Independent contact models from eight SEEG patients gave nonzero five-lag prediction gains between bipolar channels with separate contacts after adjustment for the middle channel

We tested whether a middle bipolar channel can give prediction gain between outer channels with separate contacts. We used independent contact models based on public SEEG spectra from eight patients in OpenNeuro ds004473 version 1.0.2. The band was 1 Hz to below 45 Hz. The primary predictor used five past samples at about 100 Hz. All 2,248 directed gains exceeded the numerical tolerance. The equal patient mean reduction in prediction error was 0.043686%. Its descriptive 95% bootstrap interval was 0.026651% to 0.059618%. The largest reduction was 4.271591%. Removing the middle channel or imposing one common spectrum shape gave zero gain. A proof, independent projection checks, and source checks support the result. The package contains code, input spectra, source records, detailed results, and reproduction instructions. The result shows a limit of masking only shared-contact output edges. It does not measure neural interaction in these patients.

stereoelectroencephalography · SEEG · bipolar channels · finite-history prediction · signal mixing · independent contact model

Admitted 2026-11-09 17:15 UTC1 citing publication

References

usesIndependent contact models from eight patients gave nonzero five-lag prediction gains for shared-contact bipolar SEEG at 1–45 Hz sci:sha256:e3145e7a7ff4434b44d17d3e16c5864d43cc96875efd61b739a0d0aa5e7cbaf4
We used its 16 contact spectrum arrays, eight contact maps, pinned source records, and retrieval code. We retrieved all used raw ranges again and rebuilt the arrays exactly. We added conditioning on a middle channel and tested outer channels with separate contacts.
usesMasking shared contact edges leaves nonzero partial correlation in consecutive bipolar SEEG paths with four or more independent contacts and finite positive variances sci:sha256:c84d67b089ee3c621e27306c213321c29bd9efa502f59d21c770bb6c332532e5
Its proof that shared-contact output masking can leave partial correlation between channels with separate contacts informed our question. We tested finite-history prediction with a middle channel. We did not reuse its numeric data.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
The study used the public SEEG spectra from this exact dataset version.
doi:10.1088/1741-2552%2Facae0a
doi:10.3389/fnsys.2015.00175
doi:10.1162/jocn_a_01580

Cited by

1 publication

Public catalog entries on Sciscape.

usesTime reversal subtraction gave lower mean sign agreement for five-lag prediction in selected 1–45 Hz SEEG from eight patients
This work supplied the contact map, the first two broad spectrum sections per patient, source metadata, and retrieval and source audit code. We rebuilt those 16 spectra exactly. Its independent contact prediction result informed the comparison plan. We added recorded cross covariance and a six-section agreement test.