English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

A fundamental challenge in neuroscience is to understand what structure inthe world is represented in spatially distributed patterns of neural activityfrom multiple single-trial measurements. This is often accomplished by learninga simple, linear transformations between neural features and features of thesensory stimuli or motor task. While successful in some early sensoryprocessing areas, linear mappings are unlikely to be ideal tools forelucidating nonlinear, hierarchical representations of higher-order brain areasduring complex tasks, such as the production of speech by humans. Here, weapply deep networks to predict produced speech syllables from cortical surfaceelectric potentials recorded from human sensorimotor cortex. We found that deepnetworks had higher decoding prediction accuracy compared to baseline models,and also exhibited greater improvements in accuracy with increasing datasetsize. We further demonstrate that deep network's confusions revealedhierarchical latent structure in the neural data, which recapitulated theunderlying articulatory nature of speech motor control. Finally, we used deepnetworks to compare task-relevant information in different neural frequencybands, and found that the high-gamma band contains the vast majority ofinformation relevant for the speech prediction task, with little-to-noadditional contribution from lower-frequencies. Together, these resultsdemonstrate the utility of deep networks as a data analysis tool forneuroscience.

Deep learning as a tool for neural data analysis: speech classification and cross-frequency coupling in human sensorimotor cortex