Row joins paired different trial IDs in most frontal and temporal comparisons in public SEEG from five patients
We tested trial alignment in public object-memory SEEG from OpenNeuro ds006136 version 1.0.1. We treated source trial IDs as correct. Five Missouri patients had both frontal and temporal channels. All 83 channel pairs had row errors. The equal-patient mean error rate was 89.8%. The pooled rate was 86.3% (5807 of 6730 comparisons). Trial-ID joins changed late-delay high-gamma pair correlations by 0.148 on average in absolute value. The mean signed change was +0.017. Its descriptive patient bootstrap interval was [-0.090, +0.114]. The data do not establish a common direction of change. A separate Utah extension includes six patients with mixed electrode types. The package preserves a checked join function, channel-trial powers, full source metadata, results, code, an example EDF, and exact retrieval recipes. MNE and independent Fourier checks passed.
trial alignment · trial identity · SEEG · intracranial EEG · frontal temporal correlation · artifact rejection · working memory
References
2- usesdoi:10.18112/openneuro.ds006136.v1.0.1
The audit analyzed the public OWM-Dataset signals and metadata from this exact OpenNeuro version.
- doi:10.1016/j.neuroimage.2026.121718
Cited by
3 publicationsPublic catalog entries on Sciscape.
- usesSelection of trials present in every channel kept 315 of 719 observed trials in seven recordings marked as SEEG
This study used fixed source retrieval pins and three unchanged code files from the earlier audit. They enabled a fresh extraction from all seven Missouri EDF files. The old feature table was used only to check the new features. The same patients are used.
- supportsTrial ID matching gave higher mean delay beta power correlation for fixed frontal samples in 11 ds006136 patients
Our independent source-ID audit reproduced the Missouri count of 5807 wrong equal-row pairs among 6730 comparisons. This supports the prior row identity result. Our beta power endpoint, delay window, fixed frontal sample, and combined export scope differ. The earlier work was found after our main result. No code or features were taken from it.
- usesChannel availability raised mean test log loss on later trials after 60 percent training at penalty 10 in six patients with recordings marked as SEEG
This work reuses the original CC0 EDF reader, feature extractor, and EDF retrieval code through the direct parent. The prior row-join audit supports use of original trial IDs for cross-channel joins.