Variance weights from five baseline pulses gave no clear AUC gain for eight cosine features at 86 SEEG sites in eight patients
We tested baseline variance weights for a compact reader of public SEEG stimulation responses. At 86 fixed sites in eight patients, both readers used eight cosine features and five early fit pulses. Each site had five later test pulses. Mean AUC was 0.9379 with variance weights and 0.9339 with uniform noise weights. The difference was +0.00405. A descriptive 95% patient resample interval was -0.00314 to +0.01221. Three patients gained, three lost, and two had equal AUC. The fixed 100 to 500 ms check gave a negative mean difference. Variance weights also increased baseline calls at the fitted threshold. The evidence does not show a clear gain for new patients. The package preserves selected voltage, all scores and fits, three figures, source hashes, and code. It includes 86 compact readers with eight weights and one threshold each. Independent numerical checks and a fresh output run passed. This run retrieved one complete public derivative and matched all 61 selected pulses exactly. The result is exploratory evidence from a reused clinical cohort. It does not test neural specificity or a live interface.
SEEG · electrical stimulation · noise weighting · cosine features · AUC · brain-computer interface · neural response
References
6- usesAn AR(16) model with a 1 percent floor gave no clear 20–500 ms SEEG detection gain after five calibration pulses in eight epilepsy sessions sci:sha256:f498186f5bf9aa390e84158392afda66f1813af870a8cf045712bf3db794d97b
Used its full temporal noise-weighting study as context for the question and scope. The present rule instead estimates diagonal noise variances in eight fixed cosine features. The studies share the clinical cohort and are not independent confirmations.
- usesA fixed spatial noise rule gave no clear 20 to 500 ms AUC gain after five early pulses in eight SEEG sessions sci:sha256:48f93098d9fe4ee439817f49e745f4c586103d8b79d4985cbb51b1bf7dc5d020
Used its spatial noise-weighting study as context. The present study keeps one selected bipolar contact pair and weights its temporal features. It does not add new patients or test the spatial rule again.
- usesEight donor shapes gave less forecast error and 0.002 lower mean response AUC than eight cosines after five fit pulses at 86 SEEG sites in eight sessions sci:sha256:2a6f7f03fdc0a3d478dcf18e72daf6d41c9857ca607df84cba991a3aaed7402c
Used its 16 selected voltage and pulse-map files, fixed readout choices, pinned source metadata, and retrieval code. All data file hashes matched. Independently reproduced 516 uniform cosine and full-waveform site AUC values. The new comparison adds local feature variance weights on the same reused cohort.
- usesFive calibration pulses gave mean later signed template AUC of 0.939 compared with 0.859 for one pulse in eight SEEG sessions sci:sha256:1d2aa103275ad7400befec4c899740ede8f2eb59a09897bdc3f4d4e190e54418
Retained its first-five-pulse contact choice among 886 eligible readouts, selection tables, and credited selection code. This analysis is conditional on those fixed choices and does not rerun every candidate score.
- usesdoi:10.18112/openneuro.ds004696.v1.0.1
Source dataset, version 1.0.1.
- supportsdoi:10.1523/JNEUROSCI.2201-22.2023
Reports the electrical stimulation experiment represented in the analyzed dataset and supports the described cohort context.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.