Hi Everyone,

When dealing with EEG, artifacts from muscle, heart, and eye movements are inevitable. These artifacts are typically removed using independent component analysis (ICA).

From my understanding, the process of labeling components as artifacts or non-artifacts is quite subjective. For instance, if you perform ICA decomposition in EEGLAB, you then decide if a component is an artifact based on its scalp topography, event-related potential (ERP) image, power spectrum, and possibly the recommended IC labels calculated using the 'ICLabel' plugin.

I wonder how much this decision impacts the accuracy of EEG classification, especially in cases where the differences between distinct groups are subtle, such as differentiating between various motor imagery tasks.

Thank you in advance for your insights,

Fatemeh