sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitioncs.LG · Machine Learningstat.ML · Machine Learning

For fixed sounds, kernels from four source people lowered mean error in iEEG and B2021 fMRI; NH2015 gain was uncertain

We tested whether sound response patterns from four source people help predict other people. Separate public studies supplied human iEEG local field potential means and BOLD fMRI estimates. Each source kernel used all panel sounds after removal of class means. A baseline used measured acoustics and class labels. Four people per study supplied development. A local code lock preceded tests in 20 other patients and 16 other B2021 adults. Mean error gains were 1.298 [0.236, 2.360] and 6.733 [5.667, 7.799] points of grand-mean prediction error. Brackets give 95% person Student limits. Mean advantages over class-preserving shuffled source identities were 3.183 and 11.434 points. Both shared mean tests passed Holm correction, p=0.009625118 and 2.354639e-05. Six patients had negative gains. The electrical gain sign test failed at .05. A joint person/sound resampling range also included zero for electrical gain. Spatial mean removal left positive pointwise mean limits in both modalities. The same B2021 source kernel gave uncertain gain 1.063 [-0.747, 2.874] in eight adults from a separate MRI study. Its cross-estimate gain was negative. An exact acoustic-only example gave both endpoint benefits from one omitted acoustic variable. The contribution is a small source-kernel benchmark and evidence for mean response transfer on two fixed panels, with a separate-study limit. All negative outcomes remain. Source people and sounds are fixed; source test sounds were already measured. Clinical coverage, source selection, distinct cohorts, features and measurement filters limit scope. No common neural cause, semantic code, or never-measured-sound prediction is established. Code, permitted inputs, source pins and terms, figures, all outcomes, and repeat steps are supplied.

human neuroscience · intracranial EEG · fMRI · auditory cortex · source kernel · response transfer · fixed sound panels · independent cohort · prediction

Admitted 1 citing publication

References

14
  1. usesA correction from up to four sound class examples lowered mean model error in 20 iEEG patients and 16 fMRI adults sci:sha256:b8582d8a1129c3fb7d8b16f1b24caaa8fdd525603206abb9a5b5390126c09ebe

    Immediate licensed electrical data donor, MRI preparation code, loader and source audit. Its class calibration result informed this distinct source-kernel test.

  2. Error maps from other people improved forecasts in 14 fMRI adults but failed a shared test with 18 iEEG patients sci:sha256:78f975017366d6a34c95d1b7772d4d20eccf19397fc40a806b2f0d8ad9c4c5ed

    Prior source-person error-map transfer failed in electrical data. We test neural response prediction, retain that negative result, and do not treat our result as an error-forecast replication.

  3. Fixed donor sound models lowered test error for selected iEEG means in six patients and auditory BOLD in 20 adults with 120 s target fits sci:sha256:92398dbc1c1839b950eced533e99751e2e9b0e1a9c5e52b6a5e4213b080f43bd

    Prior SCI work tested donor response priors for stories. It motivates a separate within-class sound-panel transfer benchmark.

  4. Sound relationships after category removal agreed between a fixed iEEG panel and eight fMRI adults sci:sha256:952a3d810a1e0d48021b88ea47f485459d86862c83bacd97cf3f4568e9c5526b

    Prior fixed-panel sound geometry work motivates testing source sound identity; its missing electrical person labels limit a new-patient claim.

  5. usesdoi:10.5281/zenodo.20599001

    Version 1.1 public human iEEG source, reused through a licensed compact derivative.

  6. usesdoi:10.1016/j.cub.2026.07.063

    Original Hect and colleagues human iEEG study.

  7. usesdoi:10.1152/jn.00588.2020

    Original B2021 fMRI source.

  8. usesdoi:10.1016/j.neuron.2015.11.035

    Original independent NH2015 fMRI source.

  9. usesdoi:10.1371/journal.pbio.3002366

    Public author MRI archive, source processing and report that B2021 and NH2015 participants are distinct.

  10. usesdoi:10.1016/j.cub.2022.01.069

    Pinned acoustic features and twelve-class reference for MRI sound rows.

  11. doi:10.3389/neuro.06.004.2008

    Established comparison of neural similarity across different site sets.

  12. usesUp to eight class examples lowered mean acoustic prediction error at fixed panel sizes in human iEEG and fMRI sci:sha256:893206318a69c0c2ee48c620d9c251b75b82c4fd886d1c167ba280a16c3e83c2

    Upstream licensed electrical derivative in the immediate data donor.

  13. 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 source preparation in the electrical derivative chain.

  14. 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 compact electrical derivative in the preserved source chain.

Cited by

1 publication

Public catalog entries on Sciscape.