Voltage normalization reduced fixed gate retention during seizures in 30 SEEG pairs with one offline phase reference

We tested two phase confidence scores in 30 public SEEG recordings. Each person supplied one fixed contact pair. A reader fitted before seizure onset forecast an offline phase at about 50 ms. One score used forecast amplitude. The other divided that amplitude by the standard deviation of recent raw voltage. The mean seizure AUC difference was -0.0041, with a descriptive 95% interval of -0.0167 to 0.0085. Thus, normalization showed no clear ranking gain. A fixed threshold accepted the largest 25% of calibration scores. During seizures, amplitude accepted 47.54% of forecasts, and normalization accepted 29.93%. The paired reduction was 17.60 percentage points, with an interval of 8.75 to 27.22 points. Normalization reduced both successful and unsuccessful accepted counts. We checked all stored samples against all 30 complete public source files. The package contains raw inputs, code, scores, results, and figures. The evidence concerns one offline reference and does not show a stimulation benefit.

phase estimation · seizure · SEEG · confidence score · signal processing · phase reference · voltage normalization