Work with the annotations from polysomnographic data

💬 Support & Discussions
1

Dear Mne team,

I work with polysomnographic data marked with different annotations (REM sleep, EOG REM, EMG phasic, EMG Tonic, etc).

I extracted the period of interest with mne.events_from_annotations, but I cannot find a solution to define the period with the superposition of annotation. For example, I would like to compare REM sleep+EOG REM Vs EOG REM without REM sleep.

How can I extract this period of signal excluding or cumulating different annotations?

Best.

William

2

I would create my own annotations that has new annotation entries for overlapping periods.
Then put a description that is the concatenation of the 2 overlapping descriptions and then reuse
events_from_annotations.

Alex

3

Hello Alex,

Thank a lot for the answer.

Okay, no possibility of implementing a conditional selection of epoch based on annotation.

I should use mne. annotation on continuous data and manually selecting overlap?

Best.

William

4

what do you mean manually?

given an Annotations object called annot you have annot.onset, annot.duration and annot.description
that are 3 numpy arrays. You can update / manipulate them.

then you can recreate a new annotation with mne.Annotations(new_onset, new_duration, new_description)
then raw.set_annotations(new_annotations)

clear?

Alex

5

Okay, with that, but in practise, I have one difficulty. For example on my continuous data I have these annotations:

raw_eeg.annotations.description
array([‘EMG : EMG médium’, ‘EOG : REM’, ‘EMG : EMG très bas’, ‘Vidéo’,
‘Note : début essais’, ‘EOG : REM’, ‘YF’, ‘EMG : EMG très bas’,…]

How can I select the description ‘EOG : REM’ and get the onset and duration for this specific description?

Thank.

William

6

I did something like that:

annotations = raw_eeg.annotations

annotation_Target = 'EOG : REM'

start_times = raw_eeg.annotations.onset[annotations.description == annotation_Target]
Duration_times = raw_eeg.annotations.duration[annotations.description == annotation_Target]
1 Like