Back to the main page.
Bug 3445 - magnes 3600 issue with reference coil locations in file
| Status | NEW |
| Reported | 2018-09-01 11:15:00 +0200 |
| Modified | 2019-06-21 11:30:29 +0200 |
| Product: | FieldTrip |
| Component: | fileio |
| Version: | unspecified |
| Hardware: | PC |
| Operating System: | Mac OS |
| Importance: | P5 normal |
| Assigned to: | |
| URL: | |
| Tags: | |
| Depends on: | |
| Blocks: | |
| See also: |
Jan-Mathijs Schoffelen - 2018-09-01 11:15:53 +0200
As discussed with Jean-Michel at biomag: there's an issue with the representation of the positions of the reference coils in 4d datasets, where (at least with the linux version of the acquisition software) the positions of the reference coils is wrong. This causes potential problems with the balancing of the forward model with the digital weights. Anecdotally, in HCP we have had quite some discussions about the poorer quality results when using the balancing weights in the forward model, as compared to not using them, so this might be related. In any case, JMB has communicated with the people from 4D and he has obtained a set of 'factory coordinates', which should be more or less the positions of the actual coils (give or take the slight changes with respect to the calibrated positions). We discussed the following plan of action: 1) investigate what the effect is of a mismatch between actual ref-coil position, in relation to the balancing weights, when accounting for the weights in the forward model (w.r.t. no balancing, and w.r.t. the correct positions with balancing). This will be done using simulations, for which JMS will work with JMB to create a set of gradiometer descriptions that are needed. 2) investigate whether other systems (e.g.: Glasgow, St.Louis (HCP), Tel Aviv) have the same issue. 3) create a patch in the FT-reading functions to optionally check and fix the coil positions of the reference coils.
Jan-Mathijs Schoffelen - 2018-10-02 14:10:51 +0200
Created attachment 872 visualization of the grad array from a Glasgow dataset (front view)
Jan-Mathijs Schoffelen - 2018-10-02 14:11:34 +0200
@jean-michel: does the z-coordinate of the reference coils in the attached figure make sense?
Jean-Michel Badier - 2018-10-19 17:47:45 +0200
Created attachment 875 Marseille sensor (recent callibration with a priori no bug)
Jean-Michel Badier - 2018-10-19 17:48:48 +0200
Hello, Sorry for my late answer. I missed the comment and saw it when I went back to the point. The comment of JMS is correct. There is a bug in the 4D software that tends to change the distance of the reference sensors from the other sensor after each callibration. The callibration is suppose to give the real location and orientation of the sensor coils from measurments. So these values change a little from the original factory values. As a result of a bug, the reference coils also change as they should not in the procedure. That results normaly in an incresased distance between the set of references and the measurment sensors. That is the case I think for the Glasgow sensor (from the picture). The distance between the sensor A1 to the central reference channel should be less that 15cm. We had in the past more than 27 cm (thanks again to François that mentionned that). I do have coordinates that are the one from our sensor. The coordinates for the references should be the factory ones. For some strange reason one gradiometer locations seems wrong in my data files. I check it and send you the coordinates asap. 1. OK 2. I will also ask the Erlangen facility if they have the same problem. 3. I guess the factory locations for the references could be used if there is any noticable effect. I suppose the geometry is similar for all centers (at least with the 3600WH)
Jean-Michel Badier - 2018-10-19 18:02:16 +0200
Hello, Sorry for my late answer. I missed the comment and saw it when I went back to the point. The comment of JMS is correct. There is a bug in the 4D software that tends to change the distance of the reference sensors from the other sensor after each callibration. The callibration is suppose to give the real location and orientation of the sensor coils from measurments. So these values change a little from the original factory values. As a result of a bug, the reference coils also change as they should not in the procedure. That results normaly in an incresased distance between the set of references and the measurment sensors. That is the case I think for the Glasgow sensor (from the picture). The distance between the sensor A1 to the central reference channel should be less that 15cm. We had in the past more than 27 cm (thanks again to François that mentionned that). I do have coordinates that are the one from our sensor. The coordinates for the references should be the factory ones. For some strange reason one gradiometer locations seems wrong in my data files. I check it and send you the coordinates asap. 1. OK 2. I will also ask the Erlangen facility if they have the same problem. 3. I guess the factory locations for the references could be used if there is any noticable effect. I suppose the geometry is similar for all centers (at least with the 3600WH)
François Tadel - 2018-10-20 09:16:38 +0200
Created attachment 876 4D coils
François Tadel - 2018-10-20 09:17:00 +0200
Created attachment 877 CTF coils
François Tadel - 2018-10-20 09:17:22 +0200
Created attachment 878 Neuromag coils
François Tadel - 2018-10-20 09:17:44 +0200
Good that this issue gets finally addressed! FYI: Brainstorm offers nice tools for representing the orientation and direction of the various coils: right-click on the channel file > Display sensors > "4D coils (all)".
Jan-Mathijs Schoffelen - 2019-04-01 09:08:03 +0200
Hi J-M/Francois, How to proceed here?
Jan-Mathijs Schoffelen - 2019-06-21 11:21:52 +0200
Just a friendly reminder that this has not yet been resolved. What would be the proper way forward here?