Trend correction kept SEEG decisions fixed under added linear voltage drift with fixed power inputs in 11 patients
We tested a voltage trend correction in public SEEG data from OpenNeuro ds005415 v1.0.0. We fit a line to ten voltage bins before each cue. Its slope corrected ten bins after the cue. Eleven patients supplied 405 early training trials and 211 later test trials. All readers used eight pairs selected from early power. Correction changed mean later balanced accuracy from 57.0% to 54.2%. The paired change was -2.8 percentage points. The 95% patient bootstrap interval was -8.2 to +2.3 points. This result showed no clear accuracy gain. With fixed power inputs, added linear voltage drift changed 27.0% of original decisions at one training slope scale. Corrected decisions stayed fixed at all tested scales. Algebra and 22 raw trial checks confirmed voltage-feature cancellation. The proof requires the same added linear drift across the control and response windows. Power changes, nonlinear drift, online use, and brain stimulation were not tested. The package contains source code, input arrays, raw audits, figures, predictions, source hashes, and reproduction instructions.
stereoelectroencephalography · SEEG · voltage drift · brain-computer interface · signal correction · balanced accuracy · ridge regression · OpenNeuro
References
5- usesPower plus voltage gave 57.0% balanced accuracy for later format reading in 11 SEEG patients with eight pairs selected from early power sci:sha256:8fdc57541722e7816d5646e19ad1395ebfdd41ba149a4a4049c48b264d073a04
This study used its power and voltage arrays, retained trials, early/later split, raw audits, source records, and extraction code. It reproduced the original combined and power accuracy, then tested voltage trend correction.
- usesA fixed eight-pair SEEG reader reached 47.4% balanced accuracy between digits and spoken words on later trials in 11 patients sci:sha256:faa69c746a9855ed7e397517c4cc6f939bf3d1b0651515aaf4fff294a136c80f
This study used its inherited power features, trial eligibility, contact pair rules, and source extraction code through the direct voltage-feature parent.
- usesA window before late Cue contained samples after recorded onset in 187 of 226 beep trials from six public SEEG runs sci:sha256:dcd1254c5527f89b3aa0296f948abc5e289cdea3f238729607e9c688c14d7346
This study used its inherited pinned source object records, EDF headers, BIDS metadata, and event interpretation. The spoken-word onset uses Cue Start to avoid the late audio Cue problem.
- usesdoi:10.18112/openneuro.ds005415.v1.0.0
The analysis used the Numbers dataset, OpenNeuro ds005415 v1.0.0.
- doi:10.1371/journal.pone.0313155
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesLine removal prevented effects from added linear drift and increased clean error in 50 ms phase forecasts with 100 ms histories in 43 public SEEG records
Its earlier evidence of drift protection and clean accuracy cost for SEEG task features helped define the phase-reader question. We copied no task data from this work.
- usesLine removal prevented effects from added linear drift and increased clean error in 50 ms phase forecasts with 100 ms histories in 43 public SEEG records
Its earlier evidence of drift protection and clean accuracy cost for SEEG task features helped define the phase-reader question. We copied no task data from this work.