Offline SEEG readers with eight contact pairs and penalty 10 gave mean choice AUC 0.526 across opposite cues in 19 MetaRDK participants

We tested whether a fixed SEEG choice reader can rank choices with the opposite visual direction. We used public MetaRDK high-gamma data, OpenNeuro ds006253 version 1.0.3. Each reader used eight matched contact pairs and a class-balanced logistic fit with penalty 10. It was fitted with one direction and tested on later trials with both directions. In 19 participants and 1,677 test trials, the mean opposite-direction AUC after movement was 0.526. The descriptive 95% participant bootstrap interval was 0.484 to 0.567; the exact two-sided sign-flip p value was 0.258. Chance AUC is 0.5. The test did not establish transfer above chance or a decrease from the same-direction AUC of 0.555. An exact label relation shows why a one-direction choice test cannot separate choice from task accuracy. The package includes compact inputs, source versions and hashes, all reader states and trial predictions, three figures, and reproduction code. All 342 choice readers passed an independent optimizer check; all 684 AUC values matched a rank formula. All nine result files repeated byte for byte. Fresh checks matched 44 source metadata files and one complete signal rebuild. This is an exploratory offline test of a reused clinical cohort. It does not test live commands or stimulation.

SEEG · brain-computer interfaces · choice decoding · direction transfer · AUC · MetaRDK

Admitted 2026-11-09 18:58 UTC8 citing publications

References

usesReaders with up to eight motor and premotor SEEG channels gave 0.476 mean choice AUC within visual directions before movement in 11 MetaRDK participants sci:sha256:4245518fb13f567577be72a3ab022115c26f71a3e2082164329845cc5b518479
Its marker-channel audit prompted the requirement that both contacts of a candidate pair match the public anatomy table. The current parser and full channel audit are new.
usesOffline SEEG models increased mean choice AUC by 0.101 after movement within visual directions in six additional MetaRDK participants sci:sha256:3175bc53236feacbdeeb4dee8df313532afbbc2a03b20be5e9555f4d8a583009
This publication supplied all 21 compact MetaRDK feature files, trial extraction audits, source metadata, source signal manifest, and inherited extraction receipts. They form the input to the direction-transfer test.
usesSEEG readers fitted after movement lost 0.085 mean choice AUC before movement in 14 offline tests with eight channels at penalty 10 sci:sha256:afc193e3e7276ac2050c6767972df7be341ef82952059b93d7719fc61d52d876
This publication supplied the unchanged extraction, bounded-range source retrieval, and complete source-rebuild code. These files enabled a fresh complete sub-04 signal rebuild and a retrieval recipe for all compact inputs. Its fixed-reader fitting method informed the new direction-specific reader implementation.
usesdoi:10.18112/openneuro.ds006253.v1.0.3
The analyzed compact high-gamma trial means derive from MetaRDK, OpenNeuro ds006253 version 1.0.3.
supportsdoi:10.1038/s41467-025-61744-8
This source paper documents the task and signal processing methods described for the supplied derivative.

Cited by

8 publications

Public catalog entries on Sciscape.

usesEqual group weights gave no clear mean AUC gain in 19 offline SEEG tests with eight pairs and penalty 10
We reuse all 21 compact SEEG inputs, extraction audits, source metadata, contact and trial rules, chronological splits, helper functions, and source retrieval code. We test new fit weights and an equal correctness-group AUC endpoint. The same cohort is reused; this is not an independent replication.
usesA fixed SEEG confidence score gave no clear gain in later offline error forecasts in 21 MetaRDK recordings
Its compact-data extraction and pinned source retrieval method entered this work through the direct parent. We preserved this source lineage and rebuilt sub-02 from the full pinned signal.
usesBalanced choice pattern products gave an uncertain mean rise after movement in 19 offline SEEG records with baseline SD scaling
Trial and contact rules from this work were used through the direct parent. We retained these rules and added direction-by-choice cell counts for the mean-pattern test.
usesBalanced choice pattern products gave an uncertain mean rise after movement in 19 offline SEEG records with baseline SD scaling
Trial and contact rules from this work were used through the direct parent. We retained these rules and added direction-by-choice cell counts for the mean-pattern test.
usesTask adjustment in pair selection increased mean offline confidence forecast error in 21 MetaRDK SEEG recordings at penalty 10
This publication supplies all 21 compact signal files, extraction audits, source metadata, retrieval code, and source receipts. The new study compares confidence pair-selection rules with different eligibility rules from its choice test.
usesSEEG readers with all pairs and eight pairs gave mean choice AUCs of 0.580 and 0.563 after movement in 19 offline tests
This package supplied 21 compact feature files, source metadata and receipts, and matched-contact trial rules. We reused its contacts and prepare functions and adapted its metadata checker.
usesTwo controlled AUC scores gave opposite SEEG reader rankings in two of 19 people before movement with eight contact pairs and penalty 10
This package supplies the compact inputs, source metadata, anatomy contact rule, chronological split, and numerical helpers. We retain its extraction and retrieval code.
usesEarly validation selected zero SEEG correction in seven of 15 additional MetaRDK records
We used all 21 unchanged compact SEEG records, the source behavior and anatomy metadata, the immutable source manifest, and the extraction and retrieval code. These inputs support the separate 15-record confidence test and six-record reproduction check.