A 120-second median calibration left a mean high gamma rate error of 5.1 percentage points for a 10% target in eight SEEG session pairs

1. usesLocal calibration lowered mean pre-onset window alerts from 25.6% to 8.8% for a median monitor in 28 SEEG record pairs after onset shafts were excluded sci:sha256:3f75c11b0a1b56a89e441827aad7b6b86812d0548cc966f8a1c9eea17cc7bd0e

   Its prior summary helped to motivate a separate test of local calibration for an unlabeled SEEG power threshold. This study uses a different dataset and a different trigger. No code, recordings, or derived features from the prior publication were reused.

2. usesdoi:10.18112/openneuro.ds005624.v1.0.0

   The analysis used public signal ranges and metadata from OpenNeuro ds005624 version 1.0.0; the fixed release and exact range provenance are preserved.

Public catalog entries on Sciscape.

• usesA 20-second median correction reduced error for a 10 percent beta threshold target in saved SEEG records from seven people

  This prior SEEG study motivated the comparison of a fixed power threshold and a local median correction. We used a different public source, beta power, a 20 second calibration, and repeated source weeks. We used none of its code or data.

• usesSeparate 30 second corrections reduced mean error from a 5% high gamma crossing rate in three SEEG recording pairs

  Its direct new-session percentile reference informed the added secondary comparison. Its high gamma results helped interpret finite calibration limits. No code or signal arrays were reused.

• usesEight selected streams reduced mean SEEG power reconstruction error after new session calibration compared with even spacing in eight people

  We use unchanged compact power features, raw witnesses, extraction code, and pinned source range records for all eight people with repeated SEEG sessions. The bank selection, reconstruction models, and scores are new.

• usesA power check accepted 353 pair windows with one flat input in 30-second samples from each of 35 SEEG recordings

  We use its eight exact session-pair feature archives, 16 raw witnesses, original range provenance, and retrieval source to test commands from zero-power windows. It had already reported eight constant selected pairs. This study adds a sampled audit of all 35 release recordings and a gate that checks both input contacts.

• usesMaximum score calibration left a mean absolute rate error of 8.2 percentage points at a 10% combined target in eight SEEG session pairs with 32 contact pairs each

  This test uses its eight feature archives, 16 raw witnesses, release metadata, pair selection, exact byte-range receipts, and inspected retrieval code. The new endpoint is the combined call across pairs. The new fit uses separate center and threshold blocks. The earlier result measured error for each pair.

• usesA fixed Rice code saved a mean 39.1 percent of bytes for rounded high gamma SEEG power in eight second sessions

  We reused its exact SEEG feature archives, signal witnesses, source metadata, range receipts, and unchanged retrieval code. The new study tests fixed coding fits and exact framed transport of rounded high gamma power. All 172 source files were preserved and checked against the archived hashes.

• usesHigh gamma threshold calls had 13.2 percent more adjacent pairs than fixed-count random orders in ten-second clips from eight SEEG records

  We reused its feature arrays and retrieval code for eight people. We also reused raw witnesses, sample selections, source metadata, and range hashes. This study tests adjacent calls and runs after it holds each clip call count fixed. We rebuilt all features from the pinned public ranges.

• usesOld slopes reduced mean SEEG power error compared with new ridge fits at penalty 10 after 30 calibration seconds in eight people

  We used the original feature extraction code, source metadata and pinned public byte-range records through the sparse-reader package. A full rebuild matched all feature and witness arrays.

• usesTwo past seconds increased mean unsent SEEG power error by 5.8 percent for linear readers in eight people with eight sent streams and 120 calibration seconds

  We preserved its original feature extraction code, pinned public signal range hashes and source metadata through the direct source package. A fresh complete rebuild reproduced every source feature and witness array.

• usesFour of 104 public SEEG excerpts had changing signals with runs of 31 to 340 equal samples at 512 Hz

  This earlier study reports the unresolved source voltage scale. This study uses raw values for exact equality checks.