sEEG Artifact Detector — specifics

Menu Process → Signal → SEEG Artifact Detector 1.0.0. See the shared methodology for the five criteria and the common parameters — this page only covers what's specific to sEEG.

Not yet validated

This plugin is functional but has not yet been run against a real sEEG recording with known bad contacts. Every default below, including the same-shaft neighbor logic itself, should be treated as a reasoned but unverified starting point.

Neighbors: same electrode shaft, not 3D nearest contacts

This is the one part that genuinely differs from MEG/EEG, not just in its numbers. sEEG electrode shafts are independent trajectories through the brain: two contacts that happen to sit close together in 3D space but belong to different shafts are generally not functionally or anatomically related, unlike two nearby MEG sensors or scalp EEG electrodes. A plain 3D nearest-neighbor search would therefore compare a contact against the wrong population.

Instead, "neighbors" (criteria 2/3) are the neighbor_count nearest other contacts on the same electrode shaft, ordered by contact number. This is determined by parsing the channel name, not by 3D position: shaft letters, an optional trailing apostrophe (the common left-hemisphere convention, e.g. A'), then the contact number — for example A1, A'3, TB12. Channel names that don't match this pattern (most commonly a bipolar derivation like A1-A2, whose name is just its anchor contact A1 and parses fine, or a non-standard label) are excluded from criteria 2/3 with a warning.

neighbor_count defaults to 2, not 5

A shaft typically only has 8-18 contacts, spaced close together (~3.5 mm): only the immediate 1-2 contacts on each side are plausibly comparable tissue — contact 6 vs. contact 1 on the same shaft can already span gray matter, white matter, and a lesion. Hence a lower default than MEG/EEG's 5.

Bipolar montages: the clinical norm, not a warning sign

Unlike EEG, bipolar derivations are the standard way of reviewing sEEG (they remove common far-field signal), so their mere presence does not trigger a warning. What does get a warning is a partial mix of monopolar and bipolar channels in the same montage — that is the genuinely inconsistent case for criteria 2/3's cross-channel amplitude/spectral comparisons.

PSD frequency bands — configurable, inherited from EEG

Same parameters and defaults as EEG Artifact Detector (psd_band_low_hz/psd_band_high_hz, psd_peak_low_hz/psd_peak_high_hz — 20-70 Hz / 2-100 Hz). These are inherited from the EEG plugin purely as a starting point: the rationale itself is weaker for sEEG than for EEG, since EEG's 20-70 Hz default targets scalp/jaw EMG contamination, a surface phenomenon far less prominent on intracranial contacts.

Command line

--run "SEEG Artifact Detector" — see using these plugins from the command line.