Merge same conditions from different runs (MEG)

Discussions
1

Hi to all,

I have 10 runs of four conditions per subject recorded with Neuromag. I want to merge same conditions from different runs. However, while the head points digitized with Polhemus are the same for all 10 runs, the MEG sensors’ position referenced to the head (“Neuromag channels” in the BS database) is different for each run. During the recordings the MEG system automatically checked the position of the coils for each run and allowed it to be in an acceptable range.
So, how can I merge the same conditions from different runs with different Neuromag channels? Or should I first retrieve the maps of each condition and each run and then merge the maps?

Thanks in advance for your reply,
Chrysa Lithari

2

Hi Chrysa,

You have two options for combining information from different acquisition runs:

  1. At the [B]sensor level[/B]: drag all your files to the “Process” list, and select the process “Standardize > Co-register MEG runs”:
  • This is a quick way to correct for small movements across runs, the fields for each run are re-interpolated on a common head position
  • 1 head model / 1 inverse model
  • The method is currently under evaluation.
  • It should be used only for small head movements between runs (<= 1cm).
  1. At the [B]source level[/B]: first estimate the sources for all the conditions and runs independently, then average the sources.
  • nRuns x nConditions head models and inverse models
  • This is much slower, more difficult to handle in terms of file managing, and it takes much more storage space.
  • The result is supposedly more accurate.
3

Hi Francois,

thanks for your reply.
I am going to follow option 1 for now. Are you planning to evaluate this method and have some feedback soon?
Is there any way to know how much was the head moved from one run to the following? Is this information in the .fif files?

Thanks,
Chrysa Lithari

4

Hi Chrysa,

A student is trying to find good metrics to evaluate this function. It may take some time. But for small displacements, I think you can use it without worrying too much.

If you use the newest brainstorm version, you can see after calling the process “Co-register MEG runs” what was the maximum displacement for one sensor in the report window.

To get a visual estimation of the subject’s movements between two runs: select the channel files for the two runs at the same time (holding the CTRL key), right click > Display sensors

Cheers,
Francois

5

Hi again,

I have imported the individual anatomy (MRI + meshes) for a subject. Every time I import a new run (fif file containing Polhemus points), I am asked to allow the ICP algorithm to run in order to refine the registration. Do I have to run it for every run, although I intend to co-register the runs later?

thanx again,
Chrysa

6

Each of those runs have different head positions. You need to have all the runs as accurately registered as possible before running the process “Co-register MEG runs”.
So yes, you need to do the registration for all the runs.

7

Thank you for your reply.
I am sorry to bother you again, but I am new and I have desperately many data…
I realized that I can import more than one runs together for each subject. Since the event channel is the same for all runs, Brainstorm (which is wonderful) groups correctly the trials of the same conditions.
However, the head position relatively to the sensors is different. So, if I import all my runs together, then there is no option for coregistration; the channel file is common. Is this right?

thanks,
Chrysa

8

Hi Chrysa,

You define the way the channel files are shared between subjects and/or runs by setting the properties of the subject. Double-click on the subject (or right-click > Edit subject), and see the option “Default channel file”. Click on the Help button for more information.

Unfortunately, there is no way to classify at the same time the runs and the conditions. The folders you see for each subject represent either experimental conditions OR runs.
There is one level of definition missing in the database structure. We will work on this problem at some point, but I don’t have any quick solution to help you right now.

Cheers,
Francois

9

Hello, I have bumped into this Discussion because I also have 6 runs per subject collected with Elekta neuromag and the headpoints were taken once at the beginning of the experiment. The tsss files were MaxFiltered. Before this post, I was computing a head model per run and noticed the Gain matrix was not exactly the same between runs, I first considered your option 2, but it is indeed slow, so I now looki into your option 1 and I was wondering why the headmodel Gain matrix is still not the same between runs? Shouldn't it have created one head model at the Subject level?
Thank you!