An intercept fitted to 15 development patients reduced mean SOZ label Brier score for a fixed balanced logistic model in 16 ds004100 SEEG test patients
We tested whether a development intercept reduces clinical seizure onset zone (SOZ) label probability error in public stereoelectroencephalography (SEEG) data. A fixed logistic model used 15 development patients and 16 test patients from OpenNeuro ds004100 version 1.1.3. The model gave each development patient and each label class equal weight. Five number features and six spectrum features were fixed. Five patient folds supplied logits for the intercept fit. Test patients supplied no fit values. Mean test patient Brier score fell from 0.21892 to 0.09402. The difference was -0.12490. Its conditional 95 percent patient bootstrap interval was [-0.16076, -0.08650]. Its approximate development and test refit interval was [-0.17481, -0.07262]. Error fell in 14 of 16 test patients. Mean patient area under the receiver operating characteristic curve (AUC) stayed at 0.76071. Brier advantages over the constant and adjusted number models had intervals that included zero. This internal test used clinical contact labels from one hospital and a previously studied cohort. It does not establish clinical benefit. The package preserves source spectra, metadata, retrieval receipts, code, all predictions, model states, bootstrap draws, and checks.
SEEG · seizure onset zone · probability calibration · Brier score · logistic regression · patient bootstrap
References
5- 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
We used the archived spectra, source code, and pinned raw-data retrieval records. A fresh reconstruction checked all 34 source spectrum files. These files supplied the spectrum features for the new probability test.
- usesIn 30-second HUP SEEG samples, reference changes reduced mean 30–100 Hz slope rank agreement more than run changes sci:sha256:b9c7b4646dba303292f492835f91dbb8026d4f35fcce938519d1d3fa6bcbff63
We used the earlier source-processing record for the 30-second signal sections, eligible contact pool, reference definitions, and Welch spectrum method. We retained the inherited reader and checked four sections with a second EDF reader.
- usesAdding six spectrum features increased mean SOZ label AUC for a contact number model in nested CAR logistic tests of 31 ds004100 SEEG patients sci:sha256:fcd88a615d5ea2329b707a35c2f4b90cbffcd52b30e922a59163788221cf8358
We used the fixed 31-patient cohort, contact number features, spectrum features, input build code, and source metadata. We fitted new models and tested label probability error on a fixed 15-patient development and 16-patient test split.
- doi:10.1093/jamia%2Focac093
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analysis used the SEEG recordings and clinical contact labels from OpenNeuro ds004100 version 1.1.3.
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesA prior fit without labels increased mean squared label error for a fixed HUP model in 18 external SEEG patients
Its fixed-model source intercept result informed the source fit rule. This study makes new source patient exclusion fits before the external comparison.