find_eog_events improvement

✉️ Mailing List Archive (read-only)
1

Hi all,

I'm using find_eog_events to determine whether a patient is asleep or awake based on the frequency/number of blinks detected. However, when I run find_eog_events on six hours of cEEG, only 2 eog events are detected (there should be many more). Any suggestions on how to solve this problem?

Thanks,
Jake

EEG_channels = ['F3', 'F4', 'C3', 'C4', 'O1', 'O2']
EOG_channels = ['LOG', 'ROG']
ch_types = ['eeg', 'eeg', 'eeg', 'eeg', 'eeg', 'eeg', 'eog', 'eog']
eog_event_id=998

info = mne.create_info(ch_names=EEG_channels+EOG_channels,
                       ch_types=ch_types,
                       sfreq=Fs)
rawEEG = mne.io.RawArray(data, info)#raw.shape is [n_channels, n_data_pts]
raw.resample(sfreq=eog_resample_fs)#resample to 50 Hz for blink detection
eog_events = mne.preprocessing.find_eog_events(raw, eog_event_id)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20190712/b16df96c/attachment.html

2

External Email - Use Caution

It's entirely possible that `find_eog_events` only works reasonably when
blinks are somewhat consistent through time. You could try adjusting the
"thresh" param, as the default might not be suitable for your data. If that
(and maybe also changing the filtering parameters) does not work, you could
write some dedicated routine for detecting blinks from the EOG channels
based on the particular statistics of your data.

Eric