I am currently starting to use a KIT system, in which there's a
reference sensor a bit further away from the head than the other
sensors, and which is used to specifically record environmental noise.
I was wondering whether some of you had tried this method, and
compared it to other denoising approaches such as SSS? If so, is there
a python implementation?
we have an SSS implementation in Python, but it's still under validation,
right Eric?
How is the data quality on your KIT system? Do you see an urgent need to
decompose your signals / regress out extrnal influences? AFAIK on a 4D
system the reference sesnors are already used during acquisition for online
compensation.
I have an open issue <https://github.com/mne-tools/mne-python/issues/2112>
on github along these lines. I know that there was a MEG denoise repo
<https://github.com/pealco/python-meg-denoise> that has the tsPCA approach
along with DSS and SNS. It was done by a graduate student at UMD.
For the KIT system, the compensation isn't done online. It seems like the
exact use case you are mentioning, but I haven't had the time to evaluate
where it was properly implemented. I would be interested in helping where I
can but I wouldn't be able to lead that integration.
I would be interested in the comparison of an SSS approach and the tsPCA
approach.
I would be interested in the comparison of an SSS approach and the tsPCA
approach.
We haven't determined exactly how Maxwell filtering (SSS) should deal with
compensation at this point. But the algorithm is just about good to go for
at least Neuromag data, and could probably be tested on uncompensated CTF
data.