The IED and recall-classifier protocol produced no data results because downloads stalled

This package documents a protocol and Python code to test whether interictal epileptiform discharges (IEDs) affect an encoding-period band-power classifier of word recall from stereoelectroencephalography (SEEG). The question matters for closed-loop memory stimulation: if IED words receive low predicted recall probabilities, a system that triggers stimulation at low probabilities may stimulate preferentially during IEDs. The planned data were RAM FR1 recordings from OpenNeuro ds004789, session 0, using bipolar EDF files and associated event and channel tables. The proposed pipeline extracts mean log10 Morlet power at eight frequencies from 3 to 140 Hz during the 0–1.6 s after word onset, detects candidate IEDs with a thresholded 10–60 Hz Hilbert envelope, and evaluates an L2 logistic regression classifier with leave-one-list-out cross-validation. Planned analyses compare recall and classifier performance for IED and IED-free words, use random word removal and IED count controls, simulate a low-recall stimulation trigger, and apply a subject-level sign-flip test. The download stalled at about 105 MB per subject, and no subject finished during the 45-minute run. Therefore, this package reports no data results; the analysis and figure scripts were not run on real data, and the threshold detector has no expert validation. The dataset URL does not identify an immutable release, so no version-specific dataset DOI is asserted.

interictal epileptiform discharges · SEEG · word recall · closed-loop stimulation · logistic regression · OpenNeuro · protocol · no results