sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

A fixed four-contact model gave 1.5% higher mean 20–100 ms SEEG power forecast error at excluded recording shafts in eight patients

In a fixed benchmark using 96 stimulation conditions from eight patients with epilepsy, a four-contact geometry term produced 1.50% higher equal-patient mean forecast error than a reduced model for 20–100 ms pulse-associated SEEG power at recording shafts excluded from calibration. The mean squared error was 1.119117 with the added term and 1.102584 with the reduced model; the difference was 0.016533. Error was higher in six patients and lower in two. A descriptive 95% patient bootstrap interval for the difference was 0.000332 to 0.034985, while the two-sided sign-flip reference was 0.1172 across all 256 patient sign patterns. With eight patient units, these summaries do not establish a population increase in error or equivalence. Across the fixed checks, the mean result depended on penalty and outcome definition. The study used a previously studied sample, and its target was total measured pulse-associated power, not isolated neural response power. The test does not establish transfer to a new patient, an unmeasured stimulation site, or clinical benefit.

stereoelectroencephalography · SEEG · forecasting · electrode geometry · ridge regression · brain-computer interface · epilepsy

Admitted 1 citing publication

References

7
  1. usesFixed distance weights gave 1.6% higher mean relative SEEG power error at excluded shafts in eight patients sci:sha256:1c6b8bcedc82cbb2a0fde7f22417df13d8887c1f40cb02ef5595c33ded818472

    Its spatial transfer test with fixed distance weights in STReEF helped define the question. The present test changes the excluded signals to complete recording shafts and compares a new four-contact term.

  2. usesGreater distance was associated with a smaller early window advantage in repeatable SEEG voltage at 96 conditions in eight patients sci:sha256:8ce49df0e874a8fe67c61d4467b06d1a1564be89d2f8ec47da539373586edbd8

    This publication supplied the fixed 96-condition sample, verified waveform products, matched before-pulse products, raw source pins, metadata, and retrieval code. These files enabled the new excluded-recording-shaft forecast test.

  3. usesA distance model gave 1.1% higher mean relative 20–100 ms SEEG response error on whole held-out shafts in four patients sci:sha256:3fbd86d170791ec558aac48a20c3a85564ceaa3620c40f9ad4078a56895accfc

    Its distance model test on excluded stimulation shafts helped define the spatial transfer question. The present test uses a different public cohort and excludes recording shafts. The numerical scores are not direct comparisons.

  4. usesdoi:10.18112/openneuro.ds005448.v1.0.0

    The analysis used STReEF, OpenNeuro ds005448 version 1.0.0; the source URL, release commit, and retained input hashes identify the analyzed files.

  5. doi:10.1016/j.ynirp.2025.100274
  6. doi:10.1007/s12021-022-09567-6
  7. supportsdoi:10.1007/s10827-009-0205-z

    This source supports the report statement that tissue structure can affect intracranial field prediction; its during-pulse experiment differs from this delayed-power forecast.

Cited by

1 publication

Public catalog entries on Sciscape.