Hi Megan,

It seems to me that the most likely problem would be that associated with the vibrations of the speaker membrane  which would certainly not be ‘stationary’ in time. But do you have reason to believe that the artefact has a ‘stable’ waveform/topography at the sensors? 

As a mere user of maxfilter, all I could come up with is trying with more a aggressive correlation limit, I suppose you've tried lowering it? I know Jukka can do some magic in the SSS component space; if you have sampled the waveform going into the speaker, perhaps that could be used to identify/correlate ‘offending’ components and flag them for removal? Another related approach could be to run ICA on the tSSS’d data and try to see if a component would stick out (again: a sound waveform for correlation could help).

I’d appreciate it if you could send the list a follow-up on whether/how this gets sorted, it’s something many of us deal with from time to time ;)

Chris

--
Christopher Bailey, MSc
MEG Engineer, MINDLab Core Experimental Facility
Center of Functionally Integrative Neuroscience (CFIN)
Aarhus University, Denmark

On 1 Sep 2017, at 01.42, Megan Schendel <[log in to unmask]> wrote:

Hi Samu,
Thanks for your response.

I didn't realize I had only replied to Miikaa, earlier, describing our situation....

We are running a baby study where a speaker is positioned on the floor, well away from the MEG, but in the MSR.  We are not having success removing the artifact via tsss, and it's been troublesome creating a projector after the processing.  So we thought to try this.

It should be a stationary artifact, so I would expect a projector to work well.

We have a lot of data already collected in this manner, so too late to change speakers!

Does this affect your suggestion any?

Thanks again for your help!
Megan

MEG Technician
The Mind Research Network
1101 Yale Blvd. NE
Albuquerque, New Mexico 87106

On Wed, Aug 30, 2017 at 10:00 PM, Samu Taulu <[log in to unmask]> wrote:

Hi Megan,

In principle, you can apply SSP prior to movement compensation and tSSS. However, in practice one has to first compensate for the effects of the orthogonal projection on the sensor-level signals and only then apply (t)SSS. This can be done, e.g., by a minimum-norm reconstruction with SSP applied on the forward model vectors or by applying SSP on the SSS matrix before the actual SSS operation. There are a few extra things to take into account here, especially if the head is moving substantially.

It would probably be easier to calculate new SSP vectors from the (t)SSS-processed data and apply them as the last step to suppress any residual artifacts, such as blinks, that you may have in the processed data.

I hope this clarifies the situation. 

Best regards,

Samu

Samu Taulu, D.Sc.

Research Associate Professor, Physics

Director, I-LABS MEG Brain Imaging Center

Institute for Learning & Brain Sciences (I-LABS)

Portage Bay Bldg. Room 201E

Box 357988

University of Washington

Seattle, WA 98195-7988

USA

On Wed, Aug 30, 2017 at 1:03 PM, Megan Schendel <[log in to unmask]> wrote:

Hi all,

Is it possible to apply an ssp prior to movement compensation and tsss?

Thanks,

Megan

MEG Technician
The Mind Research Network
1101 Yale Blvd. NE
Albuquerque, New Mexico 87106