Causal ripple triggers fired later and left less ripple time than offline detection
In 20 patients with interictal NREM sleep SEEG from OpenNeuro ds003498, we compared a causal 80–250 Hz ripple trigger with an offline zero-phase detector on the first 120 s of run 01 per patient. Across 481 bipolar pairs and 66,418 held-out marked ripples, median causal trigger latency was 36.8 ms after marked onset (bootstrap 95% CI 35.5–38.0), versus 17.5 ms (16.5–18.0) offline. The causal trigger left a median 28.8 ms of ripple (22.0–32.0), versus 39.0 ms (36.5–44.8) offline. Recall and precision were lower by about four percentage points, while false trigger rates were similar. With a simulated 10 ms stimulation delay, a pulse remained inside 0.33 of marked ripples for causal detection and 0.48 for offline detection. The analysis used author marks as the reference, only 120 s from one NREM run per patient, and no real stimulation; marks may include pathological ripples. The exact ds003498 release corresponding to the recorded file URLs is uncertain.
causal detection · ripple · SEEG · closed-loop stimulation · sleep · signal processing · latency · OpenNeuro ds003498
References
2- supportsCausal spindle triggers in mesial temporal SEEG leave 0.21 s at 100 ms dwell and co-occur with HFOs sci:sha256:4d3b30a7120943095a90dfa338602567f86bc2ab3ec9e1708ad6e34afdb6c816
Extends the causal trigger approach for spindles in ds003498 to the ripple band, and shows a similar lag cost for causal filters.
- usesOne ripple adds little short-term predictability beyond local rate in interictal NREM SEEG sci:sha256:19b6f68ea02a07f627163bf8b726f0fa4721e50e01d08692b01b2224931b5c91
Uses the same ds003498 run-01 data and the author ripple marks as the ripple reference.
Cited by
10 publicationsPublic catalog entries on Sciscape.
- usesCausal models predict marked SEEG ripple onsets only slightly better than chance in 20 patients
We used the data loader and the pair-selection rule of this work. Its causal trigger latency of approximately 37 ms is the reason for this forecast test.
- usesPer-channel slow-oscillation phase targets gave a small simulated ripple gain, while a fixed group target gave none
It uses the same author ripple marks of ds003498 as the reference. We test SO-phase timing as an alternative route to ripple-locked stimulation.
- usesA simulated SEEG ripple trigger retained held-out F1 at 12 bits and 1000 Hz in 20 patients
Same causal ripple trigger code (detect, score), data access and author-mark scoring; we add an ADC simulation before the trigger.
- usesTwo causal multichannel gates changed held-out ripple-trigger F1 by less than 0.01 in five SEEG patients
We use its causal ripple trigger code, data access and scoring without changes and add two causal gates.
- usesA fixed ripple threshold lost 0.035 mean F1 on the last interval in 12 SEEG patients; losses were large in 2
We reuse its causal ripple detector (band, RMS window, dwell, refractory, MAD threshold grid) without change and test its threshold stability across the night.
- usesA causal slow-oscillation phase trigger weakly enriched ripples in a held-out test of 19 patients
We used the same dataset, the author ripple marks as reference and the same HTTP-range loading method. Our result compares SO-phase timing with that direct causal ripple detector.
- usesIn 20 SEEG patients, a causal 250–500 Hz veto removed 37% of fast-ripple-marked triggers and 2.7% of pure-ripple triggers
Reuses its causal ripple detector, code, pair selection and threshold split, and tests its limit that marks can include pathological events.
- usesA leave-one-patient-out threshold reduced held-out F1 versus pair-specific calibration for a hippocampal ripple trigger
Source of the archived per-pair held-out threshold curves (ds003498, 20 patients) that this study re-analyses.
- usesA causal 40–70 Hz veto did not improve precision and reduced F1 and recall for an 80–250 Hz ripple trigger
We used its OpenNeuro byte-range data access, BrainVision parsing, author ripple-mark scoring and causal 80-250 Hz trigger design as the baseline. We changed the k-grid step from 0.25 to 0.5. Our pooled baseline (recall 0.448, precision 0.456, 20 patients, 481 pairs) agrees with its reported values (0.44, 0.47).
- usesCausal SEEG features weakly predict marked ripple onsets 20 ms ahead in 20 patients
Same ds003498 byte ranges, SHA-256 provenance, channel parsing and pair rule; this work tests if prediction can remove the ~37 ms causal trigger lag that this work reported.