Back to the main page.
Bug 3170 - output of ft_prepare_leadfield and thus ft_sourceanalysis depend on unit of the grid when using grads and mags
|2016-08-02 16:25:00 +0200
|2019-08-10 12:41:12 +0200
Thomas Hartmann - 2016-08-02 16:25:36 +0200Created attachment 807 script to reproduce the bug hi, the attached script does a lcmv source analysis using NEUROMAG data. both magentometers and gradiometers are used. the only difference between the two calculations is that one is done with a grid and leadfield in meters, the other in centimeters. when i do that with an old ft version, the output of the calculations (i.e., the lcmv power as well as the leadfields) differ but are a mere scaling of each other. i saw that there were previous reports of that topic. this is only confusing but something we could live with. BUT in the latest fieldtrip versions, the tra field is set for the gradiometers. when the sensors are converted to a different unit, only the tra entries for the gradiometers get scaled accordingly while the entries for the magnetometers are still 1. this leads to entirely different leadfields and thus entirely different results after source analysis. the attached script should run with any fif file. if you need an example fif file, please let me know and i will provide you with a link. best, thomas
Jan-Mathijs Schoffelen - 2017-01-24 12:41:46 +0100The metric units influence the outcome, because of the gradiometer units depending on distance. I guess that this has been discussed at Biomag (I recall a somewhat heated discussion at least). Can we close this, or does it need follow up?
Thomas Hartmann - 2017-01-24 12:53:48 +0100(In reply to Jan-Mathijs Schoffelen from comment #1) yes, it is what we discussed at the biomag and i took the advice by robert and can confirm that he is right. this bug can thus be closed.
Jan-Mathijs Schoffelen - 2017-01-24 12:55:38 +0100OK, will do. Just to make a point: the advice of Robert was the same as the stuff I wrote in a personal e-mail :o).
Robert Oostenveld - 2019-08-10 12:35:01 +0200This closes a whole series of bugs that have been resolved (either FIXED/WONTFIX/INVALID) for quite some time. If you disagree, please file a new issue on https://github.com/fieldtrip/fieldtrip/issues.