time. 3) Yes, that's good. I generally recommend normalizing time-frequency power data anyway. If you apply a baseline normalization, the original units don't matter.
@@mikexcohen1 Thank you Mike for your detailed answer! Really appreciate it. I actually just paid attention that CSD activity is still a voltage and thus converting it to dB before TF analysis must be done using 20*log10(channel_data_in_uV_per_cm2) since it is voltage and not power.Is this correct or am I missing something?
Hi Mike! This is a great video-series. I noticed that in your book (which is also great) you mention that the SL filter should not be applied to MEG data. Would you mind elaborating on why this is the case, or refer me to a source that does? Many thanks!
Hello Mike I have registered now for the course but a have a question I am doing my preprocessing on EEGLAB is surface Laplacian is available to be done on EEGLAB. I have been searching and I found a CSD toolbox, could you please clarify how to use?
Hi Lina. Yes, you can use the CSD toolbox. I have a function that also implements the same method (from Perrin in the 1980's) that is a bit more efficient than the CSD toolbox. You can get that from my book code (available from my website sincxpress.com; you don't need to buy the book to get the code), and it's also available in some of my online courses.
I have been watching a couple of your videos and they are brilliant! Really helps me grasp these technical concepts! You are the best!
Thank you kindly, Shehani. I'm glad you're finding the vids useful :D
Hi! Thank you for the useful videos, I am a fan. At 9:54 when you talk about low spatial frequency attenuation by Laplacian, why does the Power at low spatial frequency higher than that at a high spatial frequency ?. It seems counter-intuitive. What am I missing?
Hi Radha. That's an empirical observation in neuroscience -- lower spatial frequency activity tends to have larger power. It's probably a combination of brain organization plus the intrinsic biophysics of EEG: small-scale activity is too small to be picked up from outside the head, while larger-scale activity produces fields that are powerful enough to be measured through tissue types and at a distance.
Hello Mike! Thanks for this new version of your course. I have been watching your videos before the release of the new version, they are both great! I have couple of questions about Surface Laplacian:
1) Apart from functional connectivity, how wise it is to apply SL on ERP data? Or for time-frequency analysis?
2) In your old version of the course, you mentioned that after SL, the unit changes to uV per unit area square (mostly cm^2). Does this change in unit change anything about our interpretation of the EEG data?
3) You also mentioned in the old version to pay attention to the unit issue in case we are going to do TF analysis after SL. In other words, no normalize the data before TF analysis. Can we for example do: 10*log10(channel_data_in_uV_per_cm2)?
Apologies if any of my questions doesn't make sense
Hi Haneen.
1) Yes, it's a good idea. SL improves SNR overall and also attenuates some artifacts like muscle activity. You can look up a special issue on the SL in International Journal of Psychophysiology from several years ago.
2) Nope, not really. It's still changes in voltage over time.
3) Yes, that's good. I generally recommend normalizing time-frequency power data anyway. If you apply a baseline normalization, the original units don't matter.
@@mikexcohen1 Thank you Mike for your detailed answer! Really appreciate it. I actually just paid attention that CSD activity is still a voltage and thus converting it to dB before TF analysis must be done using 20*log10(channel_data_in_uV_per_cm2) since it is voltage and not power.Is this correct or am I missing something?
Yeah, "must" is a strong word. I generally recommend normalizing power to eliminate the original units of the data. But it's not always necessary.
Hi Mike! This is a great video-series. I noticed that in your book (which is also great) you mention that the SL filter should not be applied to MEG data. Would you mind elaborating on why this is the case, or refer me to a source that does? Many thanks!
The SL is a model of voltage spreading over the scalp. MEG isn't voltage ;)
Hello Mike
I have registered now for the course but a have a question I am doing my preprocessing on EEGLAB is surface Laplacian is available to be done on EEGLAB. I have been searching and I found a CSD toolbox, could you please clarify how to use?
Hi Lina. Yes, you can use the CSD toolbox. I have a function that also implements the same method (from Perrin in the 1980's) that is a bit more efficient than the CSD toolbox. You can get that from my book code (available from my website sincxpress.com; you don't need to buy the book to get the code), and it's also available in some of my online courses.