Added remote SEEG power did not lower mean error for moving versus stationary visual effect labels in a fixed test of 17 people
We tested whether remote SEEG power after visual stimulation improved prediction of moving versus stationary visual effect labels in a fixed test of 86 records from 17 people. Each test person was excluded from preprocessing and model fitting. Mean log loss increased from 0.621998 to 0.635153 nats per record; mean gain was -0.013156 nats, with a descriptive 95% person bootstrap interval of -0.039763 to 0.017763. Four people had lower error. The within-person permutation test gave p = 0.546. This reused-cohort test did not establish an added prediction benefit and does not rule out other SEEG readers.
SEEG · visual effects · prediction · power spectra · cross-person validation · log loss · brain-computer interface
References
3- usessci:sha256:3d1647c9a767ea0b389c4475596fc0a13c65b2f5d64e1cb07e481cdea7a95f9f not known to Sciscape
We reused all 112 record spectra and the source metadata, hashes, extraction logs, and raw signal code. These inputs enabled the new visual effect label test.
- usesdoi:10.18112/openneuro.ds005169.v1.0.0
The analyzed source records and metadata came from OpenNeuro ds005169 version 1.0.0.
- doi:10.1002/hbm.25749
Background method paper describing the cortical stimulation study; this publication uses marked pre and post intervals and does not apply its full artifact separation method.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesA reader with early post-stimulation SEEG power had lower AUC than a descriptor reader in a fixed pair test of 17 people
This study reused its 86-record eligibility rules, 33 features, five fixed readers, weighted logistic fit, source spectra, metadata, and raw retrieval code. Its one-person predictions were reproduced exactly. The new analysis applies a grouped pair AUC test and repeats all fits after each person is removed.