Hello Tatu,
I can answer some steps as the MaxFilter developer, but you may need to get more input from experts of ICA afrer maxfiltering.
First, tSSS is safe to use even if you would not have any magnetized pieces attached. If there is any head movement, EEG electrodes may sometimes cause movement-related disturbances. Monitor the program output (option -v or the GUI log dialog), and see what it reports about the number of projected waveforms. If there are no sensor-space disturbances, the program simply reports 'projected 0 waveforms' and essentially performs ordinary SSS. Note that tSSS does not affect eye-movement signals because their sources are in the head and far from sensors.
The warning about "diminisihing brain signals arising from very strong, superficial sources" is no worry in normal research use. However, it is fairly easy to demonstrate e.g. with shallow phantom dipole and 1000 nAm amplitude that tSSS may reduce the phantom signal if the origin is moved e.g. 2 cm lower. In addition, in some extreme cases like really strong epileptic source (>1000 nAm) located at the most superficial parts of the brain far from the origin, and spiking at short time intervals, tSSS application may show amplitude reductions.
The tSSS a s such does not reduce signal rank, but the number of SSS components depends on the head position. Head position transformation result does also depend on the choice of the SSS origin. The default value, 0 0 40, is often fairly low in adult subjects, and transforming to the default position (i.e. device frame origin) may cause the position to shift over 25 mm, which in turn causes increased reconstruction noise in MaxFilter. Thus, you should try to find the origin according to subject anatomy:
1. If you have MRI data, fit a sphere to inner skull curvature and use the resulting origin.
2. Digitized points can be used, provided that there is a large enough number of them on the scalp. MaxFilter provides option '-origin fit', which uses all digitized points.
3. However, large number of points on the face (nose and eyebrows) usually pushes the result too low. If this is the case, I can send you a small utility program which leaves the facial points away from the origin fitting and produces better origin for maxfilter.
Last, have you checked the result after ICA if you don't to '-trans default'? If you compare the result in source space, the head position transformation step is unnecessary, but you would still profit to select the SSS origin close to the subject's anatomy.
We have a small utility tool to visualize the initial head position in the helmet, and I can send it if you are interested to try.
Best regards, Jukka
Jukka Nenonen | Method Development Manager
Elekta Oy
Box 34, FI-00531 Helsinki, Finland
Visiting Address: Siltasaarenkatu 18-20 A, Helsinki
Office: +358 9 756 24085 | Mobile: +358 400 249 557 | Fax: +358 9 756 24011
[log in to unmask] | http://www.elekta.com
________________________________________
From: Announcement for the Neuro MEG list [[log in to unmask]] On Behalf Of Tatu Huovilainen [[log in to unmask]]
Sent: 20 January 2016 23:27
To: [log in to unmask]
Subject: Maxfiltering parameters with ICA
Hi all,
I have quite a few questions about choosing the Maxfiltering parameters, with the aim of working with data in the independent component domain. I'm working with combined M/EEG data (with cHPI) that is full of eye-movements.
For the ICA I have to do movement correction to fix the source-sensor configuration and to transform the different trials to same head position. I'm transforming the location to default head position, with changing the default location on some subjects where Maxfilter reports too large corrections otherwise (the 25mm limit). About the 'adjust initial hpifit consistency' option: It seems to be off by default, but isn't it better to always assume the HPI coils could have moved? Doing this will not introduce unnecessary variation to the source-sensor stationarity?
Then about tSSS: I've read that tSSSing will result in different parts of the data having different rank, which could be a problem for ICA. Would I be better off with doing only SSS? As a note, I'm not looking to remove the eye movement generated signals at this point as the ICA method I've chosen works great with those. But given that I have the EEG also, will I have to use tSSS?
Also the "Warning: Spatiotemporal signal space separation (tSSS) may diminish brain signals arising from very strong, superficial sources" is worrying. What would be considered very strong source? Would this include the eye movement related signals?
Sorry if these are very basic questions, my background is in EEG and all these possible adjustments of Maxfiltering settings seem important considering ICA, but I can't seem to find any relevant discussions. I'm also asking these as I've now done ICA (with combined M/EEG, tSSS'd data, that's moved to default head position with adjusted hpifits) and the results were pretty much on par with doing everything with the EEG sensors only, which seems a let down. I'm also seeing a lot of eye movement components that have weird topographies when looking at the gradiometers, while with EEG only I'm getting much fewer and clearer eye movement ICs.
Best regards,
Tatu
P.S.
Maxfilter version is 2.2.15
Please consider the environment before printing this e-mail.
The contents of this e-mail message (including any attachments) are confidential to and are intended to be conveyed for the use of the recipient to whom it is addressed only. If you receive this transmission in error, please notify the sender of this immediately and delete the message from your system. Any distribution, reproduction or use of this message by someone other than recipient is not authorized and may be unlawful.
|