Using ICALABEL to detect line noise

💬 Support & Discussions
1
  • MNE version:1.8.0
    Hello everyone
    I have a batch of data with line noise, and I want to use ICALABEL to help me identify eye movement and line noise components.
    According to ICALABLE’s suggestion, I trained ICA in epochs of 1-100Hz filtering and obtained some components that may be eye movement and line noise.
    Afterwards, I should remove these components on epochs of 0.1-30Hz.
    But I think this is a bit strange because the 30Hz low-pass filter has already removed high-frequency noise.
    Will removing high-frequency noise from the IC again lead to excessive removal?
    Or should I remove IC of noise on original raw data
    微信截图_20240928205101
    微信截图_202409282051011920×1020 301 KB

    without filter?
2

Hello, if your data has already been filtered such that line noise is not an issue anymore, why not simply ignore the line noise components that MNE-ICALabel reports, i.e., just don’t remove those. This would be the easiest approach in my opinion. MNE-ICALabel allows you to pick components that are for example related to blinks only.

Best wishes,
Richard

3

Hello richard,
Thank you for your help!
But

微信截图_20240929174236
微信截图_202409291742361390×550 77.4 KB

there are still line noise on 10-30hz (looks like)after 0.1-30hz filter , I don`t know what component it is.
Maybe it is not line noise, Have you ever meet this problem?
Thank you again!

4

Honestly I don’t see anything there :sweat_smile:

What you can try is remove the line noise components and check what the signal looks like after cleaning :slight_smile:

Edit: Oh you mean the generally higher power of those two channels? If you suspect there might be a problem, these channels would actually be candidates for exclusion (i.e., marking them bad) – before ICA!

1 Like
5

Thanks richard, Yes, higher power of those two channels.
I will follow your advice and repair them before ICA!
You have answered me more question than my supervisor who focus on social psychology, You are such a nice person. If you come to BeiJing ,I will have you a big meal!

1 Like
6

Line noise is a spike on the power spectrum at 50 Hz (EU) or 60 Hz (US) + harmonics, 50, 100, 150, … or 60, 120, 180, …
As suggested by @richard I reocmmend to mark those 2 channels as ‘bads’ and interpolate them later if you need them.

Mathieu

2 Likes
7

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.