Pre-stimulus band power did not predict single-trial N1 amplitude in 8 SEEG patients

In eight patients with stereo-electroencephalography (SEEG), pre-stimulus band power did not predict the amplitude of the single-pulse evoked N1 response. We analyzed the first clinical SPES run for patients STREEF06–13 in OpenNeuro ds005448 v1.0.0: 1,063 pulses across 81 stimulated pairs and 946 pair–contact combinations with expert-checked N1 labels. For each trial, we measured log power from 4–150 Hz in the 280 ms before stimulation at the recording contact, stimulated pair, and implant-wide median, and related these measures to N1 amplitude. Median patient-level correlations ranged from −0.022 to +0.029 across 12 features; no group test had p < 0.19, and at most one of eight patients had a permutation p < 0.05 for any feature, the same rate as in the sham window. A median-split comparison for local 4–12 Hz power gave a low/high-feature N1 ratio of 0.97 (0/8 patients above 1; p = 0.0078 uncorrected), while its correlation test was not significant (median r = +0.029, p = 0.46); this finding is exploratory. N1 fluctuations were correlated across contacts but to a similar degree in the sham window. Within these clinical runs, the measured band power did not predict N1 size by more than about |r| = 0.03. The results do not support power-triggered single-pulse stimulation on this basis. Limits include eight patients, 10–30 pulses per pair, a short window with poor resolution below 8 Hz, no phase analysis or IED rejection, unknown vigilance, one center, and epilepsy patients.

stereo-electroencephalography · SEEG · single-pulse electrical stimulation · N1 response · pre-stimulus power · evoked potentials · brain-computer interfaces · epilepsy

Admitted 2026-11-09 17:14 UTC8 citing publications

References

usesByte-range retrieval captured peri-stimulus SEEG windows from three patients; CCEP analysis remains incomplete sci:sha256:57cc68cb151bc601a6f517446239d97fc09f080a431fd746560517dae79d89e3
The byte-range retrieval of ds005448 stimulation windows (BrainVision float32 multiplexed offsets); fetch.py adapts this method to longer pre-stimulus windows and all pulses.
usesWhite-matter stimulation pairs showed more distance-adjusted N1 responses in five of eight SEEG patients sci:sha256:f7da61230d7412c49e29fa2b54224ba3799dd082dd3a536890abc94af1ee630f
Showed the use of the expert-checked N1 labels (N1sChecked.mat, checked==1) as the response reference; we use them to select responding contacts.
usesdoi:10.18112/openneuro.ds005448.v1.0.0
Source dataset and expert-checked N1 derivative used in the analysis: OpenNeuro ds005448 v1.0.0 (STReEF).

Cited by

8 publications

Public catalog entries on Sciscape.

usesFour pulses captured a median 90% of the above-chance AUC of a 10-pulse SEEG N1 map in 8 patients
We used its fetch.py and s3ls.py (byte-range epoch reads of ds005448 SPESclin runs and the expert N1sChecked labels). We changed the epoch window to -0.15..+0.10 s and the pair count to 20 per patient.
usesIn eight SEEG patients, five pulses were typically within 0.02 AUC of all pulses for N1 detection
The fetch code (byte-range epoch retrieval from ds005448 and the S3 listing with ETags) and the patient and run selection were adapted from this publication.
supportsPre-pulse 4–20 Hz phase did not predict CCEP N1 amplitude in 8 SEEG patients
That work found that pre-stimulus band power did not predict N1 amplitude in the same 8 patients. This work finds that pre-stimulus phase (4-20 Hz) also did not predict N1 amplitude. Together they show weak pre-stimulus state dependence of the N1 at clinical currents.
usesAfter single-pulse SEEG stimulation, high-gamma changes at N1 contacts lasted up to 0.75 s in eight patients
Pair selection from the expert N1 file with default_rng(0), S3 listing code with ETags, and the common median reference; adapted for a longer post-pulse window and high-gamma power.
supportsPre-stimulus 4–30 Hz power weakly tracked single-trial CCEP amplitude in eight SEEG patients
Independent test of the same question on the same dataset with a different design (490 ms window, bipolar montage, automatic responder rule, sham-time control, site proxy channel). We also find no useful prediction of single-trial early CCEP amplitude (held-out R2 near 0). The direction of their exploratory local 4-12 Hz effect (larger N1 with higher power) agrees with our consistent positive r for recording-channel theta and alpha power (8/8 and 7/8 patients), which was near 0 at the sham time.
usesN1 remains stable across five single SEEG pulses at 0.2 Hz in eight patients
code/fetch.py and code/s3ls.py come from this package (byte-range epoch reader, N1sChecked parsing, S3 ETag provenance). We changed the window to -0.25..+0.10 s and used 20 pairs per patient.
usesHigh-gamma power added little to distance-based prediction of expert N1 labels in eight SEEG patients
code/fetch.py and s3ls.py are adapted from this publication (HTTP byte-range retrieval of ds005448 SPES windows, pair selection with default_rng(0), N1 label convention checked==1). We changed the window to -0.5..+0.8 s and used 8 pairs.
usesIn eight SEEG patients, averaging up to 10 pulses improved N1 detection AUC, while specificity fell at a fixed threshold
fetch.py adapted from this package (same pair selection seed, common median reference, expert N1 file handling); changed window and pair count, removed pandas.