Removal of shared voltage reduced site identification accuracy for a fixed method in two SEEG recordings with six contacts
We tested the cost of shared voltage removal for stimulation-site identification. The data were 630 pulses from 63 sites in two public SEEG recordings from OpenNeuro ds004774, release 1.0.0. Each recording had six depth contacts. A fixed method used 20 to 90 ms after each pulse. The first five pulses per site trained the reader. The next five tested it. Removal of shared voltage reduced accuracy from 49.7% to 40.6% in UMCU21 and from 68.0% to 54.0% in UMCU25. Paired whole-site bootstrap limits for these changes were -15.15 to -3.64 and -21.33 to -7.33 percentage points. These limits condition on the fitted models and tested sites. All fixed checks kept the direction of the accuracy loss. Before-pulse controls gave accuracy near the uniform guess rate. Contact differences retained 65.7% and 54.3% of a signed split-pulse voltage cross-product. These are not shares of information. We also derived an exact worst-case common-reference error radius for the fixed full reader. All 315 test decisions passed checks on both sides of their computed boundary. The package preserves four figures, inputs, code, models, predictions, and source records. Fresh retrieval checked all 630 used raw ranges. A numerical replay matched the saved results. The study does not separate neural activity from stable stimulation artifacts. It does not establish performance in a live brain interface.
stereoelectroencephalography · site identification · common reference · contact differences · brain-computer interface · bootstrap
References
7- usesCovariance angles kept 37.9 percent mean site accuracy under nine simulated positive common voltage gains in five offline SEEG sessions sci:sha256:0dbd4e1ba615799ff752cce0c46663f54b7af5ebdf8dd5a55d3ee4cebeb486e4
The earlier reader study under simulated common gains supplied a comparison for the error model. Our exploratory radius instead treats additive common reference voltage with a fixed model.
- usesTen-pulse means had higher response label AUCs than first pulses for a fixed 10–90 ms absolute score in two SEEG recordings sci:sha256:1dfa7a52fe9acf5f4a7a85e2004bf3a19d22575d694466d89d391e4150e0d226
This publication defined the original ten-pulse trial selection. We kept its sites and pulses through the direct parent package.
- usesPulse sign tests across six contacts gave 9% call rates before stimulation in two SEEG recordings sci:sha256:a5c2810414a877f27a01d90bf6b7432c2bb98cb5b20da654ecd16a6bb9a199fb
This publication supplied the unchanged SEEG epochs, trial timing, source metadata, pinned range hashes, and retrieval code. We checked the used raw ranges again and applied a new site identification analysis.
- usesSubtracting the median voltage of five other contacts reduced source response label AUC estimates at 10–90 ms in two SEEG recordings sci:sha256:4ff9a58e12ce8e2fbbbeb0aa4b58e845e694168f72dd43f752dc3a6a0187a1de
The earlier reference analysis on these recordings supplied the comparison with response labels and shared energy in ten-pulse means. We used that scope to define a separate test of repeated site differences and later-pulse site identification.
- usesdoi:10.18112/openneuro.ds004774.v1.0.0
The analyzed source data are from OpenNeuro ds004774 version 1.0.0.
- doi:10.1016/j.jneumeth.2025.110389
- usesdoi:10.1016/j.neuroimage.2015.12.012
The split-pulse cross-product method and its noise assumptions are described with this methods reference.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.