Migrating from other analysis software

Here we offer some tips on how to migrate from other analysis software.

EEGLAB

To read in data exported from EEGLAB, MNE-Python includes an .edf reader mne.io.read_raw_edf() and a set file reader. To read in set files containing raw data, use mne.io.read_raw_eeglab() and to read in set files containing epochs data, use mne.read_epochs_eeglab().

This table summarizes equivalent EEGLAB and MNE-Python code for some of the most common analysis tasks. For the sake of clarity, the table below assumes the following variables exist: the file name fname, time interval of the epochs tmin and tmax, and the experimental conditions cond1 and cond2. The variables l_freq and h_freq are the frequencies (in Hz) below which and above which to filter out data.

Processing step	EEGLAB function	MNE-Python
Get started	addpath(...); eeglab;	import mne from mne import io, Epochs from mne.preprocessing import ICA
Import data	EEG = pop_fileio(fname);	raw = io.read_raw_fif(fname) raw = io.read_raw_edf(fname) raw = io.read_raw_eeglab(fname)
Filter data	EEG = pop_eegfiltnew(EEG, l_freq, h_freq);	raw.filter(l_freq, h_freq)
Run ICA	EEG = pop_runica(EEG, 'pca', n);	ica = ICA(max_pca_components=n) ica.fit(raw)
Epoch data	event_id = {'cond1', 'cond2'}; Epochs = pop_epochs(EEG, event_id, [tmin, tmax]);	events = mne.find_events(raw) event_id = dict(cond1=32, cond2=64) epochs = Epochs(raw, events, event_id, tmin, tmax)
Selecting epochs	Epochs = pop_epochs(EEG_epochs, {cond2});	epochs[cond2]
ERP butterfly plot	pop_timtopo(EEG_epochs, ...);	evoked = epochs[cond2].average() evoked.plot() evoked.plot_joint()
Contrast ERPs	pop_compareerps(EEG_epochs1, EEG_epochs2);	mne.combine_evoked([evoked1, -evoked2], weights='equal').plot() mne.viz.plot_compare_evokeds([evoked1, evoked2])
Save data	EEG = pop_saveset(EEG, fname);	raw.save(fname) epochs.save(fname) evoked.save(fname)

Potential pitfalls

• Many of the MNE-Python objects have methods that operate in-place to save memory (i.e., the data in the Raw object is changed when you call raw.filter(lfreq, hfreq)). If you do not want this, it is always possible to first call the object’s copy() method (e.g., filtered_raw = raw.copy().filter(lfreq, hfreq)). In addition, some MNE-Python functions have a boolean copy parameter that achieves the same purpose.

• The concept of channel types is critical in MNE because it supports analysis of multimodal data (e.g., EEG, MEG, EOG, Stim channel, etc) whereas most EEGLAB functions assume all channels are of the same type (EEG). To restrict channels to a single type, see mne.pick_types(), raw.pick_types, epochs.pick_types, evoked.pick_types, etc.