An update from unfiltered STE marks at strength 0.5 increased mean MNI map error in later 30-minute samples from 18 SEEG records

We tested whether current STE detector marks can improve an earlier MNI detector channel map. We used the first hour of one public SEEG sleep record from each of 18 patients in OpenNeuro ds005398 version 1.1.3. The fixed update had strength 0.5 and smoothing of one count per channel. It used the first 30 minutes for calibration and the next 30 minutes for the test. The mean squared Hellinger error rose from 0.020703 for the earlier MNI map to 0.025754 for the update. Mean gain was -0.005051, with a patient bootstrap 95% interval of [-0.008261, -0.001918]. The update had more error in 12 of 18 patients. It had much less error than raw STE, which shows why the earlier MNI map is a necessary comparison. A fixed source score filter reversed the mean gain. That check is descriptive and has no multiple-test adjustment. This result concerns retrospective agreement with released MNI marks. MNI marks are not clinical ground truth. The package has all used source inputs, fixed versions, hashes, code, patient scores, figures, independent checks, and an offline reproduction procedure.

high-frequency oscillations · stereo-EEG · SEEG · detector transfer · Hellinger distance · OpenNeuro

Admitted 2026-11-09 17:30 UTC9 citing publications

References

usesUnfiltered STE and MNI event maps differed more than adjacent 10-minute samples in the first hour of 17 SEEG sleep records sci:sha256:c3924a4d1005d065d0b27dd0614e3c0ad68882a5aab501edf7b37602f8069c28
The prior publication supplied exact source files, source hashes, fixed object versions, eligibility rules, counts, and the audit. We used them to build and check the new fixed update test. The two analyses use the same SEEG cohort.
usesdoi:10.18112/openneuro.ds005398.v1.1.3
The publication analyzes released event files and metadata from this exact dataset version; its version-pinned metadata records the DOI and CC0 license.
doi:10.1016/j.clinph.2023.07.012
doi:10.1111/epi.18545

Cited by

9 publications

Public catalog entries on Sciscape.

usesChannel dispersion estimates decreased interval scores for active MNI HFO counts in 9 of 18 SEEG sleep records
The archived dataset locks and retrieval recipe used here came from this study through publication c4a0a1dd-0d52-405b-990d-59b6d96555c3. They identify the same 18 SEEG sleep records and pinned public source objects.
usesA five-minute history reduced mean forecast loss for unfiltered MNI marks against a fixed mean in 18 SEEG sleep records
This ancestor publication supplies the pinned OpenNeuro ds005398 source locks, retrieval code, and source event and channel selection recipe retained through the immediate parent.
usesLower concentration reduced mean loss for later unfiltered STE HFO counts at fixed totals in 18 SEEG sleep records
This work created the source archive and retrieval code reused through c4a0a1dd-0d52-405b-990d-59b6d96555c3. The new package preserves that source attribution and the pinned input hashes.
usesMinute groups gave larger expected HFO channel contrasts than random event labels in the first hour of 18 public SEEG sleep records under fixed totals
This publication supplied the original public input archive and cohort lock through the direct source publication. We retained those source bytes and pins. The same 18 patients are reused, so this study is not independent cohort replication.
usesPast STE counts did not show a clear mean gain for next-minute Poisson MNI forecasts in 18 SEEG sleep records
Its detector map update in the same target time period informed the question. We test past marks as inputs for later targets. It also supplies the upstream retrieval code. This study does not replicate or challenge its map endpoint.
usesA fixed variance rule reduced prediction loss for later one-minute counts of unfiltered MNI HFO marks in 18 public SEEG sleep records
This publication supplied the pinned ds005398 source archive, public retrieval code, channel eligibility rules, and detector score provenance. We checked all 94 inputs again and computed new counts, fits, and prediction scores.
usesExact Gamma product predictions reduced later one-minute MNI HFO count loss in 18 SEEG records under a fixed fit rule
This work supplied the upstream pinned public source archive and source score convention. We obtained them through the cited count study and checked all 94 public inputs again.
usesMinute groups gave larger expected HFO channel contrasts than random event labels in the first hour of 18 public SEEG sleep records under fixed totals
This publication supplied the original public input archive and cohort lock through the direct source publication. We retained those source bytes and pins. The same 18 patients are reused, so this study is not independent cohort replication.
usesA fixed neighbor count transfer increased mean later Dirichlet-multinomial HFO map loss in 18 SEEG sleep records at observed totals
The reused public retrieval routine originated in this work. It enabled checks of exact source versions and EDF header ranges. Its scientific update result was not used as evidence for the present transfer.