spikes in MUSE data when preprocessed in EEGLAB

[daga3467]

I am getting very 'spiky' output spectra plots in EEGLAB when attempting to analyse MUSE data in EEGLAB (see plot attached)

download/file.php?mode=view&id=148

This is after going through a preprocessing pipeline in EEGLAB, filtering, removing artefacts visually etc. and these data look 'ok', not wildly erratic.

Some data did not appear to do this initially, and a quick look analysis wise yielded the following (only 2 subjects' data albeit, but much smoother plots):

download/file.php?mode=view&id=149

but all subsequent data display this characteristic. Consequently (I assume there is a relation but correct me if am wrong) I get a 'subscripted sasignment dimension mismatch' error when I try to plot/analyse multiple data sets (in EEGLAB).

download/file.php?mode=view&id=147

Am no expert as yet on EEGLAB (let alone MATLAB) so does anyone have any advice on why my data show these characteristics or the error prevents further analysis?

many thanks,

Dave

[James]

Have you made sure you're using 256Hz RAW EEG data for the input?
Mind Monitor's default is to record at 1Hz. If you're doing your own FFT's then you need to go to "Recording Interval" in Settings and set it to "Constant".

[daga3467]

yes, recording interval on constant then just importing csv files straight into EEGLAB and following the processing pipeline there...

[James]

Did you filter out the blank lines with the blink/markers from the CSV?

[daga3467]

not sure what you mean. I imported csv direct into EEGLAB, removed epoch baseline, filtered at 1hz and then 40hz then plotted the channel data and visually inspected/removed artefacts (two passes through the data to reject artefacts) that appeared to be blinks etc) - this on the advice of a colleague of mine who specialises in electroencephalography

[James]

As well as EEG data, the CSV also contains data markers for Blink, Jaw Clench and optional user data markers.
They look like this in the file:

2021-02-11 11:02:07.605,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,/muse/elements/blink

[daga3467]

ok I found this thanks - there are minimal so I would not have noticed many blank lines in the dataset (approximately 8 per 46000 rows of data). Will filter these out in each file BEFORE importing into EEGLAB and see if this resolves the spiky issue (will report back)!

[James]

For reference, the MATLAB import script on the website ( https://mind-monitor.com/FAQ.php#matlab ) automatically splits the data into two separate tables, one for the main data called "museData" and one for the elements such as blinks, jaw clenches and other markers called "museElements".

[daga3467]

Thanks. One other related question: to what degree are the csv data already artefact filtered for the non-raw data (columns B to U) and in this output format would it be meaningful to analyse these data and establish means (eg. mean alpha for TP9 for one condition versus another experimental condition) - would this more or less correlate with EEGLAB preprocessed data then subject to analysis?

[James]

The RAW EEG in the CSV is unfiltered.
The brainwaves have blink artefact removal and power line noise removal that is done in Interaxon's closed source code, so I can't tell you exactly what's been done, only that it is.