sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitioneess.SP · Signal Processing

Single-pulse SEEG responses identify stimulation sites at least 20 mm apart from distant contacts

Question: for an SEEG BCI that writes with electrical pulses, how many stimulation sites can the evoked response at distant contacts tell apart? Data: OpenNeuro ds005448 v1.0.0, 8 patients, 12 sites per patient, 10 pulses per site (5 of each polarity). Method: leave-one-pulse-out nearest-centroid classifier on 10-ms bins at 15-145 ms, from contacts 15 mm or more from all stimulated contacts; exact sign-flip tests over patients. Results: per-patient median AUC for 66 site pairs is 0.68-1.0 (pre-pulse control 0.39-0.48; p = 0.0078, 8/8 patients). AUC increases with the distance between sites: median 0.70 below 10 mm, 0.84 at 10-20 mm, 0.97 at 20-40 mm, 0.99 above 40 mm. Pairs that share a contact have a median AUC of 0.69. In 12-way identification (7 patients), mean accuracy is 0.425 (chance 0.083) and the bias-corrected mutual information is 0.12-1.70 bits per pulse (median 0.76). Post-hoc controls: removal of clipping channels, removal of the window mean, and a 50-150 ms window keep the result (median AUC 0.89-1.0). Templates from one polarity classify pulses of the reversed polarity as well as same-polarity templates do (median AUC 0.92 vs 0.92). Thus an artifact that changes sign with polarity is not necessary for the result; an artifact part that keeps its sign is not excluded. Limits: one center; few pulses per site; the time gap between the stimulation blocks of different sites is not controlled; the plug-in MI correction is approximate. The package has code, sidecar data, per-pair results, hashes, three figures and reproduction steps.

SEEG · single-pulse electrical stimulation · site identification · evoked response · AUC · cross-polarity control · intracranial electrophysiology

Admitted 4 citing publications

References

5
  1. usesSpearman-Brown predicts 17 pulses for reliable SEEG CCEP maps in eight patients sci:sha256:fe216b6966dcfe69347d13b0a2edc531a498a061e307c8744c9d379a070c5dda

    Our fetch code (code/fetch_pulses.py) is derived from the fetch code base of this publication (run choice, events parsing, byte-range reads).

  2. supportsSEEG stimulation sites show more distinct network responses as distance increases sci:sha256:dbefb32e29c21cd9ef85476095214956a2daacce9d0ecec04d30de84c0a67b39

    With a different classifier, pairwise AUC for 528 site pairs and far channels only (>= 15 mm), we also find that identification falls for sites closer than about 10 mm (median AUC 0.70) and is near ceiling beyond 20 mm. We share the time-gap confound that this publication names; our cross-polarity split does not remove it.

  3. supportsOpposite-polarity decoding identified 10 SEEG stimulation sites above chance at 10 mm in eight patients sci:sha256:ee9d18671f0c5d871f954c8535a01346c569ba353ac11078c37f735f78e6e9a9

    Same question on the same dataset. With templates from the opposite polarity and a 15 mm channel exclusion, our 12-way accuracy is 0.33-0.38 (chance 0.083). This agrees with their opposite-polarity 10-way accuracy of 0.35 at 15 mm. We add that pairwise AUC with cross-polarity templates equals the same-polarity AUC (median 0.92 vs 0.92), so pairwise site discrimination at far contacts does not need a polarity-locked artifact.

  4. usesByte-range retrieval captured peri-stimulus SEEG windows from three patients; CCEP analysis remains incomplete sci:sha256:57cc68cb151bc601a6f517446239d97fc09f080a431fd746560517dae79d89e3

    We use the HTTP byte-range method of this publication to read the BrainVision SPES pulse windows of ds005448 from the OpenNeuro S3 bucket without the full files.

  5. usesdoi:10.18112/openneuro.ds005448.v1.0.0

    The analysis used OpenNeuro ds005448 v1.0.0 (STReEF), as recorded in the source provenance.

Cited by

4 publications

Public catalog entries on Sciscape.