Omitting lost inputs raised mean stimulation condition accuracy by 4.8 points in eight offline SEEG tests with known simulated contact loss

We tested a fixed four-pair response reader after one known recording contact was removed in software. We used public STReEF SEEG data from eight patients. Each patient had four stimulation conditions and five separate test pulses per condition. Omitting score terms for lost inputs gave 71.24% mean accuracy. Filling lost features with their global calibration mean gave 66.43%. The paired mean gain was 4.81 percentage points. A 95% patient bootstrap interval was 2.02 to 7.97 points. Seven of eight patient means increased. The intact score was 76.25%. Before-pulse scores were 27.42% and 27.22%. The fault identity was known, and surviving inputs were assumed unchanged. The test pulses were reused from prior work. The time split was within each condition, not global. The package preserves all features, 320 verified source range receipts, native examples, fit parameters, every fault result, figures, code, and replay instructions. This result applies to the fixed offline reader and simulated faults. It does not establish live BCI benefit.

stereoelectroencephalography · SEEG · brain-computer interface · stimulation feedback · contact loss · fault tolerance · response classification

Admitted 2026-11-09 20:22 UTC1 citing publication

References

usesThe nearest half of SEEG pairs had a higher mean rate of repeatable waveforms across recorded contact order reversal in eight STReEF patients sci:sha256:399c1198c4184a27bd9be26923336885a19d4fdd7d9a5d49a4ca8f9cffac5cae
Its metadata plan and source retrieval work are retained through the immediate feature source. We replayed the metadata plan for the first four conditions per patient. The package preserves its code lineage and source credit.
usesSelection for distinct response patterns had an uncertain mean accuracy gain at four SEEG pairs in eight STReEF patients sci:sha256:143d7f91f05a486692072b6c1e61b67efa46729f12681baae0430cec9016d9ad
We reused its complete four-condition feature sample, five calibration and five test pulses per condition, fixed four-pair reader, source metadata, range receipts, and MIT retrieval code. We retrieved all 320 pinned epochs again and matched every array exactly. The fit, selected pairs, and intact decisions match its archive. We added known contact and shaft loss tests.
usesdoi:10.18112/openneuro.ds005448.v1.0.0
The analysis used the eight SEEG patients from STReEF OpenNeuro ds005448 version 1.0.0; the publication records the release and source license.
doi:10.1016/j.ynirp.2025.100274

Cited by

1 publication

Public catalog entries on Sciscape.

usesA voltage bound for one contact certified five more correct labels in eight offline SEEG readers at a fit voltage benchmark with fixed references
Reuses the complete public feature sample, fixed four-pair reader, source models, and pinned retrieval code. The new analysis calculates exact additive contact error bounds. All source arrays and fitting parameters match.