Smooth scalar-on-image regression via spatial Bayesian variable selection

Jeff Goldsmith, Lei Huang, Ciprian M Crainiceanu

Research output: Contribution to journalArticle

Abstract

We develop scalar-on-image regression models when images are registered multidimensional manifolds. We propose a fast and scalable Bayes' inferential procedure to estimate the image coefficient. The central idea is the combination of an Ising prior distribution, which controls a latent binary indicator map, and an intrinsic Gaussian Markov random field, which controls the smoothness of the nonzero coefficients. The model is fit using a single-site Gibbs sampler, which allows fitting within minutes for hundreds of subjects with predictor images containing thousands of locations. The code is simple and is provided in the online Appendix (see the "Supplementary Materials" section). We apply this method to a neuroimaging study where cognitive outcomes are regressed on measures of white-matter microstructure at every voxel of the corpus callosum for hundreds of subjects.

Original languageEnglish (US)
Pages (from-to)46-64
Number of pages19
JournalJournal of Computational and Graphical Statistics
Volume23
Issue number1
DOIs
StatePublished - 2014

Fingerprint

Bayesian Variable Selection
Regression
Scalar
Gaussian Markov Random Field
Bayes Procedures
Neuroimaging
Image Model
Gibbs Sampler
Voxel
Coefficient
Prior distribution
Ising
Microstructure
Predictors
Smoothness
Regression Model
Binary
Estimate
Variable selection
Spatial regression

Keywords

  • Binary Markov random field
  • Gaussian Markov random field
  • Markov chain Monte Carlo

ASJC Scopus subject areas

  • Discrete Mathematics and Combinatorics
  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

Smooth scalar-on-image regression via spatial Bayesian variable selection. / Goldsmith, Jeff; Huang, Lei; Crainiceanu, Ciprian M.

In: Journal of Computational and Graphical Statistics, Vol. 23, No. 1, 2014, p. 46-64.

Research output: Contribution to journalArticle

@article{848b0e8ec17f49599226487c457cfa9b,
title = "Smooth scalar-on-image regression via spatial Bayesian variable selection",
abstract = "We develop scalar-on-image regression models when images are registered multidimensional manifolds. We propose a fast and scalable Bayes' inferential procedure to estimate the image coefficient. The central idea is the combination of an Ising prior distribution, which controls a latent binary indicator map, and an intrinsic Gaussian Markov random field, which controls the smoothness of the nonzero coefficients. The model is fit using a single-site Gibbs sampler, which allows fitting within minutes for hundreds of subjects with predictor images containing thousands of locations. The code is simple and is provided in the online Appendix (see the {"}Supplementary Materials{"} section). We apply this method to a neuroimaging study where cognitive outcomes are regressed on measures of white-matter microstructure at every voxel of the corpus callosum for hundreds of subjects.",
keywords = "Binary Markov random field, Gaussian Markov random field, Markov chain Monte Carlo",
author = "Jeff Goldsmith and Lei Huang and Crainiceanu, {Ciprian M}",
year = "2014",
doi = "10.1080/10618600.2012.743437",
language = "English (US)",
volume = "23",
pages = "46--64",
journal = "Journal of Computational and Graphical Statistics",
issn = "1061-8600",
publisher = "American Statistical Association",
number = "1",

}

TY - JOUR

T1 - Smooth scalar-on-image regression via spatial Bayesian variable selection

AU - Goldsmith, Jeff

AU - Huang, Lei

AU - Crainiceanu, Ciprian M

PY - 2014

Y1 - 2014

N2 - We develop scalar-on-image regression models when images are registered multidimensional manifolds. We propose a fast and scalable Bayes' inferential procedure to estimate the image coefficient. The central idea is the combination of an Ising prior distribution, which controls a latent binary indicator map, and an intrinsic Gaussian Markov random field, which controls the smoothness of the nonzero coefficients. The model is fit using a single-site Gibbs sampler, which allows fitting within minutes for hundreds of subjects with predictor images containing thousands of locations. The code is simple and is provided in the online Appendix (see the "Supplementary Materials" section). We apply this method to a neuroimaging study where cognitive outcomes are regressed on measures of white-matter microstructure at every voxel of the corpus callosum for hundreds of subjects.

AB - We develop scalar-on-image regression models when images are registered multidimensional manifolds. We propose a fast and scalable Bayes' inferential procedure to estimate the image coefficient. The central idea is the combination of an Ising prior distribution, which controls a latent binary indicator map, and an intrinsic Gaussian Markov random field, which controls the smoothness of the nonzero coefficients. The model is fit using a single-site Gibbs sampler, which allows fitting within minutes for hundreds of subjects with predictor images containing thousands of locations. The code is simple and is provided in the online Appendix (see the "Supplementary Materials" section). We apply this method to a neuroimaging study where cognitive outcomes are regressed on measures of white-matter microstructure at every voxel of the corpus callosum for hundreds of subjects.

KW - Binary Markov random field

KW - Gaussian Markov random field

KW - Markov chain Monte Carlo

UR - http://www.scopus.com/inward/record.url?scp=84901808735&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84901808735&partnerID=8YFLogxK

U2 - 10.1080/10618600.2012.743437

DO - 10.1080/10618600.2012.743437

M3 - Article

AN - SCOPUS:84901808735

VL - 23

SP - 46

EP - 64

JO - Journal of Computational and Graphical Statistics

JF - Journal of Computational and Graphical Statistics

SN - 1061-8600

IS - 1

ER -