TY - JOUR
T1 - Brain kernel
T2 - A new spatial covariance function for fMRI data
AU - Wu, Anqi
AU - Nastase, Samuel A.
AU - Baldassano, Christopher A.
AU - Turk-Browne, Nicholas
AU - Norman, Kenneth A.
AU - Engelhardt, Barbara E.
AU - Pillow, Jonathan W.
N1 - Funding Information:
This work was supported by the McKnight Foundation (JP), NSF CAREER Award IIS-1150186 (JP), the Simons Collaboration on the Global Brain (SCGB AWD1004351) (JP) and a J. Insley Blair Pyne Fund Award (to JP, BE, KN).
Publisher Copyright:
© 2021 The Author(s)
PY - 2021/12/15
Y1 - 2021/12/15
N2 - A key problem in functional magnetic resonance imaging (fMRI) is to estimate spatial activity patterns from noisy high-dimensional signals. Spatial smoothing provides one approach to regularizing such estimates. However, standard smoothing methods ignore the fact that correlations in neural activity may fall off at different rates in different brain areas, or exhibit discontinuities across anatomical or functional boundaries. Moreover, such methods do not exploit the fact that widely separated brain regions may exhibit strong correlations due to bilateral symmetry or the network organization of brain regions. To capture this non-stationary spatial correlation structure, we introduce the brain kernel, a continuous covariance function for whole-brain activity patterns. We define the brain kernel in terms of a continuous nonlinear mapping from 3D brain coordinates to a latent embedding space, parametrized with a Gaussian process (GP). The brain kernel specifies the prior covariance between voxels as a function of the distance between their locations in embedding space. The GP mapping warps the brain nonlinearly so that highly correlated voxels are close together in latent space, and uncorrelated voxels are far apart. We estimate the brain kernel using resting-state fMRI data, and we develop an exact, scalable inference method based on block coordinate descent to overcome the challenges of high dimensionality (10-100K voxels). Finally, we illustrate the brain kernel's usefulness with applications to brain decoding and factor analysis with multiple task-based fMRI datasets.
AB - A key problem in functional magnetic resonance imaging (fMRI) is to estimate spatial activity patterns from noisy high-dimensional signals. Spatial smoothing provides one approach to regularizing such estimates. However, standard smoothing methods ignore the fact that correlations in neural activity may fall off at different rates in different brain areas, or exhibit discontinuities across anatomical or functional boundaries. Moreover, such methods do not exploit the fact that widely separated brain regions may exhibit strong correlations due to bilateral symmetry or the network organization of brain regions. To capture this non-stationary spatial correlation structure, we introduce the brain kernel, a continuous covariance function for whole-brain activity patterns. We define the brain kernel in terms of a continuous nonlinear mapping from 3D brain coordinates to a latent embedding space, parametrized with a Gaussian process (GP). The brain kernel specifies the prior covariance between voxels as a function of the distance between their locations in embedding space. The GP mapping warps the brain nonlinearly so that highly correlated voxels are close together in latent space, and uncorrelated voxels are far apart. We estimate the brain kernel using resting-state fMRI data, and we develop an exact, scalable inference method based on block coordinate descent to overcome the challenges of high dimensionality (10-100K voxels). Finally, we illustrate the brain kernel's usefulness with applications to brain decoding and factor analysis with multiple task-based fMRI datasets.
KW - Brain decoding
KW - Brain kernel
KW - Factor modeling
KW - Gaussian process
KW - Latent variable model
KW - Resting-state fmri
KW - Task fmri
UR - http://www.scopus.com/inward/record.url?scp=85118580422&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85118580422&partnerID=8YFLogxK
U2 - 10.1016/j.neuroimage.2021.118580
DO - 10.1016/j.neuroimage.2021.118580
M3 - Article
C2 - 34740792
AN - SCOPUS:85118580422
SN - 1053-8119
VL - 245
JO - NeuroImage
JF - NeuroImage
M1 - 118580
ER -