[Eeglablist] Heart Rate (ECG) removal

[Bruzadin Nunes, Ugo]

Hello eeglab community,

I have been having trouble removing ECG/EKG from my datasets. At the INL lab, we noticed that even after thorough automatic ICA/PCA/ICLabel/BSS/CORRMAP and epoch rejection loops, there is still a lot of ECG bleeding in many "good" components. We could do it by visual inspection, but even then a lot of our ECG have left-over pieces spread through many components, making it almost impossible to remove it.

We use a lot of IClabel (default or beta), but it has been missing a lot of ECG or misidentifying them as brain.
MARA also isn't performing very well on our dataset, it's removing a lot of untouchable components (decreasing alpha by significant amount of power).
We also noticed that Cleanline may have been introducing errors in our data so we decided to avoid that altogether.

I have been thinking about how to solve this problem and decided to ask for help.

I considered many alternatives and wanted to know what are my other options. I thought about:
A. creating a algorithmic-generated heart-rate depending on the persons' EEG, and using it to drive ICAs to a more robust heart-rate component.
B. getting template with a good component ECG and cleaning it up (filtering & BSS) then using it for a later CORRMAP.
C. identifying the components with left-over ECG, cluster them, them and running a BSS-type algorithm to separate the ECG from the data.

However, these are very tasking ideas. What are the other alternatives that I missed out?

Any help is welcome!

Thanks in advance,

Ugo Bruzadin Nunes, PhD Candidate
PSYC 222 Lecturer - Effects of Recreational Drugs
Integrative Neuroscience Lab - Member
Brain and Cognitive Sciences Ph.D Program ||
School of Psychological and Behavioral Sciences
Southern Illinois University - Carbondale