A spindle detector decrease at seizure-onset-zone pairs is not robust to interictal spikes

In 43 patients with interictal NREM sleep SEEG, a pre-registered detector found lower spindle event rates in seizure-onset-zone (SOZ) bipolar pairs than in other pairs from the same patient (mean per-patient AUC 0.369, 95% CI 0.302–0.437; Holm-adjusted p=0.001). However, an event-independent 11–16 Hz spectral peak showed no SOZ difference (AUC 0.501; Holm-adjusted p=0.98), candidate events before rejection were similar (AUC 0.498), and the decrease was removed by robust thresholds and adjustment for 80–300 Hz power. Non-SOZ pairs with spikes also had fewer detected events. A synthetic check showed that spikes reduced detected rates under the detector. These findings indicate that spike-sensitive rejection and mean-plus-standard-deviation thresholds can produce an apparent SOZ spindle decrease; these data do not establish fewer underlying spindles in SOZ tissue. The analysis used 200 seconds per patient and has the limitations detailed in the report, including low event rates, possible anatomical confounding, and no expert spindle-mark validation.

sleep spindles · seizure onset zone · SEEG · intracranial EEG · epilepsy · interictal spikes · detector bias · signal processing