Brain Connectivity-Informed Regularization Methods for Regression

Marta Karas, Damian Brzyski, Mario Dzemidzic, Joaquín Goñi, David Kareken, Timothy W. Randolph, Jaroslaw Harezlak

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

One of the challenging problems in brain imaging research is a principled incorporation of information from different imaging modalities. Frequently, each modality is analyzed separately using, for instance, dimensionality reduction techniques, which result in a loss of mutual information. We propose a novel regularization method to estimate the association between the brain structure features and a scalar outcome within the linear regression framework. Our regularization technique provides a principled approach to use external information from the structural brain connectivity and inform the estimation of the regression coefficients. Our proposal extends the classical Tikhonov regularization framework by defining a penalty term based on the structural connectivity-derived Laplacian matrix. Here, we address both theoretical and computational issues. The approach is first illustrated using simulated data and compared with other penalized regression methods. We then apply our regularization method to study the associations between the alcoholism phenotypes and brain cortical thickness using a diffusion imaging derived measure of structural connectivity. Using the proposed methodology in 148 young male subjects with a risk for alcoholism, we found a negative associations between cortical thickness and drinks per drinking day in bilateral caudal anterior cingulate cortex, left lateral OFC, and left precentral gyrus.

Original languageEnglish (US)
Pages (from-to)1-44
Number of pages44
JournalStatistics in Biosciences
DOIs
StateAccepted/In press - Dec 6 2017

Fingerprint

Regularization Method
Brain
Connectivity
Regression
Alcoholism
Imaging
Imaging techniques
Modality
Gyrus Cinguli
Frontal Lobe
Negative Association
Penalized Regression
Neuroimaging
Drinking
Laplacian Matrix
Regularization Technique
Linear Models
Information use
Tikhonov Regularization
Dimensionality Reduction

Keywords

  • Brain connectivity
  • Brain structure
  • Laplacian matrix
  • Linear regression
  • Penalized methods
  • Structured penalties

ASJC Scopus subject areas

  • Statistics and Probability
  • Biochemistry, Genetics and Molecular Biology (miscellaneous)

Cite this

Brain Connectivity-Informed Regularization Methods for Regression. / Karas, Marta; Brzyski, Damian; Dzemidzic, Mario; Goñi, Joaquín; Kareken, David; Randolph, Timothy W.; Harezlak, Jaroslaw.

In: Statistics in Biosciences, 06.12.2017, p. 1-44.

Research output: Contribution to journalArticle

Karas, Marta ; Brzyski, Damian ; Dzemidzic, Mario ; Goñi, Joaquín ; Kareken, David ; Randolph, Timothy W. ; Harezlak, Jaroslaw. / Brain Connectivity-Informed Regularization Methods for Regression. In: Statistics in Biosciences. 2017 ; pp. 1-44.
@article{8dda87f386ba4f2f93bec7daa35f7904,
title = "Brain Connectivity-Informed Regularization Methods for Regression",
abstract = "One of the challenging problems in brain imaging research is a principled incorporation of information from different imaging modalities. Frequently, each modality is analyzed separately using, for instance, dimensionality reduction techniques, which result in a loss of mutual information. We propose a novel regularization method to estimate the association between the brain structure features and a scalar outcome within the linear regression framework. Our regularization technique provides a principled approach to use external information from the structural brain connectivity and inform the estimation of the regression coefficients. Our proposal extends the classical Tikhonov regularization framework by defining a penalty term based on the structural connectivity-derived Laplacian matrix. Here, we address both theoretical and computational issues. The approach is first illustrated using simulated data and compared with other penalized regression methods. We then apply our regularization method to study the associations between the alcoholism phenotypes and brain cortical thickness using a diffusion imaging derived measure of structural connectivity. Using the proposed methodology in 148 young male subjects with a risk for alcoholism, we found a negative associations between cortical thickness and drinks per drinking day in bilateral caudal anterior cingulate cortex, left lateral OFC, and left precentral gyrus.",
keywords = "Brain connectivity, Brain structure, Laplacian matrix, Linear regression, Penalized methods, Structured penalties",
author = "Marta Karas and Damian Brzyski and Mario Dzemidzic and Joaqu{\'i}n Go{\~n}i and David Kareken and Randolph, {Timothy W.} and Jaroslaw Harezlak",
year = "2017",
month = "12",
day = "6",
doi = "10.1007/s12561-017-9208-x",
language = "English (US)",
pages = "1--44",
journal = "Statistics in Biosciences",
issn = "1867-1764",
publisher = "Springer New York",

}

TY - JOUR

T1 - Brain Connectivity-Informed Regularization Methods for Regression

AU - Karas, Marta

AU - Brzyski, Damian

AU - Dzemidzic, Mario

AU - Goñi, Joaquín

AU - Kareken, David

AU - Randolph, Timothy W.

AU - Harezlak, Jaroslaw

PY - 2017/12/6

Y1 - 2017/12/6

N2 - One of the challenging problems in brain imaging research is a principled incorporation of information from different imaging modalities. Frequently, each modality is analyzed separately using, for instance, dimensionality reduction techniques, which result in a loss of mutual information. We propose a novel regularization method to estimate the association between the brain structure features and a scalar outcome within the linear regression framework. Our regularization technique provides a principled approach to use external information from the structural brain connectivity and inform the estimation of the regression coefficients. Our proposal extends the classical Tikhonov regularization framework by defining a penalty term based on the structural connectivity-derived Laplacian matrix. Here, we address both theoretical and computational issues. The approach is first illustrated using simulated data and compared with other penalized regression methods. We then apply our regularization method to study the associations between the alcoholism phenotypes and brain cortical thickness using a diffusion imaging derived measure of structural connectivity. Using the proposed methodology in 148 young male subjects with a risk for alcoholism, we found a negative associations between cortical thickness and drinks per drinking day in bilateral caudal anterior cingulate cortex, left lateral OFC, and left precentral gyrus.

AB - One of the challenging problems in brain imaging research is a principled incorporation of information from different imaging modalities. Frequently, each modality is analyzed separately using, for instance, dimensionality reduction techniques, which result in a loss of mutual information. We propose a novel regularization method to estimate the association between the brain structure features and a scalar outcome within the linear regression framework. Our regularization technique provides a principled approach to use external information from the structural brain connectivity and inform the estimation of the regression coefficients. Our proposal extends the classical Tikhonov regularization framework by defining a penalty term based on the structural connectivity-derived Laplacian matrix. Here, we address both theoretical and computational issues. The approach is first illustrated using simulated data and compared with other penalized regression methods. We then apply our regularization method to study the associations between the alcoholism phenotypes and brain cortical thickness using a diffusion imaging derived measure of structural connectivity. Using the proposed methodology in 148 young male subjects with a risk for alcoholism, we found a negative associations between cortical thickness and drinks per drinking day in bilateral caudal anterior cingulate cortex, left lateral OFC, and left precentral gyrus.

KW - Brain connectivity

KW - Brain structure

KW - Laplacian matrix

KW - Linear regression

KW - Penalized methods

KW - Structured penalties

UR - http://www.scopus.com/inward/record.url?scp=85037168544&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85037168544&partnerID=8YFLogxK

U2 - 10.1007/s12561-017-9208-x

DO - 10.1007/s12561-017-9208-x

M3 - Article

AN - SCOPUS:85037168544

SP - 1

EP - 44

JO - Statistics in Biosciences

JF - Statistics in Biosciences

SN - 1867-1764

ER -