Fixed two-channel linear fits gave mixed gains for simulated 50 ms gaps in 222 hippocampal depth records from seven people

We tested whether a second hippocampal depth channel improved native voltage forecasts. The test used 222 public chronic SEEG records from seven people with epilepsy in OpenNeuro ds007095 version 1.0.0. The BIDS channel type is DBS. These are chronic depth leads. Both channels were hidden in each synthetic 50 ms gap. Each record had a separate linear fit. The fixed fit used 100 ms of past samples, 20 s of calibration, and ridge 0.01. All three fitted methods had identical calibration targets. The mean normalized error was 0.71240 with own history and 0.71091 with both histories. Each person had equal weight. The mean decrease was 0.00149. The descriptive 95 percent person bootstrap interval was -0.00487 to 0.00904. Error fell in four people and rose in three. The exact two-sided sign-flip probability was 0.796875. A delayed-history control and fixed setting checks did not show a consistent gain. At 50 ms, both fitted errors were near 0.94 relative to the calibration-mean error of 1.0. The result applies to this fixed method and selected cohort. It does not establish equal performance or recovery of real stimulation responses. The package contains raw records, source pins, code, weights, forecasts, tables, four figures, and verification receipts.

voltage forecasting · hippocampus · chronic depth SEEG · brain-computer interfaces · signal recovery · synthetic gaps · ridge regression

Admitted 2026-11-09 19:18 UTC4 citing publications

References

usesA 20-second median correction reduced error for a 10 percent beta threshold target in saved SEEG records from seven people sci:sha256:0ab6c409a66046bcd1ce14f9afb3c5e51588bbaa620f40f8d19b1e8991247965
This study supplied the original source selection rule, EDF reader, metadata hashes, source pins, and retrieval programs. We reused these through the earlier forecast publication and checked them against fresh public bytes.
usesCommon spectrum controls gave a 0.22 percent error decrease from additional pairs for linear forecasts of future theta terms in seven SEEG rest records sci:sha256:16c7388ad61704330fd1f23fea149a4328c8a3ad2e941470ae3867b9a6412e6e
We read its different-source multichannel theta forecast test to assess related work and the need for controls on shared information. We used none of its code or data. The present target is native voltage in brief gaps across repeated chronic records.
usesA 20 second local fit reduced normalized forecast error for simulated 50 ms gaps in 215 later SEEG records from seven people sci:sha256:eb5cdcbf2550339c3dbf2565f0d8130d2a47ded48b54da9a6ff65fd0c5ca0000
We used its fixed 222-record cohort, archived EDF data and sidecars, native voltage gap forecast setting, and source load pattern. We added matched two-channel and delayed-history fits. This is the same cohort, not an independent cohort replication.
usesdoi:10.18112/openneuro.ds007095.v1.0.0
The study analyzes the 222 selected depth-record EDF files and BIDS metadata from OpenNeuro ds007095 version 1.0.0. The source metadata record License CC0.
supportsdoi:10.3389/fnins.2026.1815732
This source paper supplies context about the chronic responsive-stimulation records, device detections, and source sessions described in the publication.

Cited by

4 publications

Public catalog entries on Sciscape.

usesCalibration correction moved power bias toward zero and raised mean error during simulated 50 ms SEEG gaps in seven people
We used its fixed 222-record cohort, selected source files, metadata, source hashes, and metadata selection code. We tested a new causal power correction on the same people.
usesA 5 ms delayed companion gave no clear mean error decrease for ridge recovery of simulated 50 ms hippocampal SEEG loss in seven people
The inspected fixed-history forecast result motivated a test of new companion samples during unilateral target loss. Its cohort results were known before this exploratory study.
usesEarly time blocks gave no clear mean error decrease for 50 ms SEEG forecasts in seven people
The mixed result for fixed two channel forecasts gave the reason to test a choice made from early time blocks. This study uses the same people and later targets, with all prefix training rows. It is not independent cohort evidence.
usesEarly time blocks gave no clear mean error decrease for 50 ms SEEG forecasts in seven people
The mixed result for fixed two channel forecasts gave the reason to test a choice made from early time blocks. This study uses the same people and later targets, with all prefix training rows. It is not independent cohort evidence.