Removing stimulation contacts changed selected ds005557 remote SEEG power by less than 1 dB at 70–150 Hz, excluding 118–122 Hz

We tested a reference choice for a brain-computer interface that reads after stimulation. We used 299 selected word windows from 12 people and 19 recordings in OpenNeuro ds005557 version 1.0.0. We compared an average of all SEEG contacts with an average without the two labeled stimulation contacts. The main band was 70–150 Hz, with 118–122 Hz omitted. All 30,650 remote contact windows changed by less than 1 dB. The median across person summaries was 0.103 dB after stimulation and 0.102 dB in controls. An exploratory 3–30 Hz check used 19 raw audit windows and no controls. Two windows from one person had median changes of 2.113 and 3.790 dB. Small high-frequency changes did not establish a clean signal in all bands. The package preserves spectra, raw audits, source versions, hashes, code, four figures, and reproduction steps. Independent checks, fresh source retrieval, and offline reproduction passed. We did not test decoder accuracy or a causal stimulation effect.

SEEG · common average reference · electrical stimulation · spectral power · reference sensitivity · brain-computer interface