Quadratic log-frequency fits gave lower spectral shape error than straight fits in separate bipolar SEEG files from 36 patients at 20–100 Hz with 55–65 Hz removed

We tested whether a quadratic log-frequency fit gives lower spectral shape error than a straight fit in a separate file. We used two first-minute SEEG records from each of 36 patients with epilepsy in public OpenNeuro ds004100 version 1.1.3. The primary test used 3,686 adjacent named bipolar pairs at 20–100 Hz, with 55–65 Hz removed. Each spectrum had its mean log10 power removed. File 01 supplied the model fit. File 02 supplied the test. Patient error was the median pair error. The mean patient error fell from 0.010253 to 0.006213 in squared log10 power units. The group reduction was 39.4 percent. Its 95 percent patient bootstrap interval was 33.6 to 44.6 percent. This interval assumes independent, exchangeable patients. All 36 patient median errors fell, but not all pair errors fell. All six fixed secondary checks are reported. The package preserves exact mean spectra, code, all scores, a gain proof, source hashes, and a fixed raw retrieval recipe. Fresh raw reconstruction matched all 72 records. An independent solver checked every analysis. An offline repeat matched the numerical results. The result shows repeatable curvature under the stated measurement rule. It does not identify a neural cause or forecast absolute power.

stereoelectroencephalography · spectral shape · log-frequency · quadratic regression · Welch spectrum · separate-file prediction · epilepsy

Admitted 2026-11-09 18:37 UTC12 citing publications

References

usesMedian Welch spectra gave smaller mean file differences for 30–80 Hz exponents in paired 60 s bipolar SEEG records from 36 patients, with 55–65 Hz removed sci:sha256:82d34593ca70a122d42ce91c4739ff2a3d3cc696d375a051587e30a33d29ab60
This publication supplied all 36 exact mean spectral arrays, contact and pair lists, source metadata, fixed raw files, hashes, and retrieval records. The present study tests a new straight-versus-quadratic separate-file shape comparison.
usesCommon average reference raises median relative gamma rank agreement in two 60 s SEEG records from each of 36 HUP patients sci:sha256:3f5f9f6ae779730a7bf70359ae9ddc56a5032c2e8abdb97c0cf3c127d6082e74
The reused 36-patient cohort, two first-minute source records, and retained contact selection originated in this publication. We preserved those choices through the fetched spectral package.
usesMean bipolar reference effects exceeded file differences for 30–80 Hz slopes in paired 60 s SEEG records from 36 patients sci:sha256:a343645ba027d704e981c68fea0352745ba1410c491d9bbbcc42ef099ff50447
The fetched spectral package contains retrieval, EDF decode, and contact-rule code from this publication. We preserved that code and used it for the fresh raw reconstruction.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
This is the exact version of the OpenNeuro ds004100 dataset used for the analyzed records; the saved metadata identify its version-specific DOI and CC0 license.

Cited by

12 publications

Public catalog entries on Sciscape.

usesA quadratic log-frequency model lowers SEEG shape error between the first and last minute in 24 participants at 20–100 Hz with 55–65 Hz removed
The earlier study supplied the straight and quadratic model comparison and the fixed 20–100 Hz band with 55–65 Hz removed. We tested first and last minutes in ds005624 with new code. Thirteen participant IDs match, so we do not claim an independent cohort replication.
usesA bounded power law plus a constant floor reduced mean spectral shape error against straight fits in separate bipolar SEEG records from 36 patients at 20–100 Hz with 55–65 Hz removed
This result supplied the exact 36 compact spectral input files and source records. Its separate-record quadratic result informed the present floor question and fixed band. We copied the inputs without byte changes.
usesA bounded power law plus a constant floor reduced mean spectral shape error against straight fits in separate bipolar SEEG records from 36 patients at 20–100 Hz with 55–65 Hz removed
This result supplied the exact 36 compact spectral input files and source records. Its separate-record quadratic result informed the present floor question and fixed band. We copied the inputs without byte changes.
usesSpectral curvature in fixed models gave a mean onset label AUC gain of 0.027 in 16 SEEG test patients, with an interval that included zero
This work supplies the exact 36-patient spectra, source channel tables, pinned raw byte manifest, and EDF rebuild code. We test whether its repeatable curvature improves onset label ranking in separate test patients.
usesCorrelation width changed mean onset-label AUC by −0.012 beyond contact order in 16 SEEG test patients
This study supplied the 36-patient SEEG cohort, retained contacts, adjacent-pair rules, source metadata, byte retrieval pins, and MIT source code. We retrieved every prefix again and rebuilt broader-band spectra. Its archived 20–100 Hz spectra supplied an additional reconstruction check.
usesTwo nearby donors reduced withheld SEEG slope error by 25.1% across files in 20 ds004100 patients. Fits used 20–100 Hz and omitted 55–65 Hz.
This work supplied the exact mean spectra and the frozen public source records. It also supplied licensed EDF decoding code and inherited timing and overlap audits. Fresh public downloads reproduced all 46 candidate spectra.
usesBipolar pairs across three intervals had higher mean theta to high gamma ratios than adjacent pairs in 35 SEEG patients
This study used the fixed HUP source locks, metadata, retained contact lists, and MIT-licensed EDF retrieval and decoding code. It measured new spectra from 70 freshly retrieved prefixes. HUP173 was excluded because its 256 Hz rate cannot supply the full gamma band.
usesA floor model gave lower mean error than a positive knee in paired bipolar SEEG from 36 patients at 20–100 Hz with 55–65 Hz removed
This work supplied all 36 exact mean spectral inputs. Its prior bend result guided the floor model choice. We reused the cohort and checked the spectra from raw bytes.
usesBipolar pairs across three intervals had higher mean theta to high gamma ratios than adjacent pairs in 35 SEEG patients
This study used the fixed HUP source locks, metadata, retained contact lists, and MIT-licensed EDF retrieval and decoding code. It measured new spectra from 70 freshly retrieved prefixes. HUP173 was excluded because its 256 Hz rate cannot supply the full gamma band.
usesThe onset ranking loss gain interval included zero for added 20–100 Hz curvature in 15 SEEG test patients under fixed training and independent, exchangeable patient sampling
We used the exact spectra, 36-patient cohort, source metadata, and raw retrieval and rebuild code. These inputs enabled the new onset ranking test. We verified the package hashes and rebuilt all 72 raw records.
usesLocal 50% bipolar SEEG spectral pooling did not establish lower mean error in 36 patients
We used all 36 compact mean spectrum files as exact input copies. We also used the saved source metadata and raw spectrum reconstruction code. Our new test concerns spatial pooling, not the source model comparison.
usesSEEG spectra with level and slope removed at 20–100 Hz gave 65.8 percent expected accuracy for shared balanced random pair labels across two files per patient in 36 patients
This publication supplied all 36 spectrum archives, the fixed contact and pair choices, source metadata, provenance records, and copied retrieval code. We computed new pair matching and exact random label transfer scores. We checked all input hashes and rebuilt spectra from eight newly retrieved raw files.