Filtering Mind Monitor EEG Data

Mind Monitor is the main data collection application that pairs with the Muse EEG from Intraxon. I collected a 5-minute resting-state EEG, and I’m trying to analyze it in EEGLAB.

For the data I’m trying to analyze, I am low-pass filtering the data at 30Hz and high-pass filtering it at 1Hz. This was chosen just because there was one study conducted using the Muse EEGs by Cannard et al. (2021) which used this filtering technique! But beyond that:

• the high-pass filter removes activity below 1Hz. This is because there is little cerebral activity below 0.5 Hz — most activity below this frequency typically is due to motion or an electrical artifact! (Libenson. 2022)
• the low-pass filter removes activity above 30 Hz. Similarly, this is because cerebral activity measured on the scalp rarely goes above this threshold. Instead, these signals more likely are due to electrical noise or muscle artifacts.

As always, these are the standard, not the rule. As an interesting side-note, there are some avenues of research which explicitly look at activity outside these bounds! (Check out “DC Potentials”!) But this is outside the scope of my current research.

Once I’ve set the filter, I just leave the “plot frequency response” box checked and move on. EEGLAB then prompts me to rename and possibly resave this filtered version of the file. Very cool stuff. I personally think it’s always good to save any changes as a new file — you never know when your computer decides that electricity is a social construct and restarts. I’m not speaking from experience.

But OOPS — be careful to only apply one filter at a time! The EEGLAB tutorial says this:

“Note that if high-pass and low-pass cutoff frequencies are BOTH selected, low-pass and high-pass parts will have the same slopes. Frequently, the low-pass slope is therefore steeper than necessary. To avoid this problem, we recommend first applying the low-pass filter and then, in a second call, the high-pass filter (or vice versa).” (https://eeglab.org/tutorials/05_Preprocess/Filtering.html)

…That’s good to know. Let me just redo all my filtering. Don’t worry, it’s not difficult.

But no worries. Then, you can just plot it (Plot → Channel spectra and maps) and… BOOM:

Brain waves! Real, filtered, brain waves! Isn’t that incredible?

But here’s the kicker: not all the data is clean. The messy data needs to be rejected so it doesn’t screw with the data analysis.

That definitely feels like a task that should be done after a night’s sleep, and not at 12:30 AM. Goodnight then. My sleep spindles will be doing their thing, soon enough.