[FieldTrip] bug in recent versions of ft_prepare_leadfield

[Robert Oostenveld]

Dear all

We are sorry to report that with merging the recent pull request #1377 <https://github.com/fieldtrip/fieldtrip/pull/1377> about a month ago, we introduced a bug in the FieldTrip release that might have affected your computations.

Specifically, the handling of the defaults for cfg.reducerank and the cfg.backproject changed a bit, which had an unforeseen sideeffect in ft_prepare_leadfield, if (and only if) you used this function for the computation of MEG-based leadfields, using the singleshell method. Specifically, ft_prepare_leadfield would by default compute MEG singleshell leadfields with reducerank=2 (this is how it always has been), but it would wrongly backproject (i.e. the intended projection of the rank-reduced 2-column leadfield back into 3D space with a column for the x, y, and z dipole moment in Cartesian space) the rank-reduced leadfield. Rather, ft_prepare_leadfield would on line 292 <https://github.com/fieldtrip/fieldtrip/blob/af5f9822413d11e66f3821943e945e98ab766da6/ft_prepare_leadfield.m#L292> discard the last column. The error was due to MATLAB interpreting if [] as false, causing on line 295 only the first two columns to be copied to the output. Consequently, the leadfield for the z-direction was discarded.

Are you affected? If you used the master branch from GitHub (which is the development version), or a release version between 20200529 and 20200607, and if you have used ft_prepare_leadfield for MEG with singleshell models, then you are likely affected. You can check yourself: your precomputed leadfields should have three columns; if they only have two columns and you did not explicityly specify reducerank and backproject in your configuration, your leadfields are wrong. Other models than MEG singleshell are not affected.

To resolve the problem, please update to the latest 20200623 release version from the FTP server <ftp://ftp.fieldtriptoolbox.org/pub/fieldtrip/> or from the GitHub release page <https://github.com/fieldtrip/fieldtrip/releases> and recompute your leadfields, source estimates, etcetera.

best regards,
Robert

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20200624/4de33da1/attachment.htm>