High-order graph matching based feature selection for Alzheimer's disease identification.

Feng Liu, Heung-Il Suk, Chong Yaw Wee, Huafu Chen, Dinggang Shen

Research output: Chapter in Book/Report/Conference proceedingChapter

27 Citations (Scopus)

Abstract

One of the main limitations of l1-norm feature selection is that it focuses on estimating the target vector for each sample individually without considering relations with other samples. However, it's believed that the geometrical relation among target vectors in the training set may provide useful information, and it would be natural to expect that the predicted vectors have similar geometric relations as the target vectors. To overcome these limitations, we formulate this as a graph-matching feature selection problem between a predicted graph and a target graph. In the predicted graph a node is represented by predicted vector that may describe regional gray matter volume or cortical thickness features, and in the target graph a node is represented by target vector that include class label and clinical scores. In particular, we devise new regularization terms in sparse representation to impose high-order graph matching between the target vectors and the predicted ones. Finally, the selected regional gray matter volume and cortical thickness features are fused in kernel space for classification. Using the ADNI dataset, we evaluate the effectiveness of the proposed method and obtain the accuracies of 92.17% and 81.57% in AD and MCI classification, respectively.

Original languageEnglish
Title of host publicationMedical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention
Pages311-318
Number of pages8
Volume16
EditionPt 2
Publication statusPublished - 2013 Jan 1
Externally publishedYes

Fingerprint

Alzheimer Disease
Gray Matter
Datasets

ASJC Scopus subject areas

  • Medicine(all)

Cite this

Liu, F., Suk, H-I., Wee, C. Y., Chen, H., & Shen, D. (2013). High-order graph matching based feature selection for Alzheimer's disease identification. In Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention (Pt 2 ed., Vol. 16, pp. 311-318)

High-order graph matching based feature selection for Alzheimer's disease identification. / Liu, Feng; Suk, Heung-Il; Wee, Chong Yaw; Chen, Huafu; Shen, Dinggang.

Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention. Vol. 16 Pt 2. ed. 2013. p. 311-318.

Research output: Chapter in Book/Report/Conference proceedingChapter

Liu, F, Suk, H-I, Wee, CY, Chen, H & Shen, D 2013, High-order graph matching based feature selection for Alzheimer's disease identification. in Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention. Pt 2 edn, vol. 16, pp. 311-318.
Liu F, Suk H-I, Wee CY, Chen H, Shen D. High-order graph matching based feature selection for Alzheimer's disease identification. In Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention. Pt 2 ed. Vol. 16. 2013. p. 311-318
Liu, Feng ; Suk, Heung-Il ; Wee, Chong Yaw ; Chen, Huafu ; Shen, Dinggang. / High-order graph matching based feature selection for Alzheimer's disease identification. Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention. Vol. 16 Pt 2. ed. 2013. pp. 311-318
@inbook{38d35dece89b47859b067be99702cbe4,
title = "High-order graph matching based feature selection for Alzheimer's disease identification.",
abstract = "One of the main limitations of l1-norm feature selection is that it focuses on estimating the target vector for each sample individually without considering relations with other samples. However, it's believed that the geometrical relation among target vectors in the training set may provide useful information, and it would be natural to expect that the predicted vectors have similar geometric relations as the target vectors. To overcome these limitations, we formulate this as a graph-matching feature selection problem between a predicted graph and a target graph. In the predicted graph a node is represented by predicted vector that may describe regional gray matter volume or cortical thickness features, and in the target graph a node is represented by target vector that include class label and clinical scores. In particular, we devise new regularization terms in sparse representation to impose high-order graph matching between the target vectors and the predicted ones. Finally, the selected regional gray matter volume and cortical thickness features are fused in kernel space for classification. Using the ADNI dataset, we evaluate the effectiveness of the proposed method and obtain the accuracies of 92.17{\%} and 81.57{\%} in AD and MCI classification, respectively.",
author = "Feng Liu and Heung-Il Suk and Wee, {Chong Yaw} and Huafu Chen and Dinggang Shen",
year = "2013",
month = "1",
day = "1",
language = "English",
volume = "16",
pages = "311--318",
booktitle = "Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention",
edition = "Pt 2",

}

TY - CHAP

T1 - High-order graph matching based feature selection for Alzheimer's disease identification.

AU - Liu, Feng

AU - Suk, Heung-Il

AU - Wee, Chong Yaw

AU - Chen, Huafu

AU - Shen, Dinggang

PY - 2013/1/1

Y1 - 2013/1/1

N2 - One of the main limitations of l1-norm feature selection is that it focuses on estimating the target vector for each sample individually without considering relations with other samples. However, it's believed that the geometrical relation among target vectors in the training set may provide useful information, and it would be natural to expect that the predicted vectors have similar geometric relations as the target vectors. To overcome these limitations, we formulate this as a graph-matching feature selection problem between a predicted graph and a target graph. In the predicted graph a node is represented by predicted vector that may describe regional gray matter volume or cortical thickness features, and in the target graph a node is represented by target vector that include class label and clinical scores. In particular, we devise new regularization terms in sparse representation to impose high-order graph matching between the target vectors and the predicted ones. Finally, the selected regional gray matter volume and cortical thickness features are fused in kernel space for classification. Using the ADNI dataset, we evaluate the effectiveness of the proposed method and obtain the accuracies of 92.17% and 81.57% in AD and MCI classification, respectively.

AB - One of the main limitations of l1-norm feature selection is that it focuses on estimating the target vector for each sample individually without considering relations with other samples. However, it's believed that the geometrical relation among target vectors in the training set may provide useful information, and it would be natural to expect that the predicted vectors have similar geometric relations as the target vectors. To overcome these limitations, we formulate this as a graph-matching feature selection problem between a predicted graph and a target graph. In the predicted graph a node is represented by predicted vector that may describe regional gray matter volume or cortical thickness features, and in the target graph a node is represented by target vector that include class label and clinical scores. In particular, we devise new regularization terms in sparse representation to impose high-order graph matching between the target vectors and the predicted ones. Finally, the selected regional gray matter volume and cortical thickness features are fused in kernel space for classification. Using the ADNI dataset, we evaluate the effectiveness of the proposed method and obtain the accuracies of 92.17% and 81.57% in AD and MCI classification, respectively.

UR - http://www.scopus.com/inward/record.url?scp=84897573822&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84897573822&partnerID=8YFLogxK

M3 - Chapter

VL - 16

SP - 311

EP - 318

BT - Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention

ER -