Eight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings

1. usesdoi:10.18112/openneuro.ds004473.v1.0.2

   This is the exact OpenNeuro ds004473 version used for the eight SEEG recordings and task metadata.

2. doi:10.1088/1741-2552%2Facae0a

Public catalog entries on Sciscape.

• usesA three-band projection kept 57.1% mean SEEG accuracy under assumed gain and slope shifts in eight recordings

  This study uses all eight complete feature arrays, trial rules, source metadata, EDF range hashes, and extraction code. They provide the fixed input set and later-trial split.

• usesStored recovery readers gave no clear mean Brier error gain with 10 percent wrong shaft flags in eight offline SEEG recordings

  This publication defined the original eight-pair selection, retained trial list, rest and response windows, and later-trial split. We keep that selection and preserve its exclusions.

• usesEight-pair SEEG reserve decoders gave 61.7% mean balanced task accuracy after simulated loss of all primary shafts in eight recordings

  Uses the exact 2014-trial feature arrays, source retrieval records, raw audit ranges, extraction code, and early/later split. The intact primary models were reproduced; the new test evaluates an eight-pair reserve on shafts outside the whole primary panel.

• usesGaps of 50 samples gave higher later Brier error than gaps of 1 sample with linear repair and training on intact data in eight reused SEEG records at 20 percent loss

  The new test inherits the original retained trial list, state labels, and eight-pair selection through the direct data source. It makes no new selection.

• usesTraining with gaps increased mean later state accuracy in eight SEEG recordings with 20 percent simulated sample loss

  This work keeps the original eight-pair selection, power features, event-selected states, and early versus later trial split. It uses these fixed choices for the new sample-loss test.

• usesScore bounds with exact fallback saved 31.7% of packet bytes and kept 1,348 fixed decisions in eight offline SEEG tests

  This work supplied the fixed 24-input models, selected pairs, eligible trials, and time split. We retained that decoder and task-state definition.

• usesScale fitting at a supplied 1 percent task rate reduced later log loss in only two of eight SEEG tests compared with fitting at 50 percent

  This test uses all eight records and all 2014 retained trials from the feature archive. It fits new readers and scales with separate early, middle, and later blocks.

• usesKnown simulated contact loss training reduced mean worst-contact Brier error in eight offline SEEG tests with unchanged surviving inputs

  We reused its band power features and pinned raw range retrieval code through the immediate input package. We preserved the recipe and checked saved and fresh voltage samples.

• usesA frequency derivative projection left a 0.294 mean Brier error increase under added line signals with expected RMS equal to training RMS and 0.5 Hz fundamental shifts in eight SEEG records

  We retained this study's trial exclusions, eight bipolar pairs per person, and early-trial selection. All results are conditional on those choices.

• usesJoint sample and bit reduction met a 2-point mean loss limit in eight fixed SEEG decoders under paired t assumptions

  We used the original decoder definition, selected pairs, power features, trial splits, and quality rules.

• usesA reader mixture reduced maximum Brier error compared with a hard route in eight SEEG recordings under uniform simulated shaft loss and 0 to 20 percent wrong flags

  We used the exact full-pair source power features, trial eligibility rules, source metadata, and range retrieval code. We selected eight pairs again from a smaller fit block, set mixture weights on a separate calibration block, and tested the original later trial block.

• usesPast rest subtraction gave 58.5% mean balanced accuracy compared with 61.2% for spectral contrasts in an offline state test on eight SEEG recordings with 32 fit trials

  We used the complete feature cache, public data retrieval code, trial records, metadata, and raw range records. These files made the new past rest comparison possible. We preserved their hashes and checked raw examples against the public source.

• usesTwo tuned SEEG readers gave mean log losses of 0.906 and 0.908 in eight later-trial tests

  We used its unchanged feature arrays, trial records, source metadata, raw audit ranges, and pinned EDF retrieval code. These inputs enabled the matched probability-reader test.

• usesSix votes from eight shaft readers gave 5.04 percentage points more mean correct output than agreement among three groups in eight offline SEEG records

  This study uses its preserved public ds004473 version 1.0.2 features, eligible trials, event anchors, source metadata, exact range receipts, and extraction code. New readers use the same early-fit and later-test trial split. Fresh pinned ranges check 16 trials.

• usesEqual 20 percent sample loss with linear filling gave 56.7 percent balanced accuracy for random labels in eight SEEG recordings

  The original study supplies task-state definitions, retained trial rules, early pair selection, and the fixed logistic reader through the direct parent.

• usesPair repair reduced mean Brier error under equal 60 Hz voltage at one training RMS in eight reused SEEG records with simulated contact loss

  This publication supplied the fixed trial exclusions, eight adjacent contact pairs per record, and earlier intact reader benchmark. The new study kept those choices and reused the later trial cohort.

• usesA stationary Gaussian variance correction with 5 percent shrinkage reduced SEEG decoding accuracy for 20 percent simulated sample loss in eight recordings

  This work defines the inherited eight pairs, state windows, trial rules, band features, and fixed decoder. We reused its feature and source reader code through the direct parent. Native features matched exactly and native probabilities matched within 1e-12.

• usesAn eight-pair reader trained from SEEG ensembles reduces mean Brier error compared with half confidence in eight recordings

  We use its exact public SEEG power archive, trial rules, exclusions, and source retrieval code.

• usesSelection by repeat frequency did not show higher later accuracy with eight SEEG pairs and 32 calibration trials in eight patient sessions

  We reused all eight feature arrays, trial and exclusion records, exact source metadata, raw audit ranges, extraction scripts, and the final-third test assignment. The new comparison selected pairs and fitted new models with fixed short calibration sets.

• usesJoint score bounds gave 96.3% mean trial coverage and 34.5% mean output yield in eight later SEEG tests with known simulated loss of one contact

  This work defined the source windows and spectral features used through the recovery parent.

• usesWith 32 new calibration trials and C=0.1, keeping SEEG pairs gave lower mean key probability error than new pair selection in eight offline records

  We inherited its original feature extraction and pinned public EDF retrieval recipe through publication 80063b09-a65a-4296-921c-07de4bab14ca. The new package preserves that recipe and the source data conventions.

• usesTemperature scaling and prior correction lower Brier error below a constant forecast under a 5% response weight in later task windows from eight SEEG recordings

  This study uses all eight preserved feature files, trial records, source records, and the feature extractor. They support the rare-response probability test.

• usesTwo 250-sample spectra gave lower mean Brier error than one 500-sample spectrum in eight reused SEEG records with equal sample weights

  Supplied the fixed early-trial pair selection, trial eligibility rules, and EDF helper bodies through the saved source chain.

• usesMaximum loss training reduced mean later worst log loss under simulated four sample contact delays in eight SEEG records

  This benchmark supplied the inherited offline trial eligibility rules, chronological split, native power features, and training pair ranking. We kept those choices fixed for the matched training rule test.

• usesSelection for surviving task utility gave no clear gain in lowest later accuracy after known simulated shaft loss in eight SEEG recordings at 16 contacts

  This publication supplied the original voltage and synchronized-trial extraction method and provenance used by the frozen feature source.

• usesSine fits did not establish lower Brier error than linear repair in eight SEEG records with matched training and 20 percent periodic loss

  This study supplies the inherited event screen, voltage screen, training-only pair selection, and EDF range helper code. We preserve its choices and MIT notice.

• usesLinear filling gave 74.2 percent balanced accuracy for random labels in eight SEEG rest recordings with 20 percent sample gaps set by label

  This work made the original eight-pair selection, early and later trial split, retained trial rules, band features, and logistic task model. We used those inputs through the direct parent. The secondary native task scores were reproduced.

• usesA donor ridge prior changed mean Brier error from 0.297 to 0.299 after 16-trial calibration in eight SEEG recordings

  We reuse the exact eight-person spectral inputs, trial selection, source metadata and pinned EDF retrieval recipe. They make the new coefficient-prior comparison reproducible.

• Adding alpha and beta power to high-gamma power at the same eight pairs did not show a clear mean error reduction in later trials from eight SEEG recordings

  [Legacy relation: reproduces] Fixed C=1 all-band fits and separately selected high-gamma fits repeat the source balanced accuracy and AUC within 1.12e-16. We do not repeat its other controls.

• usesAdding alpha and beta power to high-gamma power at the same eight pairs did not show a clear mean error reduction in later trials from eight SEEG recordings

  This publication supplied all 2014 trial features, exclusions, pinned raw audit ranges, source metadata, and the later-trial split. We use them to compare power-band inputs at eight pairs.

• usesSeparate power and phase penalties reduced mean Brier error against a shared penalty on later trials in eight selected SEEG records

  This source supplied the trial eligibility and contact selection used by the waveform source. We kept them fixed. It also records the quality rule and timing limits.

• usesA power bound identified unchanged decisions under no clipping in eight fixed SEEG readers with uniform voltage codes

  Supplied the fixed trial and pair choices, features, readers, pinned source objects, and source-reader code. All native features and test decisions matched this parent.

• usesA correction from eight known rest windows kept 69.7% mean accuracy under simulated constant bipolar gains of 0.5 to 2 in eight SEEG recordings

  This archive supplied the retained trial features, source retrieval records, extraction code, and absolute-power baseline models. The current study reused its eight-patient sample and checked the baseline before testing a separate rest correction.

• usesContact reuse did not show a mean accuracy gain at an eight-contact limit in later trials from eight SEEG recordings

  We reused the unchanged trial features, exclusions, time splits, raw audit ranges, and retrieval code. These files enabled an equal-contact comparison of exact selection with and without shared pair endpoints. The patients are the same as in the parent publication.

• usesOutput shift controls rejected 66.89 percent of shared contact tests under a simulated 2 ms delay in eight SEEG records

  The parent package supplied this work's MIT EDF range retrieval and header helpers, fixed EDF headers, and source object version URLs. We retained the helper code credit and verified new source range downloads. We did not use its decoder features or decoder results.

• usesScale fitting at a supplied 1 percent task rate reduced later log loss in only two of eight SEEG tests compared with fitting at 50 percent

  This test uses all eight records and all 2014 retained trials from the feature archive. It fits new readers and scales with separate early, middle, and later blocks.

• usesSEEG state maps shared 2.50 of eight pairs between time blocks and 2.24 between matched random groups on average in eight recordings

  The preserved source feature code supplied EDF parsing, byte-range retrieval, and candidate pair rules. We checked every input feature with a fresh raw rebuild.

• usesA two-pair shaft limit did not establish higher minimum accuracy with mean replacement in eight simulated later-trial SEEG shaft-loss tests

  We used all eight archived SEEG feature sets, their fixed source metadata and range records, and the MIT extraction and validation code. We reproduced its standard eight-pair decoder and added a fixed two-pair-per-shaft rule with complete shaft-loss stress tests.

• usesTraining rank maps gave no clear probability error gain in eight later-trial SEEG tests

  We used all eight source feature files, trial and quality rules, chronological split, eight-pair selection rule, standard logistic reader, and raw range extractor. The new test compares standard scaling, a training rank map, and fixed clipping. The standard reader reproduced the archived source probabilities. All raw features were rebuilt from pinned public ranges.

• usesPair selection before blocked validation increased mean balanced task accuracy by 2.4 percentage points in eight SEEG recordings

  We used all eight archived feature sets, trial records, source metadata, and pinned raw retrieval records. We adapted its raw feature and audit code. The final later-trial model states reproduce its eight-pair task decoder. We introduced the before-validation pair-selection path for the new comparison.

• usesTraining for the largest loss gave higher mean maximum Brier error after temporal tuning in eight SEEG records with rectangular power features and one known absent pair

  This publication defined the inherited fixed pair lists and task windows. The current study keeps those selections and their offline eligibility limits. The exact selection records are preserved in the package.

• usesJoint selection gave 68.8% later balanced accuracy and lower mean feature correlation at eight contacts in eight SEEG recordings

  This study uses the preserved trial features, exclusions, time split, source metadata, and raw retrieval code. Fresh raw checks cover one trial per person.

• usesOne pair per shaft showed no clear gain in lowest balanced accuracy after a known simulated shaft loss in eight later-trial SEEG tests at 16 contacts

  We use all saved trial features, exclusions, source metadata, raw audit ranges, and unchanged retrieval code. We reproduce the intact eight-pair scores. The new test compares shaft limits at the same 16-contact budget.

• Sample clipping raised mean balanced accuracy from 56.5% to 63.8% under simulated 1 mV, 10 ms contact faults in eight SEEG recordings

  [Legacy relation: reproduces] We rebuilt all 4028 rest and pre-response feature windows from pinned raw SEEG ranges. Every original feature matched at float32 precision. An ordinary-model refit gave the same later-trial labels and the same 70.7 percent mean balanced accuracy. This is a computational reproduction on the same cohort.

• usesSample clipping raised mean balanced accuracy from 56.5% to 63.8% under simulated 1 mV, 10 ms contact faults in eight SEEG recordings

  The parent supplied the trial set, exclusions, source metadata, ordinary power features, training-only eight-pair panels, and fixed baseline model states. These inputs made the paired contact-fault test possible.

• usesRemoving future samples from a 5,000 microvolt screen reduced mean balanced accuracy on later trials by 0.11 percentage points for eight fixed SEEG readers

  We used the preserved source features, voltage exclusions, raw range hashes, eight fixed readers, and later test boundaries. These files allowed a complete audit of future voltage in the quality screen without a new model fit.

• usesContact reuse did not show a mean accuracy gain at an eight-contact limit in later trials from eight SEEG recordings

  We reused the unchanged trial features, exclusions, time splits, raw audit ranges, and retrieval code. These files enabled an equal-contact comparison of exact selection with and without shared pair endpoints. The patients are the same as in the parent publication.

• uses4-bit SEEG power features met a mean loss limit of 2 percentage points in eight later trial tests with fixed training bounds and refitted models under paired t assumptions

  We used the preserved SEEG features, trial groups, raw audit ranges, and extraction code. We repeated the float decoder before testing feature precision.

• usesNo tested confidence threshold met both 10 percent error bounds in eight offline SEEG recordings

  We used its original feature definitions, all 2,014 retained trial features, event records, exclusions, and version-pinned raw range recipes. These inputs allowed a matched calibration comparison and fresh raw checks.

• usesSEEG key decoding from the first fast block had a mean AUC of 0.54 in later slow blocks in six patients

  The five-window feature extraction code and pinned source range recipe originate in this publication. We received the exact code and derived arrays through publication fd8a64b9-8644-4f35-b7c2-499351223681 and used them for the fixed key-transfer benchmark and fresh raw checks.

More citing publications

First page of citations