SEEG stimulation artifacts decay with distance but remain above baseline noise on 90% of channels beyond 60 mm

We measured how far electrical stimulation artifacts spread in SEEG recordings from 102 stimulation trains in 18 patients in OpenNeuro ds005169 v1.0.0. Trains used 43 or 50 Hz biphasic 1 ms pulses at 0.2–3 mA. For each recording channel, we measured peak-to-peak amplitude around each pulse and compared it with the same measure at random baseline times. The artifact decayed with distance with a median exponent of -2.18 (IQR -2.52 to -1.77); 62% of stimulations decayed faster than distance^-2. Artifact amplitude at 20 mm scaled with current as mA^0.99 (pooled) and mA^1.03 (within subject). Median artifact amplitude per mA was 346 µV at 20–30 mm and 36 µV beyond 60 mm. The median baseline noise floor was 11.4 µV. Artifact exceeded this floor on 97% of channels at 40–60 mm and 90% beyond 60 mm. The extrapolated artifact-free radius had a median of 91 mm (IQR 77–113 mm). These results indicate that, in this dataset, stimulation artifacts often extend across the SEEG implant and should be considered on distant channels in systems that record during stimulation. We also report a polarity-invariant pulse-locked component, but do not validate it as neural: an artifact-like step shape occurs far from the stimulation site, and nearby measurements may include neural responses and nonlinear artifact. Only the first 2 s of each train and 1.95 s of baseline were analyzed. Polarity labels were inferred from the recordings, and the dataset includes only stimulations that evoked visual percepts. Distances use the supplied electrode coordinates; the artifact-free radius is a log-log extrapolation. Peak-to-peak amplitude includes the pulse and does not measure artifact-tail duration.

SEEG · intracranial EEG · stimulation artifact · electrical stimulation · artifact rejection · bidirectional brain-computer interfaces · signal processing