pick-channels

name	pick_channels
description	Select channels by name list or type (EEG, MEG, etc.)
layer	L3
group	channel
metadata	{"tags":["operator","channels","select","subset","pick"],"modalities":["eeg","seeg","ecog","meg"],"step_string":"pick_channels","analysis_goal_allowed":["classification","source_localization","feature_extraction","clinical_screening","exploratory","generic","connectivity","phase_amplitude_coupling","online_inference"],"analysis_goal_forbidden":[]}

Pick Channels

Function

Subsets the data to include only specified channels. Can select by explicit channel names or by channel type (EEG, MEG, sEEG, etc.).

Parameter Format

pick_channels:{selection}

Examples:

• pick_channels:eeg — Keep only EEG-type channels (remove EOG, EMG, stim, etc.)
• pick_channels:Fp1,Fp2,F3,F4,C3,C4,P3,P4 — Explicit channel names
• pick_channels:meg — Keep only MEG channels
• pick_channels:seeg — Keep only sEEG channels

Parameters

Parameter	Type	Description
selection	string	Channel type (eeg, meg, seeg, ecog, emg) OR comma-separated channel names

Type Selection

When a single recognized type is given, channels are filtered by their ch_types metadata:

• eeg — scalp EEG electrodes
• meg — MEG sensors
• seeg — sEEG depth contacts
• ecog — ECoG grid/strip contacts
• emg — EMG channels (usually removed)
• data — all data channels (exclude stim/misc)

When to Use

• Remove non-neural channels (stim, EOG, EMG, misc) before processing
• Select a spatial subset (e.g., only motor cortex channels C3, Cz, C4)
• Focus on specific channel group for paradigm-specific analysis
• Remove reference channels that shouldn't be processed

When NOT to Use

• When you need ALL channels for spatial methods (ICA, CAR, source localization)
• Keep EOG/ECG channels if ICA artifact detection needs them (pick channels AFTER ICA)

Ordering

• Apply BEFORE: ICA if you want to exclude certain channel types from decomposition
• Apply AFTER: ICA if you needed EOG/ECG channels for artifact detection
• Generally early in pipeline — reduces computation for subsequent steps

Common Patterns

Goal	Selection	Notes
Remove stim channels	pick_channels:eeg	Keeps only EEG-type
Motor cortex subset	pick_channels:C3,Cz,C4,FC3,FC4,CP3,CP4	MI BCI
Remove EOG after ICA	pick_channels:eeg	Apply after ICA step

Reference Code

MNE Chain (on existing Raw object)

raw.pick_channels(['C3', 'Cz', 'C4'])

Standalone (numpy, by name list)

import numpy as np

pick_names = ['C3', 'Cz', 'C4']
pick_idx = [i for i, ch in enumerate(channels) if ch in pick_names]
data = data[pick_idx]
channels = [channels[i] for i in pick_idx]

By Channel Type

# Select only EEG channels from mixed recording
ch_types = data_dict['meta']['ch_types']  # e.g. ['eeg','eeg','eog','stim',...]
indices = [i for i, t in enumerate(ch_types) if t == 'eeg']
data = data[indices]
channels = [channels[i] for i in indices]

Key API

• raw.pick_channels(names) — MNE method
• raw.pick_types(eeg=True) — alternative for type-based selection