Eight learned shapes increased mean relative SEEG prediction error with two fit pulses in five offline sessions at 20–500 ms

We tested if shapes from other stimulation shafts improve a voltage estimate made with two local pulses. The public source was OpenNeuro ds004457 version 1.0.2. Five patient sessions supplied 183 sites and 732 later test pulses. Each target shaft was excluded from shape training. Eight learned directions per recording pair gave 8.06 percent higher relative error than the full two-pulse mean. This score gives each patient equal weight and uses 20 to 500 ms. The descriptive 95 percent patient bootstrap range for error reduction was -23.09 to +1.56 percent. Three patients had less error. The same learned directions gave 6.69 percent less mean relative error than eight cosine directions. All five patients had less error in that comparison. A four-pulse mean gave 20.47 percent less mean relative error than the two-pulse mean. This reused cohort does not show a clear gain over the planned full-waveform reference. The split was later within each site but overlapped across sites in absolute time. Fresh extraction matched all archived inputs. The full numerical replay passed. The package preserves input voltage, source hashes, models, predictions, code, three figures, and reproduction instructions.

SEEG · electrical stimulation · voltage prediction · response shapes · singular value decomposition · brain-computer interface · reused data

Admitted 2026-11-09 19:55 UTC2 citing publications

References

usesA fixed covariance rule gave higher SEEG site accuracy than a diagonal rule in offline tests from five patient sessions sci:sha256:1407cd9ac712a4bcd551a3738cc90c3fab309b871572cbd6fc644b41555f7886
This is the original archive source of the binned voltage inputs and fixed recording contacts used through the response-shape publication.
usesOne shared temporal noise matrix gave lower mean SEEG site accuracy than full covariance with two fit pulses per site in offline tests of five sessions sci:sha256:80660247e5d8e43c18f311062884ad6ed208a80d726410b122af21a5b954b4c0
Supplied an independent raw reconstruction record for the same archived bins. This record is preserved with its publication metadata.
usesOne waveform learned from two pulses per site missed repeated site differences in five SEEG sessions sci:sha256:d08ca03b09b85e16bff0cce29a0e32d65988740ca9cc1eb935f43e2d66c2eadd
Supplied the hash-checked voltage bins, fixed cohort rules, and raw extraction recipe. Its response-shape result helped define the new two-pulse test.
usesEight donor shapes reduced five-pulse SEEG forecast error versus eight cosines at 86 selected sites in eight patients; full waveforms had less error sci:sha256:d5f05d9adcd5c0419c50f10231ba2502cf7acc24728e09a3c09631e52e5a88e3
Its eight-shape comparison in a different cohort informed the fixed rank and the full-waveform comparison. The present test uses other shafts and two local fit pulses.
usesdoi:10.18112/openneuro.ds004457.v1.0.2
The analysis reuses binned voltage inputs from OpenNeuro ds004457 version 1.0.2.
usesdoi:10.1523/JNEUROSCI.1325-22.2023
This is the source study for the dataset used in the analysis and is identified by the dataset attribution notice.

Cited by

2 publications

Public catalog entries on Sciscape.

usesAn 80 ms SEEG window at 100 ms gave lower site accuracy than at 20 ms in five offline 6 mA sessions
We use its voltage bins, metadata, source manifest, extraction code, and fixed cohort code. Fresh source extraction matched all inherited inputs. The patients are reused.
usesFour local pulses gave lower 20–100 ms SEEG forecast error than one pulse with disjoint-site projection in all five tested patients
This report was read after the new primary scores for comparison and for the projection loss identity discussed in PROOF.md. It tests several shapes from other shafts with two local pulses. The present test adds strict donor timing, one disjoint donor, and distinct input pulse counts.