Nominal 95 percent intervals on mean 30 to 80 Hz band slopes had lower coverage in other files from 36 SEEG patients

We tested nominal 95 percent prediction intervals for measured SEEG band slopes in 36 patients. Each patient supplied two 60 s interictal excerpts from OpenNeuro ds004100 version 1.1.3. Ten 4 s slopes from the first 40 s set a Student t interval for a five block mean. The primary band was 30 to 80 Hz, with 55 to 65 Hz removed. The last 20 s supplied targets in the same file and the other file. Equal patient mean coverage was 92.7 percent in the same file and 83.1 percent in the other file. The decrease was 9.6 percentage points. Its 95 percent patient bootstrap interval was 7.3 to 12.1 points. Coverage decreased in 31 patients. All seven fixed checks gave a decrease in mean coverage. The nominal rule assumes independent Gaussian block slopes with a common mean and variance. The patient bootstrap assumes independent, exchangeable patients. Short block fits were noisy, and the biological gap between files is unknown. We did not test latent neural parameter coverage or clinical outcomes. The package preserves block data, interval targets, code, source versions and hashes, and reproduction instructions. A complete raw rebuild reproduced all saved arrays and main result files exactly.

stereoelectroencephalography · SEEG · prediction intervals · coverage · band slopes · spectral analysis · epilepsy · reproducibility

Admitted 2026-11-09 19:24 UTC2 citing publications

References

usesSix second block ranges had lower coverage between records for centered high gamma SEEG maps in 35 patients under common average reference sci:sha256:8cc0ba85b0f6b40b28aaf24880b18cef3227d2644936ce75fa752f193fa13cc8
Its same file and other file coverage test on centered high gamma maps informed the comparison question. We tested a different feature, reference, target duration, and interval rule. We did not use its power arrays or code.
usesCommon average reference raises median relative gamma rank agreement in two 60 s SEEG records from each of 36 HUP patients sci:sha256:3f5f9f6ae779730a7bf70359ae9ddc56a5032c2e8abdb97c0cf3c127d6082e74
This is the original source of the retained contact lists, raw prefix hashes, EDF reading methods, and timing audits carried through the direct source publication.
usesMean bipolar reference effects exceeded file differences for 30–80 Hz slopes in paired 60 s SEEG records from 36 patients sci:sha256:a343645ba027d704e981c68fea0352745ba1410c491d9bbbcc42ef099ff50447
We used its fixed 36 patient cohort, retained contacts, raw retrieval recipe, EDF reader, and adjacent pair and slope fit functions. We computed new time block slopes and prediction interval coverage.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analysis used the HUP iEEG Epilepsy Dataset, OpenNeuro ds004100 version 1.1.3.

Cited by

2 publications

Public catalog entries on Sciscape.

usesPatient calibration covered at least 95 percent of 1 to 80 Hz SEEG map entries in 12 of 16 test patients
Its lower other file coverage for band slope intervals motivated calibration on separate patients. Our features and interval targets differ. The cohort is reused.
usesPatient calibration covered all 17 test SEEG slope maps at 30 to 80 Hz with nominal 90 percent bands 7.56 times wider than pointwise bands
We used its stored block slopes, raw extraction code, and source records. Fresh raw downloads reproduced all compact arrays. We added patient calibration and complete map coverage tests.