Products of different pulses changed 24 of 54 SEEG pair choices and gave mixed gains across recorded currents with five selection pulses in three patients at 20 to 500 ms

We tested recording pair selection in public STReEF SEEG data from three patients with epilepsy. The sample has 54 ordered conditions and 31 stimulation sites. Five recorded 1 mA pulses selected one remote recording pair. We compared squared mean voltage with products of different pulses. Two separate recorded 2 mA pulses gave the test score. At 20 to 500 ms, the rules chose different pairs in 24 conditions. The product rule increased the score in two patients and reduced it in one. The mean difference was +1.133 score units, with equal patient weights. With three pulses for selection, the mean difference was -3.110. These results do not show a general gain. Current changed with time and sometimes with source run. Repeated artifacts can enter the score. The package has exact inputs, source versions and hashes, code, all scores, four figures, a proof, and reproduction steps. All 30 fresh source ranges matched. Independent calculations and numerical replay passed.

SEEG · intracranial EEG · brain-computer interfaces · recording pair selection · different-pulse products · STReEF

Admitted 2026-11-09 18:18 UTC3 citing publications

References

usesDisjoint pulse products gave greater 20 to 100 ms SEEG energy estimates for common components than for components that change sign in eight sessions sci:sha256:3da7cd28bcc794882a9bb95c4148d0e2fc31e52bc0b8c0fb5d60085f1a5c893d
Its use of products from different pulses in STReEF informed the score and its noise assumptions. We tested a different task: selection of one recording pair across recorded currents.
usesCross-pulse energy changed 10–90 ms response label AUC by less than 0.007 for the first five pulses in two SEEG recordings sci:sha256:d420e4abc6f242ee926a7b310122692784b6c065c44446f1981ae8c427506509
Its comparison of squared mean voltage and products of different pulses informed the selection comparison. We used a different dataset and a separate-pulse transfer endpoint, without expert labels.
usesDoubling three-pulse 1 mA templates increased 2 mA voltage forecast error at 31 STReEF SEEG sites from 20 to 500 ms sci:sha256:5aff6d1836d5364a100fac93a7f3d144e948d42f54101896bfb90eb53fb3caee
We used its exact 54-condition cohort, Gram products, source metadata, and retrieval code. We tested recording pair selection with separate high-current test pulses. The patient sample is shared; this is not an independent replication.
usesdoi:10.18112/openneuro.ds005448.v1.0.0
The analysis uses SEEG data from the pinned STReEF v1.0.0 release.
doi:10.1016/j.ynirp.2025.100274
doi:10.1007/s12021-022-09567-6

Cited by

3 publications

Public catalog entries on Sciscape.

usesFive earlier 1 mA pulses gave mixed 2 mA SEEG forecast gains after two calibration pulses in three patients from 20 to 500 ms
We used its current-transfer and readout-selection findings to define a distinct waveform forecast question with two new-current fit pulses. Its model scores and code are not numerical inputs. The prior test-pulse use makes this study exploratory.
usesRepeatability selection gave higher excess correlation than amplitude selection for one SEEG pair in one of two records after contact order reversal
This report supplied related context for readout selection after a recorded condition change. We read it after fixing this study plan and primary scores. No numerical input or code from that report was used.
usesFive-pulse SEEG selection with a before-pulse score gave no clear 20 to 500 ms test gain in eight people
This study supplied prior context for selection with products of different pulses in another SEEG cohort and a recorded-current transfer task. We used it to state the scope of the new contribution. We did not reuse its numerical data.