Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 3 min ago
Peer-reviewedCC BY-NC-NDNature Communications

Shared latent representations of speech production for cross-patient speech decoding

  • Zachary Spalding1,2
  • Suseendrakumar Duraivel1,3,2
  • Shervin Rahimpour4,5,6
  • Charles Wang1,2
  • Katrina J. Barth1,2
  • C. Schmitz1,2
  • Shivanand Pramod Lad4,2
  • Allan H. Friedman4,2
  • Derek G. Southwell7,4,8,2
  • Jonathan Viventi9,10,11,12,2
  • Gregory B. Cogan13,14,15,16,17,2
17 affiliations
  1. 1Department of Biomedical Engineering, Duke University, Durham, NC, USA
  2. 2Duke University
  3. 3McGovern Institute for Brain Research, Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA
  4. 4Department of Neurosurgery, Duke School of Medicine, Durham, NC, USA
  5. 5Department of Neurosurgery, University of Utah, Salt Lake City, UT, USA
  6. 6University of Utah
  7. 7Department of Neurobiology, Duke School of Medicine, Durham, NC, USA
  8. 8Duke Comprehensive Epilepsy Center, Duke University, Durham, NC, USA
  9. 9Department of Biomedical Engineering, Duke University, Durham, NC, USA. [email protected]
  10. 10Department of Neurobiology, Duke School of Medicine, Durham, NC, USA. [email protected]
  11. 11Department of Neurosurgery, Duke School of Medicine, Durham, NC, USA. [email protected]
  12. 12Duke Comprehensive Epilepsy Center, Duke University, Durham, NC, USA. [email protected]
  13. 13Center for Cognitive Neuroscience, Duke University, Durham, NC, USA. [email protected]
  14. 14Department of Biomedical Engineering, Duke University, Durham, NC, USA. [email protected]
  15. 15Department of Neurology, Duke School of Medicine, Durham, NC, USA. [email protected]
  16. 16Department of Neurosurgery, Duke School of Medicine, Durham, NC, USA. [email protected]
  17. 17Duke Comprehensive Epilepsy Center, Duke University, Durham, NC, USA. [email protected]
Open at sourceDOICite

Abstract

Speech brain-computer interfaces (BCIs) can restore communication in individuals with neuromotor disorders who are unable to speak. However, current speech BCIs limit patient usability and successful deployment by requiring large volumes of patient-specific data collected over long periods of time. A promising solution to facilitate usability and accelerate their successful deployment is to combine data from multiple patients. This has proven difficult, however, due to differences in user neuroanatomy, varied placement of electrode arrays, and sparse sampling of targeted anatomy. Here, by aligning patient-specific neural data to a shared latent space, we show that speech BCIs can be trained on data combined across patients. Using canonical correlation analysis and high-density micro-electrocorticography (μECoG), we uncovered shared neural latent dynamics with preserved micro-scale speech information. This approach enabled cross-patient decoding models to achieve improved accuracies relative to patient-specific models facilitated by the high resolution and broad coverage of μECoG. Our findings support future speech BCIs that are more accurate and rapidly deployable, ultimately improving the quality of life for people with impaired communication from neuromotor disorders.

Related

  1. FloREN: Decoding Immune Regulatory Networks through Interpretable Graph Transformer Patient Representations
    Chloé Quéré, Valérie Devauchelle-Pensec, Philippe Thuillier +23
    European Journal of Nuclear Medicine and Molecular ImagingPreprintCC BY-NC-NDImmunologyHAL
  2. Multilingual Brain Decoding Reveals Shared Cortical Semantic Geometry
    Yi Guo, Yong Hu, Michael Kwok-Po Ng +1
    arXivPreprintarXiv licenceFull textNeurosciencearXiv
  3. Shared neural geometries for bilingual semantic representations in human hippocampal neurons
    Xinyuan Yan, Ana G. Chavez, Melissa C. Franch +20
    CellPeer-reviewedLicence not statedFull text8 citedNeuroscienceOpenAlex
  4. Many Brains, One Geometry: A Shared Visual-Semantic Space for Cross-Dataset fMRI Decoding
    Moein Khajehnejad, Michelangelo Tronti, Forough Habibollahi +3
    arXivPreprintCC BY 4.0Full textBioinformatics & computational biologyNeurosciencearXiv
  5. A shared manifold for scalable temporal representations in self-paced timing
    M. Serrano, M. Castelli, Y. Peng +2
    bioRxivPreprintCC BYFull textNeurosciencebioRxiv

Cite

BibTeXRIS
BibTeX
@article{spalding2026shared,
  title = {{Shared latent representations of speech production for cross-patient speech decoding}},
  author = {Zachary Spalding and Suseendrakumar Duraivel and Shervin Rahimpour and Charles Wang and Katrina J. Barth and C. Schmitz and Shivanand Pramod Lad and Allan H. Friedman and Derek G. Southwell and Jonathan Viventi and Gregory B. Cogan},
  journal = {Nature Communications},
  year = {2026},
  month = jul,
  doi = {10.1038/s41467-026-75455-1},
  url = {https://doi.org/10.1038/s41467-026-75455-1},
  note = {CC BY-NC-ND},
}
RIS
TY  - JOUR
TI  - Shared latent representations of speech production for cross-patient speech decoding
AU  - Zachary Spalding
AU  - Suseendrakumar Duraivel
AU  - Shervin Rahimpour
AU  - Charles Wang
AU  - Katrina J. Barth
AU  - C. Schmitz
AU  - Shivanand Pramod Lad
AU  - Allan H. Friedman
AU  - Derek G. Southwell
AU  - Jonathan Viventi
AU  - Gregory B. Cogan
PY  - 2026
DA  - 2026/07/16
JO  - Nature Communications
DO  - 10.1038/s41467-026-75455-1
UR  - https://doi.org/10.1038/s41467-026-75455-1
AB  - Speech brain-computer interfaces (BCIs) can restore communication in individuals with neuromotor disorders who are unable to speak. However, current speech BCIs limit patient usability and successful deployment by requiring large volumes of patient-specific data collected over long periods of time. A promising solution to facilitate usability and accelerate their successful deployment is to combine data from multiple patients. This has proven difficult, however, due to differences in user neuroanatomy, varied placement of electrode arrays, and sparse sampling of targeted anatomy. Here, by aligning patient-specific neural data to a shared latent space, we show that speech BCIs can be trained on data combined across patients. Using canonical correlation analysis and high-density micro-electrocorticography (μECoG), we uncovered shared neural latent dynamics with preserved micro-scale speech information. This approach enabled cross-patient decoding models to achieve improved accuracies relative to patient-specific models facilitated by the high resolution and broad coverage of μECoG. Our findings support future speech BCIs that are more accurate and rapidly deployable, ultimately improving the quality of life for people with impaired communication from neuromotor disorders.
SN  - 2041-1723
ER  -