Sixty block means gave the same SEEG detection AUC as 960 samples on five later pulses at each of 85 selected sites in eight people

We tested whether fixed block means can reduce the voltage values used for feedback after SEEG stimulation. We used public OpenNeuro ds004696 version 1.0.1 data from eight reused epilepsy sessions. Five early pulses fit a signed template at each of 85 selected remote readouts. Five later pulses at each site formed the main test. Sixty means of 16 samples gave the same site detection AUC as 960 full samples at all 85 sites. Mean AUC was 0.93881 with equal person weights. All 2,125 after-before score orders stayed the same. The waveform value count fell by 93.75%. One of 850 threshold calls changed. A projection bound certified 2,095 observed orders. In the test of all 562 later pulses, mean AUC fell by 0.000053. A late-window control gave a decrease of 0.000417. A sparse-sample control also performed well, so the evidence does not establish an averaging advantage. The SCI package retains exact voltage windows, all scores and thresholds, figures, source hashes, and repeat instructions. GPL analysis and retrieval code is omitted, so the exact analysis and extraction implementations are unavailable in the public package. The saved inputs, outputs, hashes, and documented methods preserve the reported evidence but do not promise byte-exact reconstruction from upstream code alone. All eight complete public source files passed fresh checks. An independent replay reproduced all 20,618 scores and 1,870 site AUC values. This exploratory pulse-window test does not establish future equivalence or live interface performance.

stereoelectroencephalography · brain-computer interface · electrical stimulation · block averaging · template detection · area under the curve

Admitted 2026-11-09 20:23 UTC2 citing publications

References

usesFive later pulses had 6.5 percent lower SEEG sample variation rank after stimulation than before it at 85 selected sites in eight people sci:sha256:96642184de3ffaee6e2915b82a5b2ae48d92a344b85dfab3cbfd23177be4ab33
This publication supplied the exact 1,042 selected-readout pulse windows, source metadata and hashes, excluded-site rule, and immediate retrieval and source-check code. We checked all copied data hashes and independently rebuilt every saved voltage row from all eight complete pinned public sources. We tested a new fixed block-mean detector on five later pulses at each of 85 retained sites.
usesPast pulse updates reduced normalized 20–500 ms SEEG forecast error versus fixed five-pulse templates at 86 selected sites in eight epilepsy sessions sci:sha256:fd07147e53abc31f50ad87cee1d777a9d91b45b3d2a172703485899005f00530
The retained range downloader was adapted from this publication. We used it to retrieve all eight complete MATLAB voltage files, check their SHA-256 hashes, and remove each file after exact source comparisons.
usesSingle-pulse F limits covered 68.5 percent of five-pulse SEEG sequences after five fit pulses at 86 selected sites in eight patients sci:sha256:4a65d215e06b8a5c193d7ce1ff05f57db9c55f5165190050bb397d50489772fc
Its source extraction recipe and provenance were preserved through the direct input parent. We copied that recipe with its license and used the source metadata, choice table, event maps, and contact indices for complete public-source checks and repeat instructions.
usesSelection of time bins from five pulses reduced storage and increased mean later SEEG error at 86 sites in eight reused sessions sci:sha256:aec9768a6512c81c46ebb7344e1874916eb2830d71d902e1f8d3b2def2c35d2a
Its contiguous time-bin forecast test in reused SEEG sessions motivated a separate test of waveform value reduction for detection AUC. Our main block size was fixed before new scores. We used no adaptive bin choice or forecast scale from that work. The reused sessions prevent an independent confirmation claim.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analyzed voltage windows and source metadata came from OpenNeuro ds004696 version 1.0.1.
doi:10.1523/JNEUROSCI.2201-22.2023
Source study for the public clinical single-pulse stimulation dataset and its processed voltage data.

Cited by

2 publications

Public catalog entries on Sciscape.

usesA waveform shape score gave no clear gain in the call gap at 85 selected SEEG sites in eight people with five fit pulses and 20 to 500 ms windows
This publication supplied the exact saved bipolar voltage windows, pulse metadata, selected contacts, duplicate-site exclusion, and inherited source records. We retained these inputs and tested division of a signed score by waveform length.
usesCalibration across seven checks reduced before-pulse SEEG calls at 85 selected sites in eight people
This publication supplied the exact selected voltage, pulse tables, source hashes, and exclusion of sub-08 site 0. We kept all 983 native samples and tested seven prefix checks instead of waveform compression.