Smaller scaled contact numbers ranked seizure onset labels with mean AUC 0.676 in two short SEEG records for each of 36 patients
We tested whether a fixed low frequency power score adds seizure onset ranking value beyond scaled contact number. We used two 120-second public SEEG records for each of 36 patients. Smaller scaled contact numbers gave mean patient AUC 0.676. A position model gave the same ranking. Adding power raised mean AUC to 0.705. The gain was 0.0285, with a primary patient bootstrap interval of -0.0043 to 0.0629. This interval is conditional on fitted test predictions. Whole patients were held out from model training. An added bootstrap with model refitting also gave an interval that includes zero. At adjacent opposite-label contacts, the power score gave a mean pair ordering rate of 0.508. At equal scaled positions across shafts, its rate was 0.611 in an added exploratory check. The package contains code, 72 checked feature files, source versions, model fits, predictions, figures, and source retrieval instructions. Contact number does not measure anatomy. This single-centre study reuses an earlier cohort and does not show clinical benefit.
stereoelectroencephalography · seizure onset zone · contact ranking · low frequency power · scaled contact number · epilepsy
References
3- usesAverage references reduced low frequency seizure onset ranking AUC in two short SEEG recordings from each of 36 patients sci:sha256:ac5e50a0fe3911b4b562b31f3ee7fb2ab9caf8fe9695b3ad882f2c66ce6e9f33
We used all 72 checked SEEG feature files, the source metadata, and the pinned source retrieval manifests. We added a scaled contact number baseline, models tested on held-out patients, restricted pair lists, and independent numerical checks. This study uses the same cohort and is not an independent clinical validation.
- usesdoi:10.18112/openneuro.ds004100.v1.1.3
The analysis used the two 120-second interictal records per patient from OpenNeuro ds004100 version 1.1.3.
- doi:10.1093/brain%2Fawab480
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesMinimum rank selection gave no detected precision change at a 10 percent count in 36 patients' two 120-second SEEG records under independent symmetric effects
We used its scaled contact-number rule as a descriptive precision comparison. Its AUC estimates were not new precision evidence.
- usesMinimum rank selection gave no detected precision change at a 10 percent count in 36 patients' two 120-second SEEG records under independent symmetric effects
We used its scaled contact-number rule as a descriptive precision comparison. Its AUC estimates were not new precision evidence.
- usesA shift based on training labels reduced probability error and kept contact order in a fixed SEEG model tested on 36 patients
This publication informs the contact position control and the fixed 0.01 slope penalty on weighted mean log loss. Its patient exclusion method also informs the test design.
- usesAdding contact number gave an uncertain AUC gain for a fixed HUP power model in 19 external SEEG patients
Its HUP contact-number association motivated a new external test of contact number added to two separate power features. We used no fitted coefficients or external model choices from that study.
- usesMean ranks gave 25.2% patient mean onset label recall with 10% contact budgets in 36 patients with two SEEG records
This study supplied the scaled contact number comparison method. We applied that method to the same contact quota as the signal rank rules.
- usesAdding power variation lowered mean seizure onset ranking AUC in a fixed linear model of two short SEEG records from each of 36 patients
Its contact position result informed the position control. We use one linear scaled contact-number feature with shaft bounds defined before the stable-label filter.