Dear Vladimir,


Thank you for your response.


The averaged eyeblink file shows what looks like an eyeblink.

The eyeblink looks to be detected, the amplitude in some files is around 6000. In the EEG data, there is a large jump of noise across all channels at the same timepoint. Could this be interfering with modelling the blink?


Many thanks,






From: Vladimir Litvak [mailto:[log in to unmask]]
Sent: 13 June 2015 15:35
To: Emily Hird
Cc: [log in to unmask]
Subject: Re: [SPM] EEG topography-based artefact correction question


Dear Emily,


I think there is no way around examining your data by eye and seeing what really happens there for those problematic files. Are there really eyeblinks? What is their typical amplitude? Are they detected by eyeblink detection function? Does the eyeblink average really look like an eyeblink? etc. You might want to play with the threshold in the eyeblink function to make sure that the average really looks like an eyeblink even if less eyeblinks are detected. For topography based correction it's not critical to detect all of them.






On Sat, Jun 13, 2015 at 5:21 AM, Emily Hird <[log in to unmask]> wrote:

Dear SPMers,


I have a question about removing eyeblink artefacts using the topography-based artefact correction described in 'Advanced topics in M/EEG artefact removal chapter (40) of the SPM12 manual.

I am cleaning EEG data.


On the majority of my data files, the topography-based correction method has worked well. However, on some files, the method fails to detect components on the topography. I successfully model the eyeblink using the 'detect artefacts' batch, and successfully epoch around the eyeblinks. The issue happens when I run 'define spatial confounds'. No blink-related components are detected. 


In some files, the topography is blank but for a few small areas. This suggests a strong artefact is creating so much noise in the data that the component is not being detected. However, I tried running a lenient (150uv) artefact rejection on the file, this causes all trials to be rejected. Running an 80uv artefact rejection on the normal (non eyeblink) data file does not cause all trials to be rejected.

Also, marking the artefacts and rejecting using the 'threshold Z-scored difference data' function (also described in chapter 40 of the manual) does not change the fact that the components are not detected.


Would anyone have any advice on how to resolve this issue? I would be very grateful for any ideas.


Many thanks!

Emily Hird