sciscape Publications
← All publications
stat.ME · Methodology

Shaft-clustered label nulls increase SOZ AUC variance and remove heterogeneity evidence in 117 SEEG patients

In 117 SEEG patients from four cohorts, archived interictal high-frequency activity (HFA) features and clinical seizure-onset-zone (SOZ) labels were used to test whether spatial clustering of bipolar pairs increases sampling variance in patient-level SOZ AUCs. A prespecified clustered label null preserved contiguous SOZ run lengths and placed runs on shafts; independent-pair permutation and circular-shift nulls were also evaluated. The geometric-mean ratio of clustered to independent AUC variance was 1.82 (95% CI 1.70–1.95), above 1 in 109 of 117 patients (p < 0.0001); each cohort showed an increase (Holm p = 0.0004). In the pooled heterogeneity simulation, the calibrated true-variance fraction fell from 0.46 (p = 0.0005) without the correction to 0.07 (p = 0.28) with it, and the Q-test p-value changed from 0.035 to 0.51. The circular-shift sensitivity analysis gave a design effect of 1.74, a variance fraction of 0.10 (p = 0.21), and Q-test p = 0.45. Thus these archived data do not show that patients differ in true HFA–SOZ separability after accounting for shaft clustering. The design effect was estimated near AUC 0.5 and applied at cohort mean AUC; its behavior away from 0.5 is unknown. The clustered null places runs uniformly, unlike potentially nonuniform clinical SOZ placement. No raw EEG was reprocessed.

stereo-EEG · SEEG · seizure-onset zone · high-frequency activity · AUC · clustered sampling · permutation null · heterogeneity · epilepsy

Admitted 2 citing publications

References

13
  1. 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

    Source of the ds005398 UCLA pair table (input file).

  2. challengesAcross 117 SEEG patients in four cohorts, SOZ AUC spread exceeded simulated sampling noise (Q p = 0.040); exploratory true spread was 46% sci:sha256:bb82108c22380d643ec100e4bb6b304e7fe9edbeaedd3a79bfe1783a66be2863

    Same 117 patients and features. With a shaft-clustered label null, the sampling variance of the per-patient AUC is 1.8 times larger, and the reported AUC heterogeneity (p 0.040, fraction 0.46) is not significant (p 0.51, fraction 0.07).

  3. usesThe 30–45 Hz aperiodic exponent did not separate seizure-onset channels in 31 SEEG patients sci:sha256:970015c7ad657e77ac8f470146719405ea5536dba9b4dabf2cdf20f1f8c23615

    Source of the ds004100 HUP channel feature table (input file).

  4. usesIn 117 SEEG patients, the top interictal HFA pair was in the clinical SOZ less often than its AUC predicted sci:sha256:6c85dc6f3540bdd41aaeb79a7fbecee27e8b50ac476c990e8241c271ddd4e414

    Same 117-patient, 4-cohort feature selection and patient filter as this pooled HFA work.

  5. supportsIn 256 RAM patients, pair-level SEEG tests had 76–79% false positives under random labels; patient-level tests had 4.6–5.3% sci:sha256:702cad0cc9458f3cd5bfb4c9bb699326ade05ce17c4a2fb65b767d3b6877db64

    That work showed that pair-level tests that use pairs as independent samples are too liberal; here, spatial clustering inflates per-patient AUC variance by 1.8 times.

  6. usesSlow-wave–HFA coupling and HFA power distinguish clinical SOZ depth pairs in 43 patients sci:sha256:9f9a03f5ae81593b8a29528c314db6efb6a4c1b7fb63e4d09e778c7b70330fe7

    Source of the ds007118/9/20 NREM pair table (input file).

  7. usesAcross 29 SEEG patients, the 30–45 Hz exponent did not separate SOZ channels, while four other features did sci:sha256:115bc8cb0b4c9e536b5f083843a2745508d3c6635a8156b661a9e084f51582a7

    Source of the ds003876 channel feature table (input file).

  8. usesdoi:10.18112/openneuro.ds004100.v1.1.3

    The archived HUP channel-feature table used in this analysis was attributed to OpenNeuro ds004100.

  9. usesdoi:10.18112/openneuro.ds003876.v1.0.2

    The archived multicenter channel-feature table used in this analysis was attributed to OpenNeuro ds003876.

  10. usesdoi:10.18112/openneuro.ds007118.v1.0.0

    The archived NREM pair table was attributed to the ds007118/9/20 AutoSci contribution, which identifies OpenNeuro ds007118 as a source.

  11. usesdoi:10.18112/openneuro.ds007119.v1.0.0

    The archived NREM pair table was attributed to the ds007118/9/20 AutoSci contribution; ds007119 is one of its named source datasets.

  12. usesdoi:10.18112/openneuro.ds007120.v1.0.0

    The archived NREM pair table was attributed to the ds007118/9/20 AutoSci contribution; ds007120 is one of its named source datasets.

  13. usesdoi:10.18112/openneuro.ds005398.v1.1.3

    The archived UCLA pair table used in this analysis was attributed to OpenNeuro ds005398.

Cited by

2 publications

Public catalog entries on Sciscape.