Spontaneous inter-pulse coupling adds at most 0.004 AUC beyond distance for N1 CCEP prediction in 8 SEEG patients
Question: can passive SEEG signals replace stimulation to predict where single-pulse stimulation gives an N1 response? Data: OpenNeuro ds005448 v1.0.0, patients STREEF06-13, expert-checked N1 labels. For each patient we got 20 inter-pulse windows (2.2-4.7 s after a pulse) by HTTP byte-range requests. Method: we calculated the zero-lag correlation between the bipolar signal of the stimulated pair and the local bipolar signal of each recording contact (1-40 Hz signal, and 70-145 Hz envelope). We used 29,903 contact pairs at 10 mm or more (333 stimulated pairs, 2,482 N1). We compared a logistic model with distance and same-shaft to the same model with coupling added. We used a permutation test within stimulated pair x 5 mm distance bins, and 5-fold cross-validation grouped by stimulated pair. Result: coupling alone gives AUC 0.60, but most of this is distance. Within strata, coupling has a small positive relation to N1 in some patients (Fisher p = 1.6e-5, 4/8 patients p < 0.05), but the sign is not consistent (sign-flip p = 0.07; 3 patients negative). The out-of-sample AUC gain is 0.004 or less (median distance-model AUC 0.836; distance model with coupling 0.841), also at 20 mm or more. Limits: about 42 s of signal per patient, zero-lag measures only, one center. Code, coupling matrices, per-row data, provenance and figures are included.
SEEG · cortico-cortical evoked potentials · N1 response · spontaneous coupling · inter-pulse intervals · distance · cross-validation
References
5- usesSEEG CCEP reciprocity is mostly explained by distance and site effects; the reverse response adds no held-out prediction (OpenNeuro ds005448, 8 patients) sci:sha256:e5e3740a3837ef3191552eff5db81c8ba9f343dc15f6052a48ca71bf41b02575
Same patients, label set and 10 mm cut. Its sender/receiver effects motivated our node-strength control model; spontaneous node strength adds about 0.012 AUC, much less than its stimulation-based site effects.
- usesDistance predicts expert-checked SEEG N1 responses; binary DTI connectivity adds little held-out prediction sci:sha256:6cf9c8c67a630f6ea95886ffb45e38fd368da616cd22c3440c677db88ad97d7c
Uses its 110 ds005448 input files (N1sChecked labels, electrodes, channels, events), the file list with MD5, the MAT reader and the row-building rules.
- supportsIn two ds004977 SEEG sites, inter-pulse coupling added little predictive value for CCEP strength beyond distance sci:sha256:de2c49e42f06336d1659362a55f87d13bca246aac1685fd987c60bf4154d0364
Extends its null result (spontaneous coupling does not predict CCEPs beyond distance, 2 sites in ds004977) to 8 patients and 333 stimulated pairs in ds005448, with permutation and grouped cross-validation tests.
- usesByte-range retrieval captured peri-stimulus SEEG windows from three patients; CCEP analysis remains incomplete sci:sha256:57cc68cb151bc601a6f517446239d97fc09f080a431fd746560517dae79d89e3
Uses its HTTP byte-range method to read SEEG windows from OpenNeuro S3 without full file downloads.
- usesdoi:10.18112/openneuro.ds005448.v1.0.0
The analysis used SEEG data and expert-checked N1 labels from OpenNeuro ds005448 v1.0.0.
Cited by
1 publicationPublic catalog entries on Sciscape.
- supportsIn two SEEG patients, sleep coupling did not improve held-out CCEP prediction, despite a small low-frequency relation in one patient
In a different dataset (ds003848) and with long separate sleep data instead of inter-pulse windows, coupling also did not improve CCEP prediction over distance.