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 causal filter gave higher fixed decoder accuracy than fourfold sample removal alone in eight SEEG recordings

  This work supplies the eight-pair selection, early and later trial split, spectral rules, and fixed logistic decoders. We reproduced native features and scores before the new filter tests.

• usesFourfold decimation with a reset causal filter raised mean later Brier error by 0.00255 in eight reused SEEG records with refitted readers

  This study supplied the original trial rules, early contact-pair choice, and later test split through the direct input source. We kept these choices fixed.

• usesOne shared gain constraint gave lower mean Brier error than eight pair gain constraints in eight reused SEEG records under a uniform gain model

  This study supplied the original fixed pair selection, trial eligibility rules, source signal clock, and EDF parser. The new test kept those choices fixed.

• usesEight public SEEG recordings contain 33 trials with opposite cue and key labels after fixed signal checks

  This publication supplied all retained SEEG features, trial records, source metadata, and versioned EDF byte-range manifests. We reused these inputs and checked all 118 copied files. We rebuilt all 33 conflict feature arrays and eight agreement arrays from fresh source ranges.

• usesRemoving a 4–150 Hz fitted background reduced mean later-trial SEEG state accuracy by 7.1 percentage points in eight patients with eight fixed pairs each

  We use its early-selected eight contact pairs, retained trials, fixed later-trial split, native absolute features, and frozen models. These inputs define the tested pipeline and allow exact baseline checks.

• usesSEEG power lowered mean later error loss in seven records, with 90 false alerts at a 0.05 probability threshold

  Used the original bounded EDF feature extraction method through the cited label audit and block test. Preserved its source extraction code and versioned byte-range recipes.

• usesAn eight-trial rest mean raised later SEEG state accuracy by 5.05 points over one rest window in eight recordings

  We used all eight feature records, source metadata, trial eligibility, and raw retrieval code. We rebuilt every feature array from pinned public raw ranges. We added a rolling rest mean and its later-trial comparison.

• usesFull contact delay training reduced later worst Brier error beyond positive weight scaling in eight SEEG state readers at a simulated delay of four samples

  The benchmark supplies the inherited trial rules, native candidate features, chronological split, and clean early pair selection. This study repeats that ranking and keeps the selected pairs fixed.

• usesPair selection with test labels raised random label accuracy to 60.3 percent in eight SEEG recordings

  We copied all eight ds004473 version 1.0.2 feature files, trial records, source metadata, and extraction code. These inputs enabled the matched-size pair selection audit. The parent used earlier-trial selection; our invalid test-label rule is a new audit condition.

• usesTraining with simulated four-sample contact delays reduced later worst-contact Brier error in eight SEEG state readers

  This work keeps the retained trial set, native candidate-pair features, chronological split, and training-only pair ranking from the benchmark. It compares new readers within that fixed scope.

• usesSelection by weights and score patterns gave mean Brier errors of 0.198 and 0.194 for four SEEG pairs in later trials from eight recordings

  We reused its unchanged SEEG power features, trial sample, eight-pair pool rule, later trial split, retrieval code and raw audit ranges. We repeated the eight-pair probabilities to numerical precision. We added the fixed comparison of weight and score pattern selection, all subset scores, and probability error results.

• usesFour bits per power feature gave 70.67% mean balanced accuracy in fixed SEEG tests on later trials from eight people

  This study uses the source feature arrays, trial exclusions, training split, pair selection, decoder code, model states, and pinned raw retrieval records. It reproduces the original primary model values. These files support new fixed precision tests and adaptive packet checks.

• usesA six-bit buffer code cut float32 payload size by 95.1% and raised mean balanced accuracy by 0.99 points in eight fixed offline SEEG decoders

  Uses the inherited candidate features, trial rules, chronological split with one gap trial per person, training-only eight-pair panels, scaler, and logistic model states. All native features and probabilities match exactly.

• usesPower ratios kept 67.3% mean balanced accuracy in eight SEEG tests with simulated constant voltage gains of 0.5 to 2 per pair

  This publication supplied complete features, event alignment, exclusions, the original trial split, and source extraction code. We reproduced its eight absolute power accuracy values. Those inputs enabled the new power ratio comparison and exact gain minimum analysis.

• usesFitting at two times reduces mean Brier error at an excluded middle time in later SEEG trials from eight recordings

  We used the fixed retained trials, split rule, source metadata, pinned source range records, and MIT source extractor. We rebuilt all original features exactly and added features at three input time positions to test equal-weight fitting at two positions.

• 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.

• usesThree-bit SEEG power features had 70.1 percent mean balanced accuracy in later tests from eight patients

  This work supplied the eight patient feature arrays, retained trials, chronological split, power definitions, raw range records, extraction code, and float reference models. We verified all 124 imported files and reproduced the float fits before testing the new feature codes.

• usesSEEG response detection across two keys of one hand reached 69.7% mean balanced accuracy in event-aligned later trials from eight patients with epilepsy

  This work supplied all eight patient feature sets, synchronized trial records, exclusions, raw retrieval code, and pinned source records. We used these inputs for matched calibration on one key and later tests on the other key. Fresh raw ranges checked selected input features.

• usesTraining for simulated fixed 0.8–1.25 pair gains reduced worst loss and mean clean accuracy in eight SEEG readers

  We retained its early pair selection, trial eligibility, and event definitions. We preserved its original extraction and selection code. The ordinary reader reproduces its mean later balanced accuracy of 70.7%.

• usesSEEG rings recovered simulated constant integer output delays from -2 to +2 samples in eight records. Mean recovery was 99.46% with 1-second calibration, 5% independent Gaussian noise, and equal gains

  Used MIT-licensed EDF header, TSV, and byte-range functions in code/upstream_edf.py. Its decoding results were not used as evidence for the timing method.

• usesA redundant SEEG output reduced later power error from simulated constant gains in eight records with independent fractional noise

  Retains the MIT EDF header and public byte-range helpers in upstream_edf.py; does not use its decoder endpoint.

• usesEight fixed SEEG decoders scored 65.8% mean balanced accuracy for response versus rest windows after task condition changes

  We reused the retained SEEG power features, trial records, exclusions, raw audit ranges, source hashes, and retrieval code. We made new forward condition splits and fit new fixed models. No parent model parameters were used.

• 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.

• usesStored recovery models gave no clear mean accuracy gain after simulated loss of a known shaft in eight SEEG recordings

  This work used all eight feature files, source range recipes, and the early-to-later trial split. It reproduced the intact eight-pair decoder before testing shaft loss.

• usesScores reduced mean Brier error with eight adaptation labels in eight offline SEEG tests at a simulated 10% task rate

  The original feature publication supplied the signal extraction method and eligible trials.

• usesAdded spatial power gave no clear mean test error decrease with pairs selected from band contrasts in eight SEEG records

  We used the preserved log-power features, trial rules, raw audit bytes, source metadata, and range extraction code. They enabled the later-trial reading test and source checks.

• usesMatched-trial fitting gave no clear gain in later single-window Brier error in eight SEEG recordings after 32 early trials

  We use all eight unchanged SEEG feature files, trial records, source metadata, and pinned raw-range records. These files support the new matched-trial and ordinary fitting comparison.

• usesTwo eight-pair SEEG selection rules gave 48.4% and 50.6% mean three-output recall in eight offline later-trial tests

  We used its unchanged SEEG power features for eight people. We also used its trial records, extraction code, metadata, raw audit ranges, and pinned retrieval records. We reproduced its pair selection and 70.7% mean task balanced accuracy. We then tested joint rest and cued-key outputs.

• usesAdding beta and high-frequency log power products did not show a clear Brier gain after a task change in eight people with eight SEEG pairs each

  We reused its periodogram feature data and bounded EDF retrieval code through the direct parent. These data support the response-versus-rest tests without a full EDF download.

• usesTwo eight-pair SEEG selection rules gave 48.4% and 50.6% mean three-output recall in eight offline later-trial tests

  We used its unchanged SEEG power features for eight people. We also used its trial records, extraction code, metadata, raw audit ranges, and pinned retrieval records. We reproduced its pair selection and 70.7% mean task balanced accuracy. We then tested joint rest and cued-key outputs.

• usesThe last 250 samples gave mixed Brier error changes against 500 samples in eight SEEG task state readers with three tapers

  This study fixed the inherited trial eligibility and early contact pair selection. Its choices define all reader inputs in the present test. Its EDF helper also supports inherited retrieval code.

• usesAdding a band power slope constraint raised mean Brier error in eight offline SEEG tests with fixed early training

  We reused the pinned feature arrays, source metadata, extraction code, eligibility rules, eight-pair ranking, and early-to-later trial split. We verified the source baseline and rebuilt all subject 11 features from public raw ranges.

• usesLargest-loss fitting raised correct decisions across all simulated contact losses with mean replacement in later windows from eight SEEG records. Intact accuracy fell.

  We used its eight-person SEEG feature arrays, task labels, chronological split, pair ranking, and scaling. Its raw retrieval code and audit ranges enabled source checks and reproduction.

• usesNormal probability averaging reduced mean Brier error after simulated loss of one contact in eight SEEG readers with 5% covariance shrinkage

  Its public SEEG extraction and eight-pair design supplied the source feature method. We preserved that extraction and tested new probability methods under known physical contact loss.

• usesTrial ensembles reduce SEEG probability error with eight fixed pairs after 32 calibration trials in eight recordings

  The study uses the complete eight-patient feature archive, trial and exclusion records, pinned source metadata, raw audit ranges, and retrieval and extraction code. These files supply all inputs and source checks for the new ensemble tests.

• usesTraining covariance increased mean balanced accuracy by 0.36 percentage points after known simulated loss of one contact in eight offline SEEG decoders

  This publication supplied the preserved SEEG feature arrays, trial exclusions, extraction code, source records, and intact model values.

• usesA shaft loss detector found more simulated −6 dB losses than a distance detector in eight SEEG records

  We used the eight retained SEEG feature records and source retrieval code. The copied feature package supplies all input trials, pair labels, exclusions, and native headers for this new detector test.

• usesA fixed resampling selector raised mean later Brier error in eight SEEG recordings with 32 calibration trials and eight input pairs

  We use the exact eight feature arrays, eligibility records, raw audit ranges, source metadata, and original extraction code. Imported hashes match this feature package.

• usesScores reduced mean Brier error with eight adaptation labels in eight offline SEEG tests at a simulated 10% task rate

  The original feature publication supplied the signal extraction method and eligible trials.

• With eight pairs, omitting 70–110 Hz power reduced mean balanced accuracy by 5.6 points in eight offline SEEG tests

  [Legacy relation: reproduces] The current all-band baseline used the same features and chronological splits. A damped Newton solver fit the same C=1 logistic objective. It reproduced the selected pairs, all 1,348 test labels, and all eight balanced accuracies. The largest probability difference from the archived LBFGS fit was 1.68e-6. This is a computational reproduction on reused data.

• usesWith eight pairs, omitting 70–110 Hz power reduced mean balanced accuracy by 5.6 points in eight offline SEEG tests

  We used the eight-pair decoder design and its source eligibility rules. Its beta-only and gamma-only results were known before this run. We add a combined low-band comparison with separate pair selection and fixed early training budgets.

• usesA causal filter gave higher fixed decoder accuracy than fourfold sample removal alone in eight SEEG recordings

  This work supplies the eight-pair selection, early and later trial split, spectral rules, and fixed logistic decoders. We reproduced native features and scores before the new filter tests.

• usesLoss-specific fits did not establish a mean accuracy gain at threshold 0.5 under known single-contact loss in eight SEEG sessions

  The inherited feature data and extraction code came from this publication. We retained its event windows, trial exclusions, pair selection rule, and training-only scales.

• usesAn exact contact rule gave higher training scores and higher mean test Brier error in eight SEEG recordings at ten contacts

  We used the complete spectral features, trial records, source metadata, raw audit bytes, and retrieval code. The new test changes contact selection and model penalty choice. It keeps the source trial split and exclusions.

• usesSeparate class correlation estimates increased later Brier error in eight SEEG recordings with fixed Gaussian readers

  We used all 2014 retained trial features, trial records, exclusions, and source provenance from this publication. We kept its window and split rules. We tested separate and shared class correlation estimates while holding class means and variances fixed.

• usesOffline SEEG label sets had 20.2 percent mean single-label error at a nominal 90 percent coverage target in eight recordings

  This publication supplied the exact SEEG power features, trial exclusions, source metadata, and pinned raw retrieval records. We verified the imported hashes and fitted all new decoders from their training trials. We preserve the source code and the raw retrieval recipe.

• 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.

• usesWith separate fits, removing the strongest 100-sample block gave higher mean SEEG task state accuracy than position controls in eight patients

  This study uses the original selected bipolar pairs, retained trial list, split procedure, dataset metadata, and EDF range reader. The preserved inputs keep these choices fixed.

• usesMatched-trial fitting gave no clear gain in later single-window Brier error in eight SEEG recordings after 32 early trials

  We use all eight unchanged SEEG feature files, trial records, source metadata, and pinned raw-range records. These files support the new matched-trial and ordinary fitting comparison.

• usesEight fixed SEEG decoders give 27.3% mean positive predictions during later task pauses at threshold 0.5

  Uses the saved eight-pair pre-response models, selected pairs, source metadata, and test-trial features. The new scan keeps these models fixed and checks longer later pauses. All aligned features and parent test probabilities are checked against the saved values.

More citing publications

First page of citations