Scaled five-pulse forecasts had more skill for centered bipolar SEEG at 20–100 ms than 100–500 ms in eight patient sessions
We tested voltage forecasts for a brain-computer interface response monitor. We used public SEEG from OpenNeuro ds004696 version 1.0.1. The sample had eight epilepsy patient sessions, 86 stimulation sites, and 886 stimulation-to-readout pairs. Each forecast used a scaled mean of the first five good pulses. Its targets were the final five good pulses, scored separately. Each bipolar waveform had its time mean removed within each fixed window. Skill means percent less squared error than a zero voltage forecast. Equal-patient mean skill was 34.77% at 20–100 ms and 17.44% at 100–500 ms. The early advantage was 17.33 percentage points and had the same direction in all eight patients. Its descriptive 95% patient bootstrap interval was 11.27 to 23.28 points. The exact two-sided patient sign-flip test gave p = 0.0078125 under independence and sign symmetry. The before-pulse control had mean skill of −1.74%. The scale rule comes from prior work. The new result compares two fixed-window scores against zero voltage. The package contains exact inputs, full results, three figures, code, source hashes, and reproduction instructions. All eight public source derivatives were rebuilt and matched the archived products exactly. Limits include unequal window lengths, separate time centering, retrospective quality selection, and one session per patient. The result does not establish neural origin, behavioral control, or transfer to a new session.
stereoelectroencephalography · electrical stimulation · forecast skill · bipolar voltage · response monitoring · patient-level inference · OpenNeuro
References
4- usesFive-pulse noise scaling reduced later single-pulse error for 20–500 ms centered bipolar SEEG in eight patient sessions sci:sha256:fbc06ab469d48eac7a6f54dd97cd24ccd9c7c5ca4b0db4d64340536b3e81428a
We used its fixed five-pulse noise scale and public source extraction method. We did not fit a new rule on test data. We extended its zero voltage control to a primary contrast between 20–100 ms and 100–500 ms.
- usesAn offline two-pulse site rank did not show lower 20–500 ms SEEG forecast error at a fixed budget in eight sessions sci:sha256:578806530d9c396756538b53e11352e83c900274801f852ef7211403fa594c3c
We used its 25 exact input files for eight patients, 86 sites, and 886 pairs. Its products preserve the corrected chronological pulse selection. We added a fixed early-to-late skill contrast against zero voltage.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
This study uses the processed SEEG data from OpenNeuro ds004696 version 1.0.1.
- doi:10.1523/JNEUROSCI.2201-22.2023
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesFive-pulse consistency selection had an uncertain 5.1-point mean scaled forecast gain over nearest SEEG pairs in eight patient sessions
This work supplied all 886 exact ten-pulse products, the scaled forecast rule, source extraction code, and the inherited eight-patient source audit. We add physical nearest-pair selection and region and shaft checks. The later pulse data are reused.