sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.ME · Methodology

A shared finger profile test failed in six ECoG patients and six fMRI adults; unequal errors gave 50.5% mean matching in a model

We tested person matching from five finger responses in separate public human ECoG and fMRI studies. Development used two patients and one MRI adult. A local code lock preceded tests in six other patients and six other adults. Each recording half supplied ten cross-repeat finger distances. Centered unit-length profiles gave mean matching advantages of 0.069975 and 0.152525. Their 95% delete-person jackknife intervals were [-0.147475,0.287424] and [-0.049059,0.354109]. Two-direction matching accuracy was 25.0% and 58.3%, against 16.7%. Only the MRI assignment tests passed. The fixed shared test failed, p=0.331944. All negative scores and fixed checks remain. Two constructed models give interpretation limits. In one model, every person had the same true profile. Independent zero-mean errors differed in size. Normalization gave 50.5208% mean accuracy over all 4096 sign cases for six people. The fixed one-cohort rule passed in 420 cases. Actual pattern crossproducts realized the distance model within 1.14e-13. Another model gave perfect matching from different fixed site samples of one identical full population. It also passed a disjoint-panel check. Neither model was fitted to human errors. Thus matched records alone need not identify different true neural geometry. This study does not establish positive cross-modal neural generalization or causality. Small cohorts, clinical coverage, different actions, source processing and unmeasured nuisance traces limit scope. The package preserves code, source pins and terms, plans, complete results, figures, model cases and repeat instructions. Source neural arrays remain external.

human neuroscience · electrocorticography · fMRI · finger movement · representational geometry · person matching · measurement noise · spatial sampling · negative result

Admitted 2 citing publications

References

11
  1. usesA fixed test did not confirm shared finger pair structure in five ECoG patients and seven fMRI adults sci:sha256:d22893af7b32a68f335d0ab3987472b958c2c0f3e092f4ead3cb8847692089c4

    We build on its source manifest, data access code and trial boundary logic. Its failed shared distance-template test motivates a different record matching endpoint.

  2. usesFixed tests did not confirm four-axis finger gains in four ECoG patients and six fMRI adults sci:sha256:c7e2ab2de928e8ff187bbffb8ce92b6eb048152e30c642d36a4a8f65a3f86f64

    We reuse its immediate access-code version, verified source pins and Welch feature reference. The present person matching endpoint and interpretation models are new.

  3. doi:10.1371/journal.pcbi.1002655

    Original human finger-flexion ECoG experiment and required task attribution.

  4. doi:10.1038/s41562-019-0678-3

    Public ECoG source library.

  5. doi:10.1038/nn.4038

    Source of the released seven-person 7 T M1 finger data.

  6. doi:10.1016/j.neuroimage.2017.08.051

    PCM tutorial data access and representational model context.

  7. doi:10.1038/nn.4135

    Prior work on person matching from brain measurements.

  8. doi:10.1038/s41598-018-25089-1

    Prior work on own versus other person similarity and identifiability.

  9. doi:10.7554/eLife.82566

    Inference and crossvalidation for representational geometries.

  10. doi:10.7554/eLife.62324

    Prior evidence that physiological and motion nuisance signals support fMRI person matching.

  11. doi:10.7554/eLife.106812.1

    Prior measurement theory for sampling-related distortion of representational geometry; reviewed preprint version 1.

Cited by

2 publications

Public catalog entries on Sciscape.