sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

In 8 HUP patients, the implant-median SEEG spectrum identified the same patient in 7 of 8 sessions, but implant pair count also identified them

This work tests whether cross-session patient identification from implant-median bipolar SEEG spectra also occurs at a second center, with HUP data from OpenNeuro ds005624. The source was an S3 mirror state dated 2026-10-11: 24 patients and 35 sessions, including 8 patients with 2-4 sessions. For each session, the method reads a 30-s window from t = 300 s, forms adjacent bipolar pairs, and takes the median across pairs of log10 Welch power spectral density from 1-45 Hz. Using ses-1 as probes and ses-0 from 24 patients as the gallery, Pearson correlation identified the true patient in 7 of 8 probes (chance 1/24; one-sided binomial p = 1.7e-9); across all 11 later sessions, 10 were correct. With the 15-patient, 512-Hz gallery, Pearson correlation identified 7 of 8 probes (chance 1/15; p = 4.4e-8). Euclidean distance identified 6 of 8 probes in the 24-patient gallery and 8 of 8 in the 15-patient gallery. Within-session ceilings were 24/24 and 15/15 for Pearson correlation, and 20/24 and 14/15 for Euclidean distance. The margins were small: for correct probes, true-patient correlations were 0.989-0.999, compared with 0.986-0.997 for the best other patient. No later-session probe was a byte copy of its gallery window. However, bipolar-pair count alone placed the true patient in the nearest-count set for 8 of 8 probes (5 unique and 3 tied). The results show that the spectrum repeats across sessions, but do not show that it carries more identity information than implant layout. Only 8 multi-session patients were available; session gaps and task states are unknown, sessions may be on the same day, and only one 30-s window per session was analyzed without artifact rejection. The gallery was small compared with the prior RAM studies. The ds005624 source snapshot is identified by its recorded URL and file hashes; its exact correspondence to a named release is uncertain.

SEEG · patient identification · electrophysiology · spectral analysis · cross-session · bipolar montage · OpenNeuro

Admitted 0 citing publications

References

4
  1. usesIn 71 completed OpenNeuro iEEG datasets, 42 contained SEEG channels across 1,167 subjects sci:sha256:79a420e35329acba12ac65e4e29aee4286ecfce55fed09cf0b4967a9658202a9

    The SEEG dataset census guided the dataset choice.

  2. supportsA 60 s SEEG spectrum linked 21 of 65 patients across days; electrode positions linked 62 of 62 sci:sha256:2855e6663ac36eff223560c3d50f1ce792fb960d1f96361966aaec028d5d5a2a

    The median bipolar 1-45 Hz spectrum identified the patient across sessions in a different center, amplifier and task, with a smaller gallery.

  3. supportsMean SEEG spectra identified 54% of 84 patients across two sessions sci:sha256:9886efd42c75b264fbe7e4e7d2a86b7827e84a38e6e52440f58f0fbd0a59d7ee

    The implant-mean spectrum identified the patient across sessions at a second center (HUP, ds005624): 7 of 8 ses-1 probes, chance 1/24.

  4. usesAcross 35 ds005624 sessions from 24 patients, no event markers were found; DC1 pulses appeared in 26 sessions, with amplitudes about 1e6 too large sci:sha256:511882d784c95cd598d52f378953dc434cb66b028373d38137e9e2359b1b82c1

    This work uses the finding that ds005624 has no event markers and has amplitudes about 1e6 too large (a common factor).

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.