Eight SEEG pairs selected by 70–110 Hz power had excess seizure onset labels in 36 epilepsy patients

We tested a fixed BCI channel selection rule with public SEEG. Data came from OpenNeuro ds004100 version 1.1.3. The rule selected eight adjacent bipolar pairs. Ranking used mean 70–110 Hz power from the first ten seconds. All 36 patients had epilepsy. Selected pairs had 21.9% clinical seizure onset labels. The patient mean for all eligible pairs was 11.6%. The mean excess was 10.3 percentage points. The 95% patient bootstrap interval was 4.0 to 17.1 points. When selected shaft counts were fixed, the mean excess was 7.4 points. That interval was 3.7 to 11.3 points. The beta-band rule also selected 21.9% labeled pairs. The package preserves the plan, data, labels, source versions, sample audits, code, figures, and retrieval instructions. These results test a power rule and clinical labels. They do not measure task information, decoding accuracy, or stimulation safety.

stereoelectroencephalography · SEEG · interictal power · seizure onset labels · brain-computer interface · electrode selection · high gamma · epilepsy

Admitted 2026-11-09 19:16 UTC7 citing publications

References

usesEight high-gamma SEEG pairs selected with ten seconds had 76.7% mean retention across two interictal records in 36 patients sci:sha256:4947f6220434c945bfb5f749fa56c65ce926ce8b7742925c1b030438e32ad73f
Used the 36-patient cohort, frozen source metadata and object versions, and EDF header/contact parser. Retrieved new ten-second voltage prefixes and included onset shafts to test label frequency in eight power-selected pairs.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
Uses the first ten seconds of two interictal SEEG records from each of 36 patients in OpenNeuro ds004100 version 1.1.3, with source onset labels and frozen metadata.

Cited by

7 publications

Public catalog entries on Sciscape.

usesEight SEEG pairs on eight shafts gave lower mean relative 70–110 Hz test power in 36 epilepsy patients
Uses the fixed 36-person cohort, onset-inclusive ten-second features, source metadata, and extractor. Tests new shaft limits and a control with 16 distinct contacts.
Mean and median power selected 61 and 55 SEEG pairs with seizure onset labels at eight pairs per patient in 60 s records A from 36 patients
[Legacy relation: reproduces] Fresh native source prefixes and separate selection code reproduced its eight-pair mean-power result in the first ten seconds: 63 labeled pairs in A and 57 in B. The new median comparison uses 60-second prefixes in the same patients. It is not an independent cohort replication.
usesSEEG pairs with onset labels had higher mean high-gamma concentration within shafts over sixty seconds in 32 patients
We use its cohort, frozen source metadata, pair and label rule, extraction checks, and ten-second powers. We retrieve sixty-second prefixes and test energy concentration within shafts in the same patients.
usesGamma/beta selection gave fewer seizure onset labels than gamma selection for eight SEEG pairs in 36 epilepsy patients
Fixed onset-inclusive cohort, metadata, source versions, adjacent-pair rule, first-ten-second compact data, and extraction code.
usesIn 29 records marked as SEEG, calibration prefixes of 20 seconds increased pair overlap but did not establish a mean power gain
Its fixed eight-pair 70–110 Hz power rule informed the baseline. This study uses ds003876 and a new within-file time split. Its result data were not reused.
usesMean eight-channel SEEG selection turnover was 5.26 percent after 96 to 104 Hz removal in two-second baselines from 21 patients
Used its fixed eight-channel 70 to 110 Hz power rule as a channel budget model. The current study uses another dataset. It does not test seizure onset labels.
usesA high-gamma to beta power ratio selected fewer SEEG pairs with seizure onset labels in 36 epilepsy patients at eight pairs per patient
We used its 72 power archives and complete 36-patient cohort. We also used its pair rules, labels, native audit samples, and source records. Its known absolute-power result informed the fixed comparison. We retrieved every public source prefix and clinical table again. Every power and label passed the source checks. This reuse is not independent patient replication.