Back to the main page.
Bug 3056 - lcmv beamformer source solution mismatch topography
| Status | CLOSED DUPLICATE |
| Reported | 2016-01-31 17:53:00 +0100 |
| Modified | 2019-08-10 12:33:17 +0200 |
| Product: | FieldTrip |
| Component: | inverse |
| Version: | unspecified |
| Hardware: | PC |
| Operating System: | Mac OS |
| Importance: | P5 normal |
| Assigned to: | |
| URL: | |
| Tags: | |
| Depends on: | |
| Blocks: | |
| See also: |
Philipp Ruhnau - 2016-01-31 17:53:07 +0100
Created attachment 770 script and example image Dear ft-developers, I've noticed something strange and I cannot figure out what the issue is (see example screen shots and analysis pipeline aattached). I'm not entirely sure this is a bug, but given the recent problem with localspheres maybe this is related? as can be seen in my screenshot in the attachment, when I beam an erf using the lcmv method with a common filter I get very odd source solutions. I noticed this on the group level (where actually my effect was inverted in polary, and sort of medial). the paradigm is super simple (hit vs. miss of a threshold visual stimulus on the right screen side) with hundreds of trials, thus there should be a quite clear erf for hits, left lateralized. but as i said on group level i get hit < miss and more something medial or, as for the example subject frontal activity. I tried numerous things to name those I still remember: leadfield normalisation, regfactor (0/5/10), creating an average from the mom field (which gives interestingly enough a quite different solution but still an odd one), different grids, fixedori, separate filters, longer covariance time window, solutions based on gradiometers/magnetometers/both, i even calculated the covariance by hand, but still all the same. I think that, if there is a problem and I didn't just kept a typo, then that the problem is somewhere in parts of the code that I do not have the mathematical knowledge to understand.... only observation is that it seems the source reconstructions of both conditions are dominated by the filter. the covariance/data don't seem to do much. I put the data and vol/mri in my dropbox: https://www.dropbox.com/sh/9qi0u1xzbmg6as5/AAAkqOxg2odu7-3wMEKHDs1ua?dl=0 sorry for these large data, but I haven't figured out how to simulate good ERF data for gradiometers.... I was considering writing this to the list first, and apologize if it is a mistake on my side. Best Philipp p.s.: the data in the dropbox contain gradiometers only, to reduce space
Philipp Ruhnau - 2016-01-31 18:00:59 +0100
sorry, submitted twice, and cannot figure out how to delete *** This bug has been marked as a duplicate of bug 3057 ***