EEG preprocessing

Hi Mike,
I am currently working on preprocessing of EEG signals which recorded from newborn children in a resting-state while they were sleeping. I tried to follow the preprocessing methods in chapter 7 and 8 in your book. I did them in the following order: Common average reference (CAR), second-order butterworth bandpass filters, epoching with 2-second length and normalization. After that, I am planning to do epoch removal and bad channel rejection.
1- Do you think these methods are suitable for our case?
2- Do you think I also need to apply the artefacts removal methods even if the data recording while the infant were sleeping? and if so, can you please suggest any automatic removal methods?
Also, I am facing some difficulties in judging the quality of the signal. So, can you please help me in this matter?
In the following, I upload each snapshot before and after the preprocessing method.

Do you think the signal looking good after each step?

Hi. The pipeline you describe sounds good. But there are some residual artifacts in the signal you’ve pasted. I can’t quite see the time scale, but my guess is that the lower plots have 50/60 Hz line noise. A notch filter or low-pass filter should get rid of that. Unless you are looking for ~50 Hz brain activity.

The first panel doesn’t really look like line noise, though. I suggest running ICA to see if that will isolate some of the artifacts.

btw, you should do visual data inspection before re-referencing the signal. If you have a huge artifact in one channel, computing the average reference will transfer that isolated artifact into every other channel.