We’ve been experiencing intermittent issues with continuous head localization (CHL) for a long time, but I have been experiencing it a lot more lately, especially when using simultaneous EEG and slightly non-standard head coil placement (on the EEG caps).
When this occurs, pre-localization, i.e. before starting the recording, works fine, but as soon as the preview screen comes up after accepting the initial position, CHL indicates large fit errors (up to 10-20%) and large distance from the initial position. Sometimes, restarting the collection will fix this, or asking the participant to change their head position slightly might help as well. But it may not go away even after restarting 2-3 times, and typically lasts for the entire run, or even multiple runs if the participant doesn’t move.
If you record while this issue is happening, CHL can not be trusted. The fit error is indicative of this, but it is highly recommended to write it down in the session log as well. For such a dataset, only the initial position should be used during coregistration and analysis. Usually only one coil will be affected (often nasion, but not always), so the other two could be used to estimate movement during the recording.
I’ve been troubleshooting this with CTF for some time, and it seems to be a software issue. I’ll keep investigating and press them for a fix, but CTF software development is beyond slow unfortunately.