Co-registration

Dear Brainstormers,

I’m currently at the stage of checking the individual anatomy of each subject that I have imported to BS,
It turns out that my subject 8 had his head slightly inclined to the right (see the picture attached). What is tricky is that the IRM also show an inclination to the right, so I wondered if there could have been some problem where the setting I made of the fiducials were not taken into account. So I play a little bit to change the fiducials, but despite trying the more bizarre positions of NAS, LPA and RPA, the head didn’t change a lot relative to the helmet. So, my question is how does BS take into account what we set as fiducial when we import the anatomy? In fact, when I tried to distinguish what changed I wasn’t sure if it was the blue or the orange points. What correspond to each of these points?

In relation to this, I guess it is always recommended to refine the MRI registration using the head points, right?

[By the time I wrote this I received confirmation that this subject did have his head inclined along all the runs, but I thought it was worthy of ask anyway]

thanks for your feedback,

José Luis

Hi Jose Luis,

The blue dots you see are the fiducial points in the way they are set in the MRI: they look good on this image.
The red dots are the fiducial points that were digitized with the Polhemus and saved in the FIF file. They are in the same space as the green points (head shape) and the MEG sensors, they cannot move relatively to them.

The SCS/CTF coordinate system used in Brainstorm is based on the fiducials defined in the MRI. The LPA/RPA/NAS points in the MRI are the central reference for all the spatial information in Brainstorm, you should never modify them after having imported the MRI.

When you import your MEG runs, the position of the sensors are fitted in the best possible way on the MRI, with this logic:

  1. First the NAS/LPA/RPA from the Polhemus are aligned on the NAS/LPA/RPA of the MRI. This is used only as a first approximation.
  2. Then the automatic registration finds a better fit between the head shape (green points) and the head surface. It is indeed recommended to run this by default on all the runs, unless there are not enough head points, or something wrong happened during the digitization and their positions are obviously wrong wrt the sensor positions.

If you modify the NAS/LPA/RPA positions in the MRI (even a lot), then import the runs WITH the automatic registration based on the head points, you may end up having no difference at all.
The registration algorithm based on the head shape is now stable enough to correct very important errors in the placement of the fiducials in the MRI.
I guess this is why you observe no difference after changing the NAS/LPA/RPA positions.

Conclusion: I think you are doing everything correctly.
Please let me know if it leads to questionable results, there might still be small bugs around.

Cheers
Francois

Thank you Francois,
All this is very useful,
One last question. I actually modify the LPA/RPA/NAS points of the MRI after I have imported the anatomies. I was no entirely happy with some of these points, and then I changed it, but only a few milimeters. Should I re-load these anatomies?

thanks
José Luis

Hi Jose Luis,
It should be ok this way, I think everything should be updated correctly if you change the fiducial positions in the MRI viewer, but I’m not 100% sure.
Maybe it causes some problem at some point. It’s safer if you can re-import it, but if it’s one subject only, if you moved the points just from a few slides, and you have all your data imported and processed, it’s probably not necessary to re-import everything.
But try avoiding this in the future.
Francois