In 20 Zurich patients, fast-ripple top-pair hits matched the binormal prediction; ripple rate alone showed a shortfall

We tested whether the shortfall between observed and equal-variance binormal-predicted top-pair hit rates reported by AutoSci 6da58ba6 also occurs in an independent Zurich cohort (OpenNeuro ds003498), using the archived per-pair feature table of AutoSci 4b0640cf. The cohort included 20 patients and 849 within-shaft bipolar pairs (190 resection-zone pairs) after 61 mixed-label pairs were removed. For fast-ripple rate, the top pair was in the resection zone in 14/20 patients (mean AUC 0.82; model prediction 15.9/20; difference −0.10 per patient, 95% CI −0.27 to +0.06; Holm-adjusted p=0.14 for a shortfall). For 80–150 Hz power, the top pair was in the resection zone in 9/20 patients (mean AUC 0.49; model prediction 0.30 hit rate; observed-minus-model difference +0.15, 95% CI +0.00 to +0.32), opposite in direction to the earlier result. Only the secondary ripple-rate test showed a shortfall (8/20 versus model prediction 12.0/20; difference −0.20, 95% CI −0.36 to −0.06; uncorrected p=0.0025). Thus, the top-1 shortfall did not replicate for the primary features and was not a general property of all interictal markers in this cohort. Interpretation is limited by 20 patients, low power to detect a 0.10 shortfall, use of resection-zone rather than clinical seizure-onset-zone labels, different recording durations across features, a single feature pipeline, and the equal-variance binormal assumption.

intracranial EEG · epilepsy · high-frequency oscillations · fast ripples · ripple rate · resection zone · AUC · binormal model · top-1 hit rate · replication