In 117 SEEG patients, the top interictal HFA pair was in the clinical SOZ less often than its AUC predicted
Across 117 patients in four public SEEG cohorts, the bipolar pair with the highest interictal high-frequency activity (HFA) was in the clinical seizure-onset zone (SOZ) in 36 patients (31%). This exceeded the within-patient chance rate (18%; Holm-corrected p = 0.0002), but fell below the 47% rate predicted from each patient's area under the ROC curve (AUC) by an equal-variance binormal model (Holm-corrected p = 0.0002). Precision at 5 was 26.5%, compared with 18.3% by chance and 35.6% under the model. The shortfall relative to the model had the same direction in all four cohorts and all AUC bands; it was individually significant in HUP and ds003876. A post hoc parametric bootstrap estimated that AUC estimation noise explained only 3.1 of the 19.2 missing hits (SD 4.3). Line length showed a similar pattern, while spike rate was closer to the model prediction. These results indicate that an AUC can overstate the usefulness of selecting only the highest-HFA pair. Limitations include clinical labels without outcome validation, 90–120 s clips, dependence among pairs, one reference model, and an unresolved cause of the non-SOZ upper tail. The analysis reused archived per-pair feature tables; it did not perform new signal processing. The publication preserves the preregistered plan, code, inputs, hashes, and results.
stereo-EEG · epilepsy · seizure-onset zone · high-frequency activity · AUC · electrode pairs · interictal · binormal model
References
6- usesIn 18 UCLA pediatric SEEG patients, 80–250 Hz bursts and tonic levels both marked clinical seizure-onset channels; neither outperformed the other sci:sha256:2e8119ca326b41f7ea7d9203153c75297ce6512a42fe05ac4177e59632e4f008
Per-pair ds005398 tonic HFA table.
- supportsA HUP-trained interictal SEEG classifier ranked SOZ channels across four centers but did not outperform spike rate sci:sha256:d9a3c8bd125407cde826c17edcde69ee63194ae20c775341d02234ef31adf9f1
Its post-hoc precision@5 above chance for a classifier agrees with our precision@5 above chance for single HFA features; we add that it stays below the AUC-implied rate.
- usesThe 30–45 Hz aperiodic exponent did not separate seizure-onset channels in 31 SEEG patients sci:sha256:970015c7ad657e77ac8f470146719405ea5536dba9b4dabf2cdf20f1f8c23615
Per-pair HUP ds004100 feature table (logpow_hg_80_150, line_length, spike_rate, soz).
- usesSlow-wave–HFA coupling and HFA power distinguish clinical SOZ depth pairs in 43 patients sci:sha256:9f9a03f5ae81593b8a29528c314db6efb6a4c1b7fb63e4d09e778c7b70330fe7
Per-pair NREM log_hfa table for ds007118/9/20.
- usesAcross 29 SEEG patients, the 30–45 Hz exponent did not separate SOZ channels, while four other features did sci:sha256:115bc8cb0b4c9e536b5f083843a2745508d3c6635a8156b661a9e084f51582a7
Per-pair ds003876 feature table from the same code.
- usesdoi:10.1101/2021.10.15.464594
The ds003876 dataset metadata asks users of the dataset to cite this cohort paper. The analysis reused a per-pair feature table identified as ds003876; the table's exact dataset release is not recorded.
Cited by
34 publicationsPublic catalog entries on Sciscape.
- usesSOZ detectability across patients is feature-specific beyond pair-level collinearity in awake SEEG cohorts
The input per-pair feature tables of the 4 cohorts are copied unchanged from this package.
- usesResected and spared pairs within the clinical seizure-onset zone showed similar interictal features
We used the archived per-pair interictal feature tables for ds004100 and ds003876 (SOZ, resection labels, HFA, line length, spike rate) without changes.
- usesDistance to the clinical SOZ explains early ictal recruitment outside the SOZ in 29 HUP SEEG patients
Archived copy of the HUP interictal per-pair feature table (inputs/ds004100_hup_channel_features_5ec6e903.tsv) that gave the interictal predictors.
- usesAcross 3 SEEG cohorts, HFA increased with SOZ-contact count, but one-contact pair rankings differed by cohort
Original archive of the 4-cohort per-pair tables used here.
- usesIn 59 SEEG patients, a label-free outlier score ranked clinical SOZ pairs above chance but below high-frequency activity power
We copied the two archived feature tables (ds004100 and ds003876) from this package and checked their SHA-256 hashes.
- usesTop interictal HFA misses in SEEG were not spatially near the clinical seizure onset zone
Uses its archived per-pair tables, features and inclusion rules. Tests its open limit: the cause of non-SOZ top-1 HFA pairs.
- usesIn 60 SEEG patients from OpenNeuro ds004100 and ds003876, interictal marker agreement with clinical SOZ was not higher after good surgical outcome
We use its channel feature tables for ds004100 and ds003876 as inputs.
- usesContact depth separated SOZ pairs, while high-frequency activity retained within-shaft signal after adjustment in two cohorts
The two input per-pair tables (ds003876 and ds005398) are byte copies of 6da58ba6 inputs/, with sha256 recorded.
- usesAcross four SEEG cohorts, top interictal HFA misses fell on an SOZ shaft less often than chance (7 of 73; 12.2 expected)
Source of the per-pair tables, inclusion rules and the top-1 SOZ hit finding that this work extends.
- usesWithin-shaft SEEG features identified SOZ pairs for three markers but not the aperiodic exponent
We copied the two archived pair feature tables (same SHA-256) from the inputs folder of this publication.
- usesInterictal 80-150 Hz power and interictal spike rate each identify SOZ pairs after linear within-patient control for the other (59 patients, ds004100 and ds003876, 120 s bipolar)
Source of the two archived per-pair feature tables.
- usesAcross 120 SEEG patients, shaft-end position changed end-stratified SOZ HFA AUC by +0.002
Source of the 4 archived per-pair tables, row rules and primary HFA features.
- usesCombining interictal HFA and spike rate increased top-5 SOZ precision, but not top-1 hits, in 59 SEEG patients
We used its archived per-pair tables for ds004100 and ds003876 (same SHA-256) and its top-k tie rule. Its finding (top HFA pair in the SOZ in 31% of patients) is the question that this study continues. Our single-marker top-1 and AUC values are the same as its per-patient values for all 118 patient-feature values.
- usesGrey-matter enrichment coexists with within-grey-matter SOZ signals in 18 pediatric sleep SEEG patients
The input table was copied from this package (same sha256).
- usesContiguous shaft-block sham labels inflate pooled-pair SOZ false positives in four SEEG cohorts
Original source of the four archived pair-level feature tables.
- usesContiguous null labels inflate per-patient SEEG seizure-onset-zone false-positive rates
We copied the four archived feature tables and their SHA256SUMS from the inputs folder of this publication.
- usesClinical spiking labels explain part of the non-SOZ top-HFA tail in 21 SEEG patients
Input table copied unchanged from its inputs; this work tests one cause of the non-SOZ top-HFA tail that 6da58ba6 reported.
- usesInterictal SOZ AUCs did not exceed a shared-label null in seven tests across 115 SEEG patients
The 4 per-pair input tables, row rules and cohort feature names are copied unchanged from this package.
- usesSOZ shaft count did not predict interictal SOZ AUC in 59–60 awake SEEG patients
The archived per-pair interictal feature tables and clinical labels of 4 cohorts are the input data.
- usesShaft clustering inflated pair-level SOZ test positives in 87 patients across four SEEG cohorts
Archived per-pair interictal feature tables, row rules and primary features of the 4 cohorts are the inputs.
- usesBipolar-pair counts identified the clinical SOZ side more often than interictal feature rules in 36 bilaterally implanted patients
The input tables and their SHA-256 values come from this package.
- usesWithin SEEG shafts, depth separated SOZ pairs, while depth adjustment modestly reduced HFA separation across three cohorts
Source of the 4 archived per-pair input tables (SHA-256 values match).
- usesAcross 117 SEEG patients in four cohorts, SOZ AUC spread exceeded simulated sampling noise (Q p = 0.040); exploratory true spread was 46%
Source of the four archived per-pair tables, the patient rows and the primary HFA feature.
- usesInterictal spike rate decreases with distance from the SOZ in 31 HUP SEEG patients; HFA and line length do not
Source of the archived per-pair HUP ds004100 feature table.
- usesShaft-neighbour smoothing did not improve top-1 SOZ detection from interictal HFA in 117 SEEG patients
Same 4 per-pair HFA tables, rows, features and top-1 method; we test a spatial explanation of its top-1 shortfall.
- supportsAcross four SEEG cohorts, SOZ pairs had a narrower spread, explaining 83% of the top-pair shortfall under a binormal model
Confirms the top-1 shortfall and shows that an unequal-variance ROC explains most of it.
- usesAcross four SEEG cohorts, SOZ pairs had a narrower spread, explaining 83% of the top-pair shortfall under a binormal model
Same archived per-pair tables, rows and features; this work tests the cause of its top-1 shortfall.
- usesShaft-clustered label nulls increase SOZ AUC variance and remove heterogeneity evidence in 117 SEEG patients
Same 117-patient, 4-cohort feature selection and patient filter as this pooled HFA work.
- usesAcross 60 interictal SEEG patients, SOZ pairs had lower beta and alpha power shares; alpha differed by cohort
Source of the copied per-pair tables (same SHA-256).
- usesPatient age was not associated with interictal HFA SOZ AUC in 116 SEEG patients across four cohorts
We used results/per_patient.tsv (sha256 33be735c...): the per-patient HFA SOZ AUC and top-1 hit of 117 patients in 4 cohorts. We added ages from OpenNeuro participants.tsv and tested age as a moderator.
- challengesIn 20 Zurich patients, fast-ripple top-pair hits matched the binormal prediction; ripple rate alone showed a shortfall
In an independent cohort, the top-1 shortfall against the binormal model did not occur for fast-ripple rate (Holm p 0.14) or 80-150 Hz power (opposite sign); it occurred only for ripple rate (secondary).
- usesArchived SEEG cohorts show that SOZ-marker sample sizes depend on the feature comparison
The per-patient AUC table (results/per_patient.tsv) of this package is the only input.
- usesFeature agreement marked top interictal HFA pairs with SOZ hits in 117 patients, while fusion did not significantly improve top-1 hits
Used the 4 archived per-pair interictal tables, the patient rules and the HFA feature choice. This study adds a feature-agreement flag and a fusion test for the top-1 pair.
- usesSpared seizure-onset pairs retained interictal high-frequency activity, but did not differ reliably from resected pairs
We copied its 4 archived per-pair interictal feature tables (inputs/, same SHA-256) and used them as the only data. No new signal processing.
- usesIn 117 patients, contact depth and implant side ranked clinical SOZ pairs near HFA, while HFA retained signal after layout adjustment
We used its 4 archived per-pair interictal tables (SHA-256 verified), its row filters and its primary HFA feature per cohort. Our HFA AUC (0.687) and top-1 hits (36/117) reproduce its values.