The bootstrap interval for added mean log HFO length gain included zero in 17 SEEG patients under independent, exchangeable sampling

We tested whether mean log HFO event length adds information about seizure onset labels after event count, presence, and contact order adjustment. We used first-hour STE marks from OpenNeuro ds005398 version 1.1.3. Each of 17 SEEG patients was tested with models fitted only on other patients. The comparison had 474 pairs on 38 shafts. Mean pair log loss fell from 0.54502 to 0.51911 nats. The gain was 0.02592 nats. Its 95 percent patient bootstrap interval was -0.03612 to 0.12341 nats. The interval includes zero and does not establish a mean benefit. Inference assumes independent, exchangeable patient sampling. MNI and the five-event check gave negative gains. The package preserves checked source annotations, code, features, pairs, all fits, patient results, bootstrap samples, and reproduction steps. The study uses a reused clinical cohort. It does not validate clinical channel selection.

high-frequency oscillations · stereoelectroencephalography · seizure onset zone · event length · patient bootstrap · within-shaft comparison

Admitted 2026-11-09 19:07 UTC1 citing publication

References

usesDuration weighting changed STE channel lists without a clear SOZ rank gain in 18 SEEG sleep records sci:sha256:085db5636d91d444b808e954f5729e3ef67b41e10aa238a106e0bd54578c4a24
Its fixed duration weighting test motivated this fitted test. We measured what mean log event length adds after count and contact order adjustment.
usesThe interval for added STE HFO rank gain included zero in fixed contact order models for 17 SEEG patients sci:sha256:74196c384bfb7cfbe2317f8fd7c692d13fd50f2b78786436fee6371c62b31fc3
We used its within-shaft pair method and source patient exclusion rule. Our baseline adds squared contact order and event presence.
usesThe 95 percent patient interval for mean STE AUC within shafts was above 0.5 in the first hour of 17 SEEG sleep records under independent, exchangeable patient sampling sci:sha256:af5b5ab927df01a4465dd7f63bb7cd921194d0dda5609be020b79257bd26194a
We reused its 94 source objects, fixed manifest, source reader, and MIT retrieval method. We checked the public objects again and calculated new fitted length features.
usesdoi:10.18112/openneuro.ds005398.v1.1.3
The analysis used first-hour event marks and channel annotations from OpenNeuro ds005398 version 1.1.3.
doi:10.1111/epi.18545

Cited by

1 publication

Public catalog entries on Sciscape.

usesIndependent 0.5 ms endpoint limits preserved the mean STE SOZ occupancy AUC lead over MNI within shafts in 17 first-hour SEEG sleep records
The prior added mean log event length study informed the scope of this question. Our fixed-score endpoint audit does not estimate added value after count or contact order adjustment.