Causal spindle triggers in mesial temporal SEEG leave 0.21 s at 100 ms dwell and co-occur with HFOs

In nine patients with temporal lobe epilepsy, we assessed whether a causal 11–16 Hz detector could trigger early enough for spindle-locked stimulation on mesial temporal SEEG, and whether its triggers co-occurred with author-marked high-frequency oscillations (HFOs). We used five minutes of NREM sleep per patient from OpenNeuro ds003498 v1.0.1, with a causal 200 ms RMS envelope, a threshold calibrated on the first 150 s, and evaluation on the last 150 s. Against an offline zero-phase spindle detector, at the pre-specified 100 ms dwell the median trigger latency was 0.52 s after spindle onset, median remaining spindle time was 0.21 s, precision was 0.40, and sensitivity was 0.70. Only 45% of detected spindles had at least 250 ms remaining. Increasing dwell from 0 to 200 ms reduced sensitivity from 0.85 to 0.41 without increasing precision. At 100 ms dwell, ripple marks occurred near triggers at 0.35 versus 0.14 for random times, and fast-ripple marks at 0.07 versus 0.04; 8 of 9 patients had higher co-occurrence (p = 0.039 for each). This result was exploratory: it was significant only at this dwell and was not corrected across three dwell settings. Triggers inside offline spindles had more HFO co-occurrence than triggers outside (0.50 versus 0.32; 8 of 8 patients, p = 0.008). The findings suggest that this threshold detector fires late for spindle-locked stimulation in these contacts. The reference was automatic rather than expert scoring, only five minutes per patient were analyzed, sleep stage within NREM was not separated, and the sample included epilepsy patients only.

sleep spindles · causal detection · SEEG · high-frequency oscillations · closed-loop stimulation · temporal lobe epilepsy · signal processing