We have a Yokogawa/KIT system with MEG160 and have been trying to import our data into Brainstorm with no avail. We have so far tried the following:
(1) Importing it directly as a *.sqd file; this tells us, unsurprisingly, that “The file have not been exported properly from the Yokogawa software. The MEG/MRI co-registration information is missing. Please export data by using the “Third Party Export” function in advance…”
(2) Following Brainstorm’s Yokogawa/KIT data-import tutorial; this failed because our MEG160 software does not have a menu item for either “BESA Binary Export” or “Third Party Export” or anything that resembles the description in the tutorial (though it does have an ordinary BESA Export option for files such as *.sfp). We called Yokogawa to ask about this (as recommended by the tutorial), but they no longer support the MEG system or the MEG160 software and insist that the only way to obtain support for the system or software is to write a (snail-mail) letter to a separate company in Japan.
(3) Converting the *.sqd data to *.fif data using MNE-Python then importing it as Elekta-Neuromag data. MNE doesn’t seem to have a problem reading the data, but this gives us the following error: Error in mne_add_coil_defs.m (line 147); Could not find an MEG coil template (coil type = 6002 accuracy = 1) for channel MEG 158.
Does anyone know of any solution to any of these problems or another good way to import Yokogawa/KIT data into Brainstorm?
Thanks in advance!
If the file hasn’t been exported using the menu indicated in this tutorial (Third Party Export), there is no way to get the correct head position from the .sqd file using the Yokogawa reader (library the company provided). We can have the positions of the head localization coils, but in an arbitrary coordinates system. We do not have access to the digitized head points.
Maybe they have implemented in MNE-Python another reader, that does not rely on the Yokogawa library. I’ll ask them.
Have you checked if the registration with the head is correct in MNE?
This makes sense; to get things working in MNE, we provide two additional files: *_HS.txt and *_Points.txt. Both are ascii files containing rows of (x,y,z) values exported by FastSCAN; the HS file contains approximately 60,000 rows and the Points file contains 8. Our MNE code passes these as the hsp (“head-shape file”, the HS file) and elp (“marker coil file”, the Points file) options to the mne.io.read_raw_kit function in MNE-Python. MNE does appear to co-register everything correctly. We have played around a bit with importing these files directly into Brainstorm (it is slow, but seems to work); unfortunately the imported points were clearly not registered to the MRI data, and we weren’t able to find a way to fix this.
Sorry for the response delay, I didn’t forget you.
I’ll try to make both approaches work.
Can you send me examples of the files you would like to read in Brainstorm?
The original .sqd file + the extra files _HS.txt and _Points.txt
=> Do they always have the same name? Can I guess what are the files to import just from reading the content of the folder?
The file convert to .fif with MNE-Python (we need to add this sensor 6002 to the file coil_def.dat)
Upload the files somewhere, then send me the link a separate email (click on my username on this forum).
PS: I already changed the function from that was generating the error "Could not find an MEG coil template (coil type = 6002 accuracy = 1) for channel MEG 158."
Now it shows a message in the Matlab command window instead.
You’ll be able to see the signals but no topography or source modeling.
We’ll have to wait until Matti Hamalainen replies to my other email asking to add sensor type #6002 to the file coil_def.dat
You can update Brainstorm if you want to try this (menu Help > Update Brainstorm).