Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification

Bilwaj Gaonkar, Christos Davatzikos

Research output: Contribution to journalArticle

Abstract

Multivariate pattern analysis (MVPA) methods such as support vector machines (SVMs) have been increasingly applied to fMRI and sMRI analyses, enabling the detection of distinctive imaging patterns. However, identifying brain regions that significantly contribute to the classification/group separation requires computationally expensive permutation testing. In this paper we show that the results of SVM-permutation testing can be analytically approximated. This approximation leads to more than a thousandfold speedup of the permutation testing procedure, thereby rendering it feasible to perform such tests on standard computers. The speedup achieved makes SVM based group difference analysis competitive with standard univariate group difference analysis methods.

Original languageEnglish (US)
Pages (from-to)270-283
Number of pages14
JournalNeuroImage
Volume78
DOIs
StatePublished - Sep 2013
Externally publishedYes

Fingerprint

Multivariate Analysis
Magnetic Resonance Imaging
Brain
Support Vector Machine

Keywords

  • Neuroimaging analysis
  • Statistical inference
  • SVM

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification. / Gaonkar, Bilwaj; Davatzikos, Christos.

In: NeuroImage, Vol. 78, 09.2013, p. 270-283.

Research output: Contribution to journalArticle

@article{3e1f13d7209049ec8a89deae73eeb229,
title = "Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification",
abstract = "Multivariate pattern analysis (MVPA) methods such as support vector machines (SVMs) have been increasingly applied to fMRI and sMRI analyses, enabling the detection of distinctive imaging patterns. However, identifying brain regions that significantly contribute to the classification/group separation requires computationally expensive permutation testing. In this paper we show that the results of SVM-permutation testing can be analytically approximated. This approximation leads to more than a thousandfold speedup of the permutation testing procedure, thereby rendering it feasible to perform such tests on standard computers. The speedup achieved makes SVM based group difference analysis competitive with standard univariate group difference analysis methods.",
keywords = "Neuroimaging analysis, Statistical inference, SVM",
author = "Bilwaj Gaonkar and Christos Davatzikos",
year = "2013",
month = "9",
doi = "10.1016/j.neuroimage.2013.03.066",
language = "English (US)",
volume = "78",
pages = "270--283",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Analytic estimation of statistical significance maps for support vector machine based multi-variate image analysis and classification

AU - Gaonkar, Bilwaj

AU - Davatzikos, Christos

PY - 2013/9

Y1 - 2013/9

N2 - Multivariate pattern analysis (MVPA) methods such as support vector machines (SVMs) have been increasingly applied to fMRI and sMRI analyses, enabling the detection of distinctive imaging patterns. However, identifying brain regions that significantly contribute to the classification/group separation requires computationally expensive permutation testing. In this paper we show that the results of SVM-permutation testing can be analytically approximated. This approximation leads to more than a thousandfold speedup of the permutation testing procedure, thereby rendering it feasible to perform such tests on standard computers. The speedup achieved makes SVM based group difference analysis competitive with standard univariate group difference analysis methods.

AB - Multivariate pattern analysis (MVPA) methods such as support vector machines (SVMs) have been increasingly applied to fMRI and sMRI analyses, enabling the detection of distinctive imaging patterns. However, identifying brain regions that significantly contribute to the classification/group separation requires computationally expensive permutation testing. In this paper we show that the results of SVM-permutation testing can be analytically approximated. This approximation leads to more than a thousandfold speedup of the permutation testing procedure, thereby rendering it feasible to perform such tests on standard computers. The speedup achieved makes SVM based group difference analysis competitive with standard univariate group difference analysis methods.

KW - Neuroimaging analysis

KW - Statistical inference

KW - SVM

UR - http://www.scopus.com/inward/record.url?scp=84877323945&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84877323945&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2013.03.066

DO - 10.1016/j.neuroimage.2013.03.066

M3 - Article

VL - 78

SP - 270

EP - 283

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

ER -