A noise-scaled linear time trend increased SEEG forecast error for the next pulse at 20 to 500 ms in eight patients with epilepsy
The tested time trend increased forecast error by 26.8% on average across eight patients. Each patient had higher error. The descriptive 95% patient bootstrap interval was 5.8% to 60.9%. This result applies to the specified rule and recordings. We compared a noise-scaled linear trend with the mean of all earlier responses using public SEEG trial voltage derivatives from OpenNeuro ds004696 v1.0.1. The test covered 85 stimulation sites, 879 site/readout instances, and 562 target pulses in a 20 to 500 ms response window. The short-gap and pulse-order checks also showed higher error. One site was excluded because two event labels mapped to the same native sample. The result does not test a stimulation control policy, therapeutic effect, task decoder, or whether a mean is best among all possible models. The cohort was small, reused, and not randomly sampled. The package preserves result products and figures, but excludes the GPL analysis implementation, so the exact calculation cannot be rerun from the SCI payload.
SEEG · brain-computer interface · forecast error · single-pulse electrical stimulation · linear trend · waveform prediction · OpenNeuro ds004696 · epilepsy
References
4- usesRetained pulse order showed no clear extra gain for mean SEEG forecasts at 20–500 ms with variance scaling in eight sessions sci:sha256:5377acec1becadfd257031661120d1cf23f4f6b6a9d92b5e883ae432fdad76ee
The earlier retained-pulse order test helped define the question and the order control. It used the same cohort. Its evidence is not an independent replication of this trend test.
- usesPower weights fit from five pulses gave no clear mean 20–500 ms SEEG voltage forecast gain in eight sessions sci:sha256:5f8ad290d211abb81202a82ed21683bfe60c8df22cceb4e18c340cf2fc14ee28
This study uses the all-pulse waveform products, source metadata, extraction code, and provenance. It tests a different time-trend forecast with all past responses and a scale fixed from the initial five pulses.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analysis uses public trial voltage derivatives and event/channel metadata from OpenNeuro ds004696, version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesA limited time trend increased SEEG forecast error by 9.6% against the past mean in five 6 mA sessions at 20 to 100 ms
Its failed noise-scaled time trend motivated the new fixed limit on forecast movement. Three raw recordings overlap between the datasets.
- usesAt rate 1, learned weights reduced 20 to 500 ms SEEG forecast error against equal weights but raised error against scaled means in eight recordings
We used the exact all-pulse Gram inputs, native voltage example, extraction helpers, and source records. We calculated new sequential mixture forecasts and errors. We rebuilt sub-05 from its complete public source and checked all metadata files again.