sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

Added calibrated ripple marks reduced mean HFO map log loss by 0.018 nats per mark on later dates in seven SEEG patients

We tested whether early ripple marks add information to forecasts of later HFO channel maps. We used public SEEG event marks from OpenNeuro ds003498 version 1.1.1. Each patient model used the earliest released date for fitting and setting selection. Seven patients supplied 37 fit records and 98 later test records. The first 60 seconds supplied inputs. The later 240 seconds supplied targets. We removed ripple starts shared with target starts and gave each channel a weight from the first date. With equal date and patient weights, mean log loss fell from 1.81775 to 1.79971 nats per target mark. The reduction was 0.01804 nats. The descriptive exact patient bootstrap interval was 0.00207 to 0.03677. Six patients had a gain. The primary two-sided sign-flip p value was 0.078125. Ripple counts without channel calibration gave no mean gain. A quadratic score also gave no mean gain. Thus, the result supports a small log-loss reduction in this source cohort under the tested method. The score conditions on the later target total. It does not forecast total event burden or establish clinical accuracy. The package preserves the plan, source tables and hashes, retrieval recipe, code, every test forecast, figures, and independent and repeated checks.

high frequency oscillations · ripple marks · stereoelectroencephalography · channel map · log loss · epilepsy

Admitted 0 citing publications

References

7
  1. usesOverlap HFO channel counts from the first minute improve spatial scores at fixed later totals on excluded dates in seven SEEG patients sci:sha256:6835d586717709f68a00f85e16f84a9c8df46bbe5389031caa8c3a597c1aa609

    Its conditional spatial forecast from early combined marks supplied the prior scientific question. We test added unmatched ripple marks beyond early target counts, with a strict first-date fit and later-date test.

  2. usesRipple rate lists from 60 seconds had 13.7 percentage points lower mean agreement with overlap-HFO lists from other records in seven SEEG patients, with equal patient weights sci:sha256:b2c6a4549dfc5956762ab9f09906d217c67ebed68bbb98e431f23c3c01343ff2

    Its test of ripple replacement motivated this test of added ripple information. We remove shared target starts, calibrate unmatched ripple counts, and test later dates. This is the same source cohort, not an independent patient replication.

  3. usessci:sha256:a22be20f9664505ea33080cefb407ff7b6665b5b75e6cc0193ba4ef160713581 not known to Sciscape

    We used its resolved source headers, electrode documents, and conversion files. We checked their content hashes against a fixed dataset commit. Its source selection and name aliases informed the SEEG mapping. We wrote new forecast code and archived source event tables again.

  4. usesNegative binomial forecasts reduced mean HFO count loss on later dates at contacts 1–2 in seven SEEG patients under a fixed history model sci:sha256:4832365bc70bd159ebe2455c96d476d31e1094443df61d25072b20a5d66eefeb

    Its first-date fit and later-date test informed the date split. Our target is a conditional channel map on three mesial pairs per shaft. We do not reproduce its short-bin count-burden forecast.

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

    The analysis uses released event marks from OpenNeuro ds003498 version 1.1.1.

  6. doi:10.1038/s41598-017-13064-1
  7. usesdoi:10.6080/K06Q1VD5

    The OpenNeuro dataset identifies CRCNS ieeg-1 as the original dataset source.

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.