In six patients, window-end Hilbert estimation outperformed AR(20) forecasting for NREM slow-oscillation phase
This pilot compared two causal estimators of the 0.5–2 Hz NREM slow oscillation (SO) phase in mesial-temporal depth iEEG from six patients with epilepsy. It analyzed 302 bipolar channels from 39–50 s at the start of one 5-min interval per patient in OpenNeuro ds003498 v1.0.1. Signals were block-mean decimated from 2000 to 100 Hz. Each estimator used a 4-s past-data window: E1 read the Hilbert phase at the window end; E2 fit an AR(20) model and forecast 0.5 s ahead. The reference was the zero-phase estimate on the full segment. At high-SO-amplitude samples, E2 had larger median absolute phase error than E1 in all six patients (mean difference +30.8°, range +24.5° to +34.7°; exact two-sided sign-flip p = 0.031). E1 median error was 39–42° and hit rate within 45° was 0.64–0.70 (chance 0.25); its phase lead was +10° to +17°. Predicting E2 an additional 100 ms ahead raised error to 73–82°. Higher SO relative power was associated with smaller error for both estimators. These findings apply to the tested AR configuration, not AR forecasting in general. The pilot used short segments, did not verify sleep stage sample by sample, and used a filtered reference rather than an independent measure of the up-state; bipolar up-state polarity was unknown. Epileptic channels were retained and interictal spikes were not analyzed.
slow oscillation · NREM sleep · phase estimation · intracranial EEG · depth electrodes · closed-loop stimulation · autoregressive forecasting
References
2- usesIn 9 mesial temporal epilepsy patients, fast-ripple maps stabilized after about 2–5 NREM intervals, but more data did not improve resection agreement sci:sha256:a93d7e88a037328125175f7eb8424ba2187abf691749276092de32e970fa45b9
We used its cohort description to identify sub-01 to sub-09 of ds003498 as patients with mesial-temporal depth electrodes.
- usesdoi:10.18112/openneuro.ds003498.v1.0.1
Source recordings analyzed in this pilot (OpenNeuro ds003498 v1.0.1).
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesIn five patients, most significantly coupled SEEG contacts had negative-half HFA peaks, while 90-s calibration did not improve trigger HFA
We use its causal SO phase estimator (E1) and test the up-state polarity question that it left open.
- supportsA calibrated constant offset matched frequency-advanced ecHT for 100 ms SO-phase prediction in held-out SEEG from 6 patients
Held-out data from the same 6 patients give the same 0 ms windowed-Hilbert SO phase error as the pilot.
- usesA calibrated constant offset matched frequency-advanced ecHT for 100 ms SO-phase prediction in held-out SEEG from 6 patients
Uses the pilot code, filter, bipolar montage, truth and scoring rule; extends the pilot to a 100 ms latency on new non-overlapping data. The 0 ms E1 error (41.2 deg) agrees with the pilot (39.1-42.3 deg).
- usesA causal slow-oscillation phase trigger weakly enriched ripples in a held-out test of 19 patients
That work measured causal SO phase error in ds003498. We used a causal SO phase estimate and tested whether that phase predicts ripples.
- usesIn 12 SEEG patients, AR(20) had larger slow-oscillation phase errors than calibrated Hilbert hold at 100 ms
The pilot SO phase estimation study that defined the E1 windowed Hilbert estimator.
- usesIn three NREM-sleep SEEG patients, bipolar SO phase prediction was weak even at 0–10 mm and error rose with distance (not significant)
Causal SO phase estimation on one channel; our result adds that the phase does not transfer between bipolar channels.
- Causal slow-oscillation triggers had 44° phase error and retained 13% of oracle HFA gain in an offline 70-subject SEEG study
[Legacy relation: reproduces] Same causal estimator E1. Phase error 44 deg vs about 40 deg and a similar phase lead, in 70 subjects from an independent dataset.