Unknown pattern in MEG data

Support & Discussions
1

  • MNE version: 1.7.0

  • Operating system: Ubuntu

  • MEG system: Vectorview, Elekta-Neuromag (306 channels).

    Dear forum members,

    We are observing a high-amplitude pattern in our MEG data across multiple subjects (pediatric ASD cohort).

    Signal morphology:

    1. High-amplitude spindle-like bursts.

    2. Basic frequency in the beta band (~15-25 Hz).

    3. The spatial topography is consistent across subjects: ICA component localizing strictly at central sensors.

    Processing pipeline: Raw data → MaxFilter (tsss + mc + trans) → High-pass (1 Hz) + notch (50 Hz) + resample (500 Hz) → FastICA.

    We hypothesized this might be Rolandic beta rhythm, but also it could be hardware artifact (or even something else). I have attached three examples for 3 patients (ICA topoplots, ICA time series, raw data) on disk.

    ICA components with artifact numbers:

  • ICA058 for AF101

  • ICA048 for AF102

  • ICA053 for AF103

    Link to Google Drive with screenshots of ICA and raw data.

    Has anyone encountered similar artifact on Neuromag systems?

2

Hey,

Could you also share frequency spectrum of your “abnormal” ICAs please? If you use EEG data, I would suspect if there is electrode contact problem at some electrodes. However, this is MEG data and I don’t want speculate anything right now. Can you give much more details ?

What kind of task subjects perform during the data collection?
Did you perform ICAs on the continuous or epoched data?

3

Peak in 25-27 Hz was observed in all components with artifact. High amplitude in beta wasn’t observed in other (clean) components. I will upload the ICA PSD plots as soon as I regain access to the remote desktop.

The data in this specific recording comes from a passive viewing task (the subject was watching a video without any tasks). Data from experiments with other tasks are currently pending analysis.

The ICA on screenshot was applied to continuous raw data (not epochs). All our actions below in processing pipeline.

Processing pipeline: Raw data → MaxFilter (tsss + mc + trans) → High-pass (1 Hz) + notch (50 Hz) + resample (500 Hz) → FastICA.