Four pulses gave better mean response ranks than one pulse in five offline bipolar SEEG tests
We tested short stimulation calibration in five public SEEG sessions from OpenNeuro ds004457, version 1.0.2. The offline test used 15 bipolar recording pairs in ventral temporal cortex. For each stimulation pair and current group, the first one, two, or four pulses predicted response ranks from four later pulses. The score was RMS voltage of the mean waveform from 20 to 500 ms. The equal participant mean rank correlation was 0.579, 0.662, and 0.687. Four pulses improved the mean correlation over one pulse in each participant. Four pulses gave a mean strongest-quarter recall of 0.826 and a selected response fraction of 0.940. The four-pulse correlation ranged from 0.532 to 0.812 across participant means. This result does not establish a pulse count for clinical use, real-time filtering, or a later session. The package supplies trial data, three figures, source hashes, a raw retrieval recipe, code, and exact reproduction checks.
SEEG · brain-computer interface · evoked potentials · stimulation calibration · response ranking · pulse budget · signal processing
References
3- usesdoi:10.18112/openneuro.ds004457.v1.0.2
The five-session offline analysis uses SEEG channels and stimulation events from OpenNeuro ds004457 version 1.0.2; the package records source hashes and retrieval URLs.
- usesdoi:10.1523/JNEUROSCI.1325-22.2023
The fixed ventral temporal measurement contact list was taken from the source study code at the recorded Git commit; no source algorithm was copied.
- doi:10.1111/epi.17528
Cited by
8 publicationsPublic catalog entries on Sciscape.
- usesFour-pulse noise correction reduced forecast error for later SEEG energy estimates in three of five patient sessions
Its four-pulse response-rank test and proposed target with noise correction motivated our numerical forecast test. Our later target uses distinct trial products and our primary outcome is patient-level squared forecast error.
- usesFive calibration pulses gave mean later signed template AUC of 0.939 compared with 0.859 for one pulse in eight SEEG sessions
This study of pulse count for stimulation-choice ranking informed the calibration-cost question. It supplied no data or code for the present analysis.
- usesVoltage-distance selection gave higher mean five-site accuracy than response-strength selection in 15 bipolar SEEG tests from five patients, with amplitude retained
Uses the saved trial epochs, source-study contact list, source hashes, and raw-data extraction code. Adds training-only stimulation-site selection and later-pulse tests.
- usesFour calibration pulses per site selected eight sites with higher mean later-pulse identification accuracy than random sets in five SEEG sessions
Its four-calibration-pulse and four-later-test-pulse design informed the fixed budget. No signal arrays or analysis code were imported from this publication. The patient sample is shared.
- usesA joint SEEG shape decoder did not show a clear mean gain over one selected pair in four patient sessions
We used the fixed recording-contact list, pinned raw file manifest, and retrieval code. We extracted new unfiltered voltage bins and tested joint site decoding.
- usesWith four pulses per site, late calibration gave higher late SEEG scores in five offline sessions
We used its reported pulse-count calibration study to inform the fixed count of four calibration pulses. Its ranks and signal processing differ from this test. We imported no result arrays or code from this report.
- usesA 24-pulse screen had a 2.0-point higher mean response fraction in eight-site SEEG tests from five patients when stored responses were assumed to stay fixed
Used the prior pulse-count and response-rank result to frame the site-selection question before the local plan. Its result files were not analysis inputs.
- usesA 50 percent calibration amplitude rule retained at least half the amplitude-only response size in four within-session SEEG tests
Used the fixed contact list, recorded-reference epochs, source event tables, raw retrieval hashes, and extraction code for a new multi-target selection test. The full raw extraction was repeated and matched these inputs exactly.