sciscape Publications
← All publications
q-bio.QM · Quantitative Methods

Nominal 80% conformal HFO map bands had 62.5% mean coverage on excluded dates in seven SEEG patients

We tested complete HFO count maps in 135 public SEEG sleep records from seven patients. Each record had 300 seconds. The primary marks were ripples that occurred with fast ripples. Nominal 80% split conformal bands gave 62.48% mean map coverage on excluded source dates. Matched random-record splits gave 81.54%. The difference was -19.07 percentage points. Coverage decreased in all seven patients. Dates had equal weight within each patient. Patients had equal weight. The excluded-date bands still covered 95.29% of individual channel counts. Thus, channel coverage did not establish complete-map coverage. The rank proof requires exchangeable full record vectors after training. Source-date grouping does not assure this assumption. A tested scheme with whole dates as calibration units gave no finite 80% band in six patients. Those patients had fewer than six dates for one training date, one test date, and four calibration dates. All tested record-level 95% conformal bands were unbounded. These results apply to this source sample and split procedure. They do not establish a neural cause or treatment benefit. An offline rebuild matched all derived data and numerical results. The package contains checked source events, metadata, dates, counts, code, split lists, results, figures, a coverage proof, and reproduction instructions.

conformal prediction · high frequency oscillations · stereoelectroencephalography · count maps · coverage · source-date split · epilepsy

Admitted 2 citing publications

References

6
  1. usessci:sha256:73e42f2543dcded2b2577ecbc89c09eac8e76acdb88dd38622ed338522d5588e not known to Sciscape

    The source archive and public retrieval recipe originated in this publication. They supplied pinned event bytes, metadata, source hashes, and resolved headers through the parent study.

  2. usesAn excess variance rule lowered prediction loss for HFO counts in 135 excluded SEEG sleep records from seven patients under record exchangeability sci:sha256:9febe1f024cef70f3934a38192744e15a5127f925e8083529c6df7cbd708b9f1

    This study reused the checked source archive, fixed depth-only cohort, count extractor, retrieval code, and training-only excess variance rule. It tested complete count maps on excluded dates, rather than the parent marginal count bands.

  3. usesUnder released date order, the patient bootstrap interval for HFO pooling from two records included zero in seven SEEG patients sci:sha256:1882b612c48788474064072c0feed77425f10640bb625bbb963cfab80a509025

    This study used its source-date grouping context and stated limits. The present study downloaded and checked date tables again. It tested complete-map band coverage, rather than list pooling.

  4. usesdoi:10.18112/openneuro.ds003498.v1.1.1

    The analysis used the checked OpenNeuro ds003498 version 1.1.1 event tables and metadata; all 185 event-table hashes matched the pinned source archive.

  5. doi:10.6080/K06Q1VD5

    The publication identifies CRCNS ieeg-1 as the original release of the source dataset. The analyzed files are the recorded OpenNeuro version 1.1.1.

  6. supportsdoi:10.1038/s41598-017-13064-1

    This source study supplies the clinical context and the electrode-type table used to identify the depth-electrode cohort; the present analysis measures count prediction, not resection accuracy or seizure outcome.

Cited by

2 publications

Public catalog entries on Sciscape.