MEG dataset for Yokogawa system (co-registration problem)

[Ryohei Fukuma]

Dear all,

I've been working on a MEG dataset in BIDS format using Yokogawa system.
I think I found a mistake in the Appendix VIII.
Moreover, I have a question relating to the co-registration.

In the recorded file using Yokogawa system, the sensor locations are recorded using coordinate relative to the Dewar centre.
This is written in FAQ of the fieldtrip toolbox (and also in API manual for Yokogawa systems; contained as a pdf file in the fieldtrip toolbox).
https://www.fieldtriptoolbox.org/faq/coordsys/
Actually, when I compared the sensor locations in the recordings of two different subjects, they matched.

On the other hand, the "Appendix VIII: Coordinate systems" says "KitYokogawa: ALS orientation and the origin between the ears".
I think this is a mistake.

A question rises here.
What is the expected coordinate system for AnatomicalLandmarkCoordinateSystem, DigitizedHeadPointsCoordinateSystem, and HeadCoilCoordinateSystem in the coordinate system JSON file (*_coordsystem.json)?
This is not explicitly written in the specification; but, as far as I read the document, expected coordinate seems to be as same as the coordinate system in MEGCoordinateSystem for the co-registration purpose.
For a MEG system that saves MEG sensor coordinates in a coordinate system relative to head, it is natural that these anatomical locations (landmarks, digitized head points, and head coils locations) are saved with same coordinate system.
Actually, in the current BIDS format, one session are allowed to have only one *_coordsystem.json, and multiple runs in a session share the same locations in the JSON file.
However, for the Yokogawa system, which saves the MEG sensor coordinates in its system coordinates, the anatomical locations cannot be shared between runs due to the head movement.
Is this correct understanding of the specification?

If so, I think there are several options for my dataset:
1. divide the runs so that each session in the dataset has only one run. In this way, *_coordsystem.json will not be shared between runs anymore.
2. supply the locations with some footer in the  *_coordsystem.json (e.g. "NAS-run1")
3. supply the anatomical locations for each run as a derivative file (and do not supply these locations in the *_coordsystem.json).

Which one is better option to take?
For machine readability, option 1 seems to be better.
If this is just misunderstanding of the specification, or if there is better method, please give me comment.

I really appreciate your help.


Yours sincerely,

Ryohei Fukuma

[Robert Oostenveld]

Dear Ryohei

Let me first give a disclaimer: none of the authors on the BIDS-MEG paper, including me, had specific expertise or experience with the Yokogawa (or Ricoh) system. Furthermore, the lab procedures and the interaction between the MEG scanner hardware, acquisition software and the original Yokohama processing/aanalysis software is not known to me; as far as I know the Yokogawa system- and MEG-VISION software documentation is not available outside the labs that have such a system.

On https://www.fieldtriptoolbox.org/getting_started/yokogawa I read that the Yokogawa system defines a MEG dewar coordinate system that is not linked to the head, but also that subsequent analysis is done using an ALS head coordinate system that is linked to the head. The latter ALS head coordinate system is specified in the FieldTrip coordsystem faq.

If you look at the BIDS coordinate system appendix, it only lists a number of vendor-specific head coordinate systems. Each of these systems also has its own device coordinate system, which for SQUID based systems we often refer to as the “dewar" corodinate system (although for OPMs we’ll have to start thinking about another term, since those don’t have a dewar with liquid helium).  

The goal of BIDS for raw data is to organize and document data such that it allows for maximal flexibility in the subsequent analysis. In an MEG lab we typically do a few recordings (depending on the system with less or more integrated acquisition software): we record the position of the anatomical fiducials, of the head localizer coils, of the scalp surface - all of these using a Polhemus or similar - and the location of the head localizer coils relative to the MEG sensors is determined using the MEG sensor array. 

In the Neuromag acquisition software (which is the most integrated) all of these are stored along with the MEG brain data in the same .fif file, and are all following automatic coregistration expressed relative to the anatomical fiducials (i.e., nose and ears). 

The CTF acquisition software is not as well integrated and runs on two separate computers (one Linux, another Windows). Following automatic coregistration the position of the MEG sensors is expressed in the .ds dataset relative to the headcoils. However, the headcoils are often placed at the fiducials (although procedures may differ between labs), so in that case the MEG sensors are indirectly also expressed relative to the anatomical fiducials (as they are the same). The head surface is also expressed relative to the fiducials. 

The Yokogawa software appears to be even less integrated. I think that in your case the MEGCoordinateSystem (which describes the sensor positions in the .con file) should be specified as Other and in MEGCoordinateSystemDescription you should specify that coordinates are expressed relative to the “dewar" (or relative to the “helmet” in case that aligns better with Yokogawa user manuals). If you want, you can copy from the official Yokogawa documentation, or you can copy to the 2nd hand information from this documentation and/or link to it. I don’t know enought about the marker-coregis.txt and whether it contains the  anatomical fiducials, and/or the head localizer coils, and/of the scalp surface, nor in which coordinates those are, but those are documented with the AnatomicalLandmarkCoordinateSystem, HeadCoilCoordinateSystem  and DigitizedHeadPointsCoordinateSystem, respectively. It might be that all three recordings are in there, and it might be that all positions in that text file are expressed in the Polhemus electromagnetic digitizer coordinate system. It might also be that the software used for recording these points on the head already expresses relative to the AnatomicalLandmark points: in that case I would expect all of them to be expressed in the Yokogawa ALS headcoordinate system (as per BIDS coordinate system appendix).

If you want, we could look at this in more detail, I could review the dataset that you are now curating and organizing as BIDS, and we could document it as a Yokogawa tutorial on https://bids-standard.github.io/bids-starter-kit/. In case you are interested, I suggest we continue the in-detail discussion on https://neurostars.org.

best regards

Robert

 

--
We are all colleagues working together to shape brain imaging for tomorrow, please be respectful, gracious, and patient with your fellow group members.
---
You received this message because you are subscribed to the Google Groups "bids-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bids-discussi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bids-discussion/4977df02-cadf-41a9-99b6-95f06e68bd20n%40googlegroups.com.

[Marc Lalancette]

Hello,

If what Robert suggests, using the Yokogawa ALS headcoordinate system, is not possible for the raw data, then the logical solution would be to allow per-run files:

sub-<label>[_ses-<label>]_task-<label>[_acq-<label>][_run-<index>][_proc-<label>] _coordsystem.json

Otherwise, solution 1 distorts the intended meaning of a session.  Solution 2 would require something like NAS_task-<label>[_acq-<label>][_run-<index>][_proc-<label>]  to uniquely match recordings.  For the current spec, I'd still recommend the above, and either add  _coordsystem.json to bidsignore, or put the per-run files in a sourcedata folder (or other folder next to the BIDS root).

Cheers,

Marc

[Ryohei Fukuma]

Dear Robert,

Thank you very much for the positive reply.

The marker-coregis.txt you mentioned does not seem to be a file that is recorded by the Yokogawa system.
Manual of the Yokogawa system says that marker locations are saved in a marker file (*.mrk).
I've neither seen marker-coregis.txt in the MEG recording directories, nor found it in the manual.

Let me first explain the analysis pipeline of the Yokogawa system (performed in "MEG Laboratory" software) shown in the manual.
1. DICOM files are loaded into the MEG Laboratory (the format of the MRI is converted in .mri file which I do not know if it's in a standard format or not).
2. By selecting the anatomical fiducials (where MEG coils were placed during MEG recording) manually, co-registration is performed. If user has 3D digitizer in this process, you can refine the co-registration manually (the manual says that optional software can make it automatic, but we don't have it).
3. Current estimation using spherical model or dipoles are performed.

I agree there is "Yokogawa coodinate" in the pipe line; but, problem is that the MEG Laboratory do not have an ability to export the original data with sensor locations in head-based coordinate.
Technically, I already acquired AnatomicalLandmarkCoordinate, HeadCoilCoordinate and DigitizedHeadPointsCoordinate in a same coordinate system (head-based, Dewar-based, and MRI-based systems).
But, the problem makes it impossible to strictly follow the current specification of the BIDS format because of the following reasons.
1. The specification requires to upload MEG file in original format. However, with the limitation of the Yokogawa system, the files are in the Dewar-based coordinate system.
2. The specification allows only one *_coordsystem.json file in a session.
2-1. If I prepare AnatomicalLandmarkCoordinate, HeadCoilCoordinate and DigitizedHeadPointsCoordinateSystem in the Dewar-based coordinate, they differs between different runs making it difficult to write them in a one *_coordsystem.json file.
2-2. If I prepare these coordinates in any different coordinate system, sensor coordinates (for each run) in the same coordinate system must be included somewhere, which is not also defined in the current specification.

If the BIDS format can be extended to fit for the Yokogawa system, I'm happy to prepare tutorial for Yokogawa system as I have an access to the manual.
If you don't mind, I will prepare draft of the tutorial and upload it to https://neurostars.org.


Yours Sincerely,

Ryohei Fukuma

2022年2月11日金曜日 3:45:38 UTC+9 Marc Lalancette:

[Ryohei Fukuma]

Dear Marc,

Thank you very much for your reply.
I agree that your suggestion to allow the per-run files is logical.
I think that is natural and the best extension for the format.

I'm going to prepare my dataset with the per-run *_coordsystem.json file.


Yours Sincerely,

Ryohei Fukuma

2022年2月18日金曜日 11:23:40 UTC+9 Ryohei Fukuma: