Deriving statistical significance maps for support vector regression using medical imaging data

Bilwaj Gaonkar, Aristeidis Sotiras, Christos Davatzikos

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Regression analysis involves predicting a continuos variable using imaging data. The Support Vector Regression (SVR) algorithm has previously been used in addressing regression analysis in neuroimaging. However, identifying the regions of the image that the SVR uses to model the dependence of a target variable remains an open problem. It is an important issue when one wants to biologically interpret the meaning of a pattern that predicts the variable(s) of interest, and therefore to understand normal or pathological process. One possible approach to the identification of these regions is the use of permutation testing. Permutation testing involves 1) generation of a large set of 'null SVR models' using randomly permuted sets of target variables, and 2) comparison of the SVR model trained using the original labels to the set of null models. These permutation tests often require prohibitively long computational time. Recent work in support vector classification shows that it is possible to analytically approximate the results of permutation testing in medical image analysis. We propose an analogous approach to approximate permutation testing based analysis for support vector regression with medical imaging data. In this paper we present 1) the theory behind our approximation, and 2) experimental results using two real datasets.

Original languageEnglish (US)
Title of host publicationProceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013
Pages13-16
Number of pages4
DOIs
StatePublished - 2013
Externally publishedYes
Event2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013 - Philadelphia, PA, United States
Duration: Jun 22 2013Jun 24 2013

Other

Other2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013
CountryUnited States
CityPhiladelphia, PA
Period6/22/136/24/13

Fingerprint

Medical imaging
Testing
Regression analysis
Neuroimaging
Image analysis
Labels
Imaging techniques

Keywords

  • Permutation testing
  • Support Vector Regression

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Biomedical Engineering

Cite this

Gaonkar, B., Sotiras, A., & Davatzikos, C. (2013). Deriving statistical significance maps for support vector regression using medical imaging data. In Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013 (pp. 13-16). [6603545] https://doi.org/10.1109/PRNI.2013.13

Deriving statistical significance maps for support vector regression using medical imaging data. / Gaonkar, Bilwaj; Sotiras, Aristeidis; Davatzikos, Christos.

Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013. 2013. p. 13-16 6603545.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Gaonkar, B, Sotiras, A & Davatzikos, C 2013, Deriving statistical significance maps for support vector regression using medical imaging data. in Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013., 6603545, pp. 13-16, 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013, Philadelphia, PA, United States, 6/22/13. https://doi.org/10.1109/PRNI.2013.13
Gaonkar B, Sotiras A, Davatzikos C. Deriving statistical significance maps for support vector regression using medical imaging data. In Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013. 2013. p. 13-16. 6603545 https://doi.org/10.1109/PRNI.2013.13
Gaonkar, Bilwaj ; Sotiras, Aristeidis ; Davatzikos, Christos. / Deriving statistical significance maps for support vector regression using medical imaging data. Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013. 2013. pp. 13-16
@inproceedings{c84614b447954e6b949f66ee93ab37d9,
title = "Deriving statistical significance maps for support vector regression using medical imaging data",
abstract = "Regression analysis involves predicting a continuos variable using imaging data. The Support Vector Regression (SVR) algorithm has previously been used in addressing regression analysis in neuroimaging. However, identifying the regions of the image that the SVR uses to model the dependence of a target variable remains an open problem. It is an important issue when one wants to biologically interpret the meaning of a pattern that predicts the variable(s) of interest, and therefore to understand normal or pathological process. One possible approach to the identification of these regions is the use of permutation testing. Permutation testing involves 1) generation of a large set of 'null SVR models' using randomly permuted sets of target variables, and 2) comparison of the SVR model trained using the original labels to the set of null models. These permutation tests often require prohibitively long computational time. Recent work in support vector classification shows that it is possible to analytically approximate the results of permutation testing in medical image analysis. We propose an analogous approach to approximate permutation testing based analysis for support vector regression with medical imaging data. In this paper we present 1) the theory behind our approximation, and 2) experimental results using two real datasets.",
keywords = "Permutation testing, Support Vector Regression",
author = "Bilwaj Gaonkar and Aristeidis Sotiras and Christos Davatzikos",
year = "2013",
doi = "10.1109/PRNI.2013.13",
language = "English (US)",
isbn = "9780769550619",
pages = "13--16",
booktitle = "Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013",

}

TY - GEN

T1 - Deriving statistical significance maps for support vector regression using medical imaging data

AU - Gaonkar, Bilwaj

AU - Sotiras, Aristeidis

AU - Davatzikos, Christos

PY - 2013

Y1 - 2013

N2 - Regression analysis involves predicting a continuos variable using imaging data. The Support Vector Regression (SVR) algorithm has previously been used in addressing regression analysis in neuroimaging. However, identifying the regions of the image that the SVR uses to model the dependence of a target variable remains an open problem. It is an important issue when one wants to biologically interpret the meaning of a pattern that predicts the variable(s) of interest, and therefore to understand normal or pathological process. One possible approach to the identification of these regions is the use of permutation testing. Permutation testing involves 1) generation of a large set of 'null SVR models' using randomly permuted sets of target variables, and 2) comparison of the SVR model trained using the original labels to the set of null models. These permutation tests often require prohibitively long computational time. Recent work in support vector classification shows that it is possible to analytically approximate the results of permutation testing in medical image analysis. We propose an analogous approach to approximate permutation testing based analysis for support vector regression with medical imaging data. In this paper we present 1) the theory behind our approximation, and 2) experimental results using two real datasets.

AB - Regression analysis involves predicting a continuos variable using imaging data. The Support Vector Regression (SVR) algorithm has previously been used in addressing regression analysis in neuroimaging. However, identifying the regions of the image that the SVR uses to model the dependence of a target variable remains an open problem. It is an important issue when one wants to biologically interpret the meaning of a pattern that predicts the variable(s) of interest, and therefore to understand normal or pathological process. One possible approach to the identification of these regions is the use of permutation testing. Permutation testing involves 1) generation of a large set of 'null SVR models' using randomly permuted sets of target variables, and 2) comparison of the SVR model trained using the original labels to the set of null models. These permutation tests often require prohibitively long computational time. Recent work in support vector classification shows that it is possible to analytically approximate the results of permutation testing in medical image analysis. We propose an analogous approach to approximate permutation testing based analysis for support vector regression with medical imaging data. In this paper we present 1) the theory behind our approximation, and 2) experimental results using two real datasets.

KW - Permutation testing

KW - Support Vector Regression

UR - http://www.scopus.com/inward/record.url?scp=84885231977&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84885231977&partnerID=8YFLogxK

U2 - 10.1109/PRNI.2013.13

DO - 10.1109/PRNI.2013.13

M3 - Conference contribution

SN - 9780769550619

SP - 13

EP - 16

BT - Proceedings - 2013 3rd International Workshop on Pattern Recognition in Neuroimaging, PRNI 2013

ER -