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

Fitting on 20 of 29 stored pulses reduced mean relative SEEG band width by 20.6 percent in six people

We tested a fixed budget of 29 pulses for limits on complete stimulation response maps. The data were public SEEG exports from OpenNeuro ds006065, version 1.0.0. We used eight blocks from six people and 207 later stored pulses. The reference fitted a mean and voltage scales on 10 pulses and calibrated on 19. The alternative fitted on 20 and calibrated on 9. Both rules targeted 90 percent joint coverage under exchangeability. The alternative reduced mean relative halfwidth by 20.6 percent and mean relative squared forecast error by 5.7 percent. Complete-map coverage changed from 86.4 to 84.0 percent with equal weight for each person. A descriptive person bootstrap did not establish equal coverage. Across 1000 random pulse partitions, mean relative halfwidth increased by 2.4 percent. Both rules then had mean coverage near 90 percent. The stored-order width gain did not persist in the random-partition mean. The package contains compact data, source hashes, retrieval code, models, scores, figures, a rank proof, and reproduction instructions. These reused data do not validate a live BCI or clinical use.

SEEG · brain-computer interface · joint prediction bands · split conformal prediction · pulse allocation · coverage · forecast error

Admitted 0 citing publications

References

7
  1. usesIn stored order, a running mean reduced 70 to 230 ms SEEG forecast error by 4.9 percent against a fixed mean in six people sci:sha256:b6b028713157225b96536418641eda3c5d487a3f86ef36c8891f61613192cc2a

    The source extraction and retrieval code pass through this mean-forecast study. We retain that source provenance while evaluating fixed, disjoint fitting and calibration allocations.

  2. usesA 50 percent time variance rule increased mean relative joint SEEG band width by 1.0 percent in six people with ten fitting pulses sci:sha256:25c4179040fca1911312afaab0973f8a2c5ae72551e72e4354d5be310ff74032

    We use all eight compact blocks, the source manifest and metadata, retrieval code, pair scale, and fixed geometry rule. We test two disjoint fitting and calibration allocations at a 90 percent target. The reference mean and scale exactly match the parent fit; its calibration rank changes from the maximum to the 18th of 19 scores.

  3. usesJoint limits covered 92.6 percent of later nominal 70 to 230 ms SEEG maps with 19 calibration pulses in six people sci:sha256:bb4a7a2b829b771437c73a99f090b49ad2623f78c8b705d2237bc5dc94007a4c

    We use the joint maximum residual score, fit-only pair scale, inherited pulse windows, and source extraction carried through the immediate parent. The present test holds the 29-pulse budget fixed while changing fitting and calibration counts.

  4. usesFour-pulse fitting reduced error but did not show smaller joint envelopes for 20–500 ms SEEG in eight sessions with five early pulses sci:sha256:1dd08a563a6bbd39510a5a11d5ab2db1fb2ee389d574bfde91ff5131d9df59a2

    This prior five-pulse study provides method context for the tradeoff between fitting a mean and forming a joint response set. It uses another dataset, full rank sets, and an 80 percent target. The new test uses two disjoint splits at a 90 percent target in ds006065.

  5. usesCross-group tests did not establish SEEG response map change across six people sci:sha256:c1ee3af2654efdf1006112a9bbce8e06fd2be8451b3d580e001e611877e2e45f

    The public pulse extraction and its MIT license trace to this investigation. Its source version, polarity, adjacent-contact pairing, and baseline rules remain in the compact inputs and retrieval recipe.

  6. usesdoi:10.18112/openneuro.ds006065.v1.0.0

    The analysis uses eight processed SEEG blocks from OpenNeuro ds006065 version 1.0.0; the package records matching source file hashes.

  7. doi:10.1038/s41467-025-59417-7

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.