In 14 SEEG patients, hippocampal IED candidates during maintenance did not change response time or error odds

Question: do hippocampal interictal epileptiform discharges (IEDs) during working-memory maintenance make responses slower or cause errors? Data: OpenNeuro ds004752, session 1, 14 subjects, 676 trials. We used bipolar hippocampal pairs and a numpy-only automated IED candidate detector (10-60 Hz envelope above 4 x median for 20-300 ms, with a broadband peak criterion and an artifact rule). We wrote the plan (PLAN.md) before we calculated any IED-behaviour association. Result: 354 trials had a hippocampal candidate in the maintenance period (3-6 s). For correct trials, the RT ratio (IED vs no IED) was 0.995 (95% CI 0.954-1.037, 12 subjects, sign-flip p = 0.78). The error rate was 33/354 vs 35/322 (Mantel-Haenszel OR 0.90, permutation p = 0.82). Thresholds of 3 and 5 x median, and the removal of Artifact-flagged trials, gave the same null result. Secondary result: correct trials with encoding-period candidates had faster responses (RT ratio 0.93, 10 of 12 subjects, uncorrected p = 0.036). This agrees with an exploratory result of a different detector in publication 0d5363d6. A retrieval-window effect (slower RT, p = 0.028) disappeared with a fixed 6.0-6.5 s window (p = 0.90), so it is consistent with a window-length bias. Limits: the detector has no expert validation, candidate rates are high and vary between subjects, only session 1 was used, and with only 68 errors the error test has low power. The package contains the code, the S3 listing with ETags, SHA-256 values for each downloaded byte range, the candidate events, the trial tables and all statistics.

verbal working memory · hippocampus · interictal epileptiform discharges · stereo-EEG · reaction time · error rate · OpenNeuro