A fixed four-band power shape matched 8 of 12 patients between two verbal memory sessions in public hippocampal SEEG data

Can a fixed power shape link a hippocampal SEEG record to the same patient in another session? We tested public OpenNeuro ds004752 version 1.0.1. Four centered band values and a fixed nearest-distance rule matched 8 of 12 session 02 records to their patients in a session 01 gallery. The exact label probability was 0.000000601 under conditional patient label exchangeability. A check within sampling rate groups gave p = 0.00000551. The unchanged gallery gave 48.7% accuracy in 25 later records from nine patients, with equal patient weight. These data contain repeatable patient information. The design cannot assign it to a stable neural trait or measure bias in a clinical model. The package contains 49 compact spectra, event and electrode files, pinned raw source versions, all results, and code. Fresh raw checks covered one trial per record. Separate feature code and a direct count of all 12! label mappings reproduced the main calculations. All 24 result files matched a second offline run.

stereoelectroencephalography · SEEG · patient matching · power spectrum · working memory · patient leakage · reproducibility

Admitted 2026-11-09 18:09 UTC2 citing publications

References

usesTesting did not detect a linear beta load association in 8 patients with 19 later hippocampal SEEG sessions, assuming independent patients and symmetric null signs sci:sha256:4113fe5a22ccdbf6344ee222a6694d8f35692027b1f1810fc928b39ed8721e48
This work uses all 25 available later-session spectra from nine patients under the broad atlas pair rule. It uses the source metadata, raw examples, pinned object records, and source hash checks. It keeps the initial gallery fixed and tests patient match performance in later records.
usesA sign flip test did not detect a mean load association in late hippocampal theta excess across two SEEG sessions in 12 patients sci:sha256:164e078c43accf3c5735784c9e3221d25b50de923c8b24c0bb701e24102c555c
This work uses the matched pair rule, initial-session trial spectra, source metadata, headers, raw examples, and source hash records from this publication. It asks a new patient match question and does not reuse the theta load test.
usesdoi:10.18112/openneuro.ds004752.v1.0.1
This is the exact version of OpenNeuro ds004752 used for the analyzed recordings and source metadata.
supportsdoi:10.7554/eLife.78677
This paper describes the source cohort and verbal working memory task used in OpenNeuro ds004752.

Cited by

2 publications

Public catalog entries on Sciscape.

usesWith fixed mean rules, theta contrast changes had less session agreement than fixation levels in 12 hippocampal SEEG patients
We read the report on patient matching from maintenance spectra. It helped to motivate the distinction between phase level information and agreement in a task phase change. No numerical input from that report was used.
supersedesA fixed four-band power shape matched 8 of 12 patients between two verbal memory sessions in public hippocampal SEEG data
This version corrects the package verifier to exclude the root service metadata folder added by sci fetch. The first archived package was downloaded and all 24 numerical result files reproduced. The scientific results and source data are unchanged.