Nested calibration gave mean Brier score 0.0879 compared with 0.0886 after frequency correction for seizure onset labels. Tests excluded each of 31 SEEG patients from fitting.
We tested probability error for seizure onset labels in 31 public SEEG patients and 2,879 contacts. Each test patient was absent from all fitting stages. A fixed logistic model used contact number and six spectral features. Nested logistic calibration gave mean patient Brier score 0.0879. A training label frequency correction gave 0.0886. The primary difference was -0.00070. A 100-draw patient bootstrap refitted the full procedure. Its 95 percent percentile interval was [-0.00344, +0.00014]. This approximate internal interval includes zero and has limited simulation precision. The uncorrected balanced score gave Brier score 0.2033. All 62 source signal prefixes passed an independent spectrum check. The package preserves source spectra, labels, pinned retrieval records, code, fit equations, all ordinary scores, bootstrap metrics, and reproduction steps. The result concerns source labels in one selected cohort. It does not establish a surgical benefit.
probability calibration · Brier score · SEEG · seizure onset labels · electroencephalography · patient-level validation
References
5- usesSix CAR spectral features raised mean seizure onset label AUC beyond contact number. Fixed logistic tests excluded each of 31 SEEG patients. sci:sha256:aee5a470b0833c4f856b9a317fa6e6005828d574f583119a67ef17a72b7a7b3f
This study fixes the cohort and the seven inputs. We used its copied features, saved model scores, and source archive.
- usesContact splits gave higher mean seizure-onset label AUC than patient splits for a fixed spectral random forest in 31 public SEEG patients sci:sha256:b6a361054e4ebea1dd4b8f14fdd99003ce2639da49c389eca05a7a1c82bba254
Its feature preparation code, source labels, and cohort audits were kept in the parent input. They enabled independent input checks.
- usesIn 30-second HUP SEEG samples, reference changes reduced mean 30–100 Hz slope rank agreement more than run changes sci:sha256:b9c7b4646dba303292f492835f91dbb8026d4f35fcce938519d1d3fa6bcbff63
Its compact spectra, signal metadata, and pinned retrieval recipes supplied the source data. They enabled new raw signal checks.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
Source SEEG signals and clinical onset labels from the recorded OpenNeuro ds004100 version 1.1.3.
- doi:10.1093/jamia%2Focac093
General background on how class imbalance correction can affect risk probability prediction.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.