With independent, exchangeable patient sampling, gain intervals included zero for SOZ rank and list agreement after a fixed STE HFO neighbor average in 18 SEEG sleep records

We tested a fixed neighbor average of released HFO counts in the first hour of 18 public SEEG sleep records. The main STE test used 1,285 channels and 23,297 events. The score used half the target count and half the mean of immediate same-shaft neighbors. Mean seizure onset zone rank AUC rose from 0.7188 to 0.7388. The gain was 0.0199, with a 95% patient bootstrap interval of [-0.0016, 0.0429]. Mean top 20% channel list agreement between separate half-hours rose from 0.7441 to 0.7487. The gain was 0.0046, with interval [-0.0565, 0.0634]. Both planned 97.5% gain intervals also included zero. The Holm test probabilities were 0.2107 and 0.8858 under the patient sign symmetry null. Population inference assumes independent, exchangeable patients. These conditions are not established by the convenience sample. All eight fixed AUC check intervals included zero. The cohort was used in earlier work. The result concerns released marks and clinical SOZ labels. It does not establish equal performance or surgical benefit. The package preserves source bytes, provenance, a dated plan, code, scores, patient results, all uncertainty arrays, and figures. Independent checks passed. A complete offline repeat matched every numerical output. Fresh public requests matched all 94 source files or EDF headers.

high-frequency oscillations · stereoelectroencephalography · seizure onset zone · neighbor averaging · patient bootstrap · channel agreement