Importing Fieltrip structure from HCP website

Bugs
1

Hello,
When importing the pre-processed data from the HCP database, these are provided as fieldtrip structure. However, there seems to be a problem with the channel file, because all sensors are located in one place. When looking into the channel file structure, all channels have a very long location and orientation matrix (3*269), which seems to be the same for all channels.
If you need further information, let me know.
Best wishes,
Esther

2

Hi Esther,

I haven’t tried importing any of the pre-processed data. The only thing that I did was to make sure we could import the transformation matrices correctly. The only things I would be able to investigate are errors occurring during one the steps of the HCP tutorial: http://neuroimage.usc.edu/brainstorm/Tutorials/HCP-MEG

Please let me know at which step of this tutorial you are getting unexpected results, and with which subjects it occurs so I can try to reproduce the issue on my end.

Thanks
Francois

3

Hi Francois,
it is not within the described pipeline. That works fine.
However, the pre-processed data are in a fieldtrip structure, which brainstorm seems to allow me to import. But then the channel information is completely messed up. The data I have been looking into are 100307_MEG_3-Restin_rmegpreproc. So could this be a problem with the import of fieldtrip data structures? If it helps I can also point you to the data located in Montreal.
Best wishes,
Esther

4

Hi Esther,
I’m having trouble downloading data from the connectomedb website right now.
Please send me directly one of the files you are having issues with, describe how you import it and what you think is wrong with it.
Thanks
Francois

5

I looked at this file and realized there is some form of conversion that I don’t know how to do here.

In the FieldTrip files, the field .grad.tra indicates how to relate the channel names (.grad.label) and the corresponding coils positions (.grad.coilpos and .grad.coilori). In the case of gradiometers, one channel corresponds to two coils and two positions. I wrote the function in_data_fieldtrip.m with examples from Elekta recordings, where the .tra was only indicating the weights to use for each combination of coils, to get the channel value.

But with the 4D files, it looks like the compensation matrices and/or other SSP/ICA projectors are also mixed in this .tra matrix. I don’t know how to get the correspondence channel-coils in this case.
In Brainstorm we need the projectors and coil weights available separately because we save and process them separately.

I posted a message on the FieldTrip mailing list, but I may not get any interesting answer.
So for now, I added a warning, and started using the channel positions instead… Which is wrong…
I’m not sure how far we can go with these files.

Francois

6

Hi Francois,
many thanks for looking into this!
How wrong would you say is the current way of importing? The channel locations are in the correct places, but what is missing is the consideration of the ICA/SSP projectors for the source modeling, correct?
Best wishes,
Esther

7

Unfortunately, the current way of proceeding is completely wrong. It is a lot of the sensor information (positions of the coils, 4D noise compensation matrix, ICA/SSP information) that we cannot import properly, because I don’t know how to separate them.

As expected, no answer from the mailing list:
https://mailman.science.ru.nl/pipermail/fieldtrip/2017-May/011546.html

Trying again with the bug tracker:
http://bugzilla.fieldtriptoolbox.org/show_bug.cgi?id=3298

You could also trying posting the same question to the HCP support.

8

I saw that there was some reaction to the bug report. So I guess I will wait before posting to the HCP support.

9

Hi Francois,
I just checked on the fieldtrip page and saw that there is a further suggestion. Does that help to import the fieldtrip structures into brainstorm?
Thank you,
Esther

10

Hi Esther,
Maybe, but I haven’t had time to check this yet.
I will in the next 2 weeks
Francois