[Eeglablist] Epochs of different length & ICA

[#HENG JIAMIN GLADYS#]

Dear EEGlab list,

I am working on a task where participants are supposed to listen 8 music clips, ranging from 20 seconds to 41 seconds.
Participants are just supposed to close their eyes and listen to the music while it plays. Thereafter, they had to complete some rating questions (i.e. there will be some movement).
The task is recorded as a single file.

I understand that we should epoch the data before ICA so that non task-related activity do not make the data "noiser" for ICA.
Thus far, I have epoched the continuous data into 8 different files with the respective duration.

However, I am just contemplating when I should append/concatenate the epoched data with regards to ICA:

Option A: Split the single file into the 8 epochs with their respective duration > Run ICA > Reject noisy ICs accordingly > Append epochs back together
Option B: Split the single file into the 8 epochs with their respective duration > Append all epochs back together > Run ICA > Reject noisy ICs accordingly.

The issue with Option A is that each epoch might have different remaining ICs after noisy IC rejection, and when I append the epochs back together at the end, they won't have equal number of ICs. Other than that, I'm unfortunately not very familiar with what are the mathematical implications of appending/concatenating the epoched-data before ICA (e.g. it probably introduces some boundary/edge artifact before ICA, which might not be ideal?)

Would appreciate any advice and thank you very much in advance.

Regards,
Gladys