Up to four class examples reduced mean response pattern error after two amplitude controls in 20 iEEG patients and 16 fMRI adults
We tested whether a few sounds correct response patterns beyond two specified amplitude directions. Separate public studies supplied human iEEG local field potential means and BOLD fMRI estimates. Four patients and four adults supplied development. A local code lock preceded tests in 20 other patients and 16 other adults. Fitting sounds set site scales, an acoustic model and two directions: uniform site activity and the mean fitting response. Correction used 0.25 times the mean error from up to four class examples. Test sounds were excluded from fitting and support. Mean projected error gains were 4.675 [2.501, 6.848] and 8.210 [7.705, 8.714] points of mean-prediction error. Brackets give 95% person Student limits. Positive counts were 16/20 and 16/16. The fixed shared gain test passed, Holm p=0.011818. The mean comparison with acoustically matched other-class examples failed, Holm p=0.082104. One positive electrical extreme made that mean uncertain. One-example electrical gain was also uncertain. Eight adults from another MRI acquisition gave gain 8.248 [7.448, 9.049]; possible person overlap was not resolved. An exact acoustic-interaction model gave a positive projected gain without a semantic variable. Thus the result supports this measured correction procedure, not semantic coding or a common neural cause. Different cohorts, sound sets, source screens, sparse electrical coverage, reference and vascular filters limit scope. Code, allowed inputs, compact statistics, all negative results, source pins, figures and repeat instructions are supplied.
human neuroscience · intracranial EEG · fMRI · auditory cortex · sound classes · response patterns · residual correction · subject inference · public data
References
15- usesA correction from up to four sound class examples lowered mean model error in 20 iEEG patients and 16 fMRI adults sci:sha256:b8582d8a1129c3fb7d8b16f1b24caaa8fdd525603206abb9a5b5390126c09ebe
We reuse its fixed acoustic model, folds, matched support selection, alpha, licensed electrical inputs and MIT code. We test a new training-only spatial projection endpoint.
- Residual sound maps agreed between a fixed iEEG panel and 20 fMRI adults under a rotation reference sci:sha256:c833e2811606e931c209b9a349ce18fad053c6a5d3d1e78c89d2e3e2cb2a83e9
Its residual geometry result informed the question. Our patient-labelled data and held-out prediction endpoint differ.
- usesUp to eight class examples lowered mean acoustic prediction error at fixed panel sizes in human iEEG and fMRI sci:sha256:893206318a69c0c2ee48c620d9c251b75b82c4fd886d1c167ba280a16c3e83c2
Licensed electrical derivative and source preparation in the input donor chain.
- usesSound labels lowered test error for averaged responses in 20 iEEG patients and 16 fMRI adults; separate BOLD fits gave no clear gain sci:sha256:495812fc27afe8b3e0fe67db013b2f4b72cecb708745c383879ab0479511bcad
Upstream loader and native MRI preparation; a source-model selection caution.
- usesFor fixed panels, readers trained on speech ranked other human voices above animal sounds in 20 iEEG patients and 16 fMRI adults sci:sha256:52e1af640ef1bde9f0d13b663fc7eb5e034cd4c61bc367532fc15aae8cef57dd
Original licensed electrical derivative in the verified SCI input chain.
- usesdoi:10.1016/j.cub.2026.07.063
Human electrical source study.
- usesdoi:10.5281/zenodo.20599001
Version 1.1 of the public Hect data release. Small files and selected neural windows are audited against this source.
- usesdoi:10.1152/jn.00588.2020
B2021 human MRI acquisition study.
- usesdoi:10.1016/j.neuron.2015.11.035
NH2015 human MRI acquisition study.
- usesdoi:10.1371/journal.pbio.3002366
Public author neural archive with both MRI studies. Exact tar bytes are pinned by SHA-256.
- doi:10.5281/zenodo.8349727
Versioned companion code release. This DOI is not claimed to archive the neural data tar.
- usesdoi:10.1016/j.cub.2022.01.069
Pinned acoustic features and 12-class reference labels for MRI.
- doi:10.1371/journal.pcbi.1005508
Established framework for neural response patterns and measurement dependence.
- doi:10.1371/journal.pbio.3001675
Patient-labelled natural-sound data inspected for clip-translation feasibility; no neural endpoints from that source were analyzed.
- doi:10.5281/zenodo.6544488
Version 1.1 inspected for label availability only. Its neural arrays do not enter the final test.
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesOne wrong example left uncertain mean calibration gain in 20 iEEG patients and positive gain in 16 fMRI adults
We reuse licensed electrical summaries, source readers, the fixed acoustic model and sound folds, and the training-only projection. We add exact one-wrong-example stress calculations.
- usesOne bit per auditory site gave mean four-class accuracy above chance in 20 iEEG patients and 16 fMRI adults
We reuse its verified licensed electrical inputs, source pins and MIT preparation and source-audit code.
- supportsPooled repeats raised the energy–variation rank relation at 24 sites in six SEEG patients and 16 fMRI adults
Its source audit documents cross-study sound-join and cohort limits.
- usesWith fixed references, model error profiles repeated in sound fMRI and late image MEG; iEEG transfer was not confirmed
We reuse its licensed patient-labelled electrical summaries, source records and MRI preparation helper. The new endpoint tests error-size transfer across people.
- usesFixed tests did not confirm shared animal–voice transfer in both directions in 16 iEEG patients and 16 fMRI adults
Upstream licensed electrical derivative and source-reader chain retained in the verified donor.
- usesAt 20 calibration sounds, mean class gains over acoustic neighbors were larger in association than early auditory parcels in 14 iEEG patients and 16 fMRI adults
Prior class-example assay and fixed 0.25 shrinkage motivate the new regional test. Prediction procedures differ.