Constructing a representational atlas for human cerebral cortex
Main Author: | Samuel A. Nastase |
---|---|
Format: | info Proceeding Journal |
Terbitan: |
, 2016
|
Subjects: | |
Online Access: |
https://zenodo.org/record/4016446 |
Daftar Isi:
- Human cerebral cortex encodes rich perceptual and semantic information in high-dimensional neural representational spaces. However, fine-grained functional topographies vary across individual brains. We describe several critical steps toward constructing a representational atlas for human cerebral cortex. A representational atlas characterizes cortical fields in terms of what information they explicitly encodeāi.e., their representational geometry. Correspondingly, a representational template can be constructed by registering individual brains according to local representational geometry rather than anatomy. Current data-driven approaches, collectively called hyperalignment, rotate neural representational spaces into a common template space that preserves fine-grained representational content and circumvents the idiosyncratic functional topographies of individual brains. These approaches use either time-locked responses to dynamic, naturalistic stimuli or inter-voxel correlations to register neural representational spaces. Importantly, naturalistic audiovisual paradigms increase ecological validity, more fully sample neural representational spaces, and can reveal neural representations that are obscured by reliance on static, controlled stimuli. A representational atlas must also accommodate how varying behavioral goals alter neural responses. We demonstrate that attentional allocation reshapes distributed response patterns to make task-relevant representations more discriminable. Ultimately, a representational atlas provides a mapping between the representational spaces of individual brains, and can provide a basis for characterizing individual differences and predicting neural responses in novel individuals.