Annotations to dataframe

đź’¬ Support & Discussions
1

Hello guys, I’m annotating my data like the code shows:

my_annot = mne.Annotations(onset=2670,  # in seconds
                           duration=170,  # in seconds, too
                           description='seizure')
print(my_annot)
raw.set_annotations(my_annot)

After a series os preprocessing I transform the data to a pandas dataframe, however the column condition only shows the number one. The annotations indicate seizure and I need them to be correctly labeled for classification purposes.

raw.set_eeg_reference()

raw.filter(1., None)

raw.apply_function (lambda x: x * 1e-6)

raw.pick_channels(['F3-C3','C3-P3', 'F4-C4','C4-P4' ])

raw.filter(l_freq=0.5,h_freq=40.0, method='iir', iir_params=iir_params, verbose=True)

epochs=mne.make_fixed_length_epochs(raw, duration=4, overlap=0.5)

epochs=epochs.to_data_frame()

image
image797Ă—481 16.1 KB

2

you don’t pass any event_id so your condition is the default so 1.

did you mean to use events_from_annotations to cut epochs in seizure intervals?

Alex

3

So if I pass the event_id only the epochs “inside” the seizure will be marked as 1?
I’m using this calculation to know the epochs that seizures are so I manually mark in the column.

epochs['condition'] =0

start_time = 2670  # example seizure start time in seconds
duration = 171  # example seizure duration in seconds
sampling_rate = 256  # EEG data sampling rate in Hz
epoch_len = 4  # epoch length in seconds
overlap = 0.5  # epoch overlap in seconds
epoch_len_samples = epoch_len * sampling_rate  # epoch length in data points
overlap_samples = overlap * sampling_rate  # epoch overlap in data points
start_sample = int((start_time - overlap) * sampling_rate)  # calculate starting sample index
end_sample = int((start_time + duration - overlap) * sampling_rate)  # calculate ending sample index
start_epoch = int(start_sample // epoch_len_samples)  # calculate starting epoch index
end_epoch = int(end_sample // epoch_len_samples)  # calculate ending epoch index
seizure_epochs = list(range(start_epoch, end_epoch + 1))  # create list of epochs that correspond to seizure

seizure_epochs 
[667,
 668,
 669,
 670,
 671,
 672,
 673,
 674,
 675,
 676,
 677,
 678,
 679,
 680,
 681,
 682,
 683,
 684,
 685,
 686,
 687,
 688,
 689,
 690,
 691,
 692,
 693,
 694,
 695,
 696,
 697,
 698,
 699,
 700,
 701,
 702,
 703,
 704,
 705,
 706,
 707,
 708,
 709,
 710]

epochs.loc[(epochs['epoch']>=667) & (epochs['epoch']<=710), 'condition']=1

epochs['condition'].value_counts()
4

Here’s an example that re-codes the equally spaced event codes based on whether they’re not in the seizure annotation, partially overlapping the seizure, or fully within the seizure:

import numpy as np
import mne

sample_data_folder = mne.datasets.sample.data_path()
sample_data_raw_file = (sample_data_folder / 'MEG' / 'sample' /
                        'sample_audvis_raw.fif')
raw = mne.io.read_raw_fif(sample_data_raw_file, verbose=False, preload=True)
annot = mne.Annotations(onset=[3, 55], duration=[15, 27],
                        description=['seizure'] * 2)
raw.set_annotations(annot)
annot_events, annot_event_id = mne.events_from_annotations(raw)
# get onset and duration in samples
onset_samp = annot_events[:, 0]
duration_samp = np.rint(annot.duration * raw.info['sfreq']).astype(int)
offset_samp = onset_samp + duration_samp

# make equally spaced events
epoch_dur = 4
events = mne.make_fixed_length_events(raw, duration=epoch_dur, overlap=0.5, first_samp=False)
event_samp = events[:, [0]]
epoch_duration_samp = epoch_dur * raw.info['sfreq']

# compute overlap / containment
overlap_with_seizure = np.logical_and(
    (event_samp + epoch_duration_samp) >= onset_samp,
    event_samp <= offset_samp
).any(axis=1)

fully_within_seizure = np.logical_and(
    event_samp >= onset_samp,
    (event_samp + epoch_duration_samp) <= offset_samp
).any(axis=1)

events[overlap_with_seizure, 2] = 2
events[fully_within_seizure, 2] = 3

Now you can use that events to do your epoching. Condition 1 is “not in seizure”, 2 is partial overlap, and 3 is fully within seizure. Of course you can skip the partial overlap step if you don’t care about those, or skip the “fully within” step if you want to treat partial and full overlap the same.