Multi-modality canonical feature selection for Alzheimer's disease diagnosis

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

13 Citations (Scopus)

Abstract

Feature selection has been commonly regarded as an effective method to lessen the problem of high dimension and low sample size in medical image analysis. In this paper, we propose a novel multi-modality canonical feature selection method. Unlike the conventional sparse Multi-Task Learning (MTL) based feature selection method that mostly considered only the relationship between target response variables, we further consider the correlations between features of different modalities by projecting them into a canonical space determined by canonical correlation analysis. We call the projections as canonical representations. By setting the canonical representations as regressors in a sparse least square regression framework and by further penalizing the objective function with a new canonical regularizer on the weight coefficient matrix, we formulate a multi-modality canonical feature selection method. With the help of the canonical information of canonical representations and also a canonical regularizer, the proposed method selects canonical-cross-modality features that are useful for the tasks of clinical scores regression and multi-class disease identification. In our experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, we combine Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) images to jointly predict clinical scores of Alzheimer's Disease Assessment Scale-Cognitive subscale (ADAS-Cog) and Mini-Mental State Examination (MMSE) and also identify multi-class disease status for Alzheimer's disease diagnosis.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages162-169
Number of pages8
Volume8674 LNCS
EditionPART 2
ISBN (Print)9783319104690
DOIs
Publication statusPublished - 2014 Jan 1
Externally publishedYes
Event17th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2014 - Boston, MA, United States
Duration: 2014 Sep 142014 Sep 18

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume8674 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other17th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2014
CountryUnited States
CityBoston, MA
Period14/9/1414/9/18

Fingerprint

Multimodality
Alzheimer's Disease
Feature Selection
Feature extraction
Canonical Representation
Multi-class
Neuroimaging
Modality
Positron emission tomography
Multi-task Learning
Medical Image Analysis
Weight Coefficient
Image analysis
Positron Emission Tomography
Canonical Correlation Analysis
Least Squares Regression
Magnetic Resonance Imaging
Higher Dimensions
Sample Size
Objective function

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Zhu, X., Suk, H-I., & Shen, D. (2014). Multi-modality canonical feature selection for Alzheimer's disease diagnosis. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 2 ed., Vol. 8674 LNCS, pp. 162-169). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8674 LNCS, No. PART 2). Springer Verlag. https://doi.org/10.1007/978-3-319-10470-6_21

Multi-modality canonical feature selection for Alzheimer's disease diagnosis. / Zhu, Xiaofeng; Suk, Heung-Il; Shen, Dinggang.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8674 LNCS PART 2. ed. Springer Verlag, 2014. p. 162-169 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8674 LNCS, No. PART 2).

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

Zhu, X, Suk, H-I & Shen, D 2014, Multi-modality canonical feature selection for Alzheimer's disease diagnosis. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 2 edn, vol. 8674 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 2, vol. 8674 LNCS, Springer Verlag, pp. 162-169, 17th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2014, Boston, MA, United States, 14/9/14. https://doi.org/10.1007/978-3-319-10470-6_21
Zhu X, Suk H-I, Shen D. Multi-modality canonical feature selection for Alzheimer's disease diagnosis. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 2 ed. Vol. 8674 LNCS. Springer Verlag. 2014. p. 162-169. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2). https://doi.org/10.1007/978-3-319-10470-6_21
Zhu, Xiaofeng ; Suk, Heung-Il ; Shen, Dinggang. / Multi-modality canonical feature selection for Alzheimer's disease diagnosis. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 8674 LNCS PART 2. ed. Springer Verlag, 2014. pp. 162-169 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2).
@inproceedings{8a146f7c800c4cd9ba518e17ddcdf18e,
title = "Multi-modality canonical feature selection for Alzheimer's disease diagnosis",
abstract = "Feature selection has been commonly regarded as an effective method to lessen the problem of high dimension and low sample size in medical image analysis. In this paper, we propose a novel multi-modality canonical feature selection method. Unlike the conventional sparse Multi-Task Learning (MTL) based feature selection method that mostly considered only the relationship between target response variables, we further consider the correlations between features of different modalities by projecting them into a canonical space determined by canonical correlation analysis. We call the projections as canonical representations. By setting the canonical representations as regressors in a sparse least square regression framework and by further penalizing the objective function with a new canonical regularizer on the weight coefficient matrix, we formulate a multi-modality canonical feature selection method. With the help of the canonical information of canonical representations and also a canonical regularizer, the proposed method selects canonical-cross-modality features that are useful for the tasks of clinical scores regression and multi-class disease identification. In our experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, we combine Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) images to jointly predict clinical scores of Alzheimer's Disease Assessment Scale-Cognitive subscale (ADAS-Cog) and Mini-Mental State Examination (MMSE) and also identify multi-class disease status for Alzheimer's disease diagnosis.",
author = "Xiaofeng Zhu and Heung-Il Suk and Dinggang Shen",
year = "2014",
month = "1",
day = "1",
doi = "10.1007/978-3-319-10470-6_21",
language = "English",
isbn = "9783319104690",
volume = "8674 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
number = "PART 2",
pages = "162--169",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 2",

}

TY - GEN

T1 - Multi-modality canonical feature selection for Alzheimer's disease diagnosis

AU - Zhu, Xiaofeng

AU - Suk, Heung-Il

AU - Shen, Dinggang

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Feature selection has been commonly regarded as an effective method to lessen the problem of high dimension and low sample size in medical image analysis. In this paper, we propose a novel multi-modality canonical feature selection method. Unlike the conventional sparse Multi-Task Learning (MTL) based feature selection method that mostly considered only the relationship between target response variables, we further consider the correlations between features of different modalities by projecting them into a canonical space determined by canonical correlation analysis. We call the projections as canonical representations. By setting the canonical representations as regressors in a sparse least square regression framework and by further penalizing the objective function with a new canonical regularizer on the weight coefficient matrix, we formulate a multi-modality canonical feature selection method. With the help of the canonical information of canonical representations and also a canonical regularizer, the proposed method selects canonical-cross-modality features that are useful for the tasks of clinical scores regression and multi-class disease identification. In our experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, we combine Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) images to jointly predict clinical scores of Alzheimer's Disease Assessment Scale-Cognitive subscale (ADAS-Cog) and Mini-Mental State Examination (MMSE) and also identify multi-class disease status for Alzheimer's disease diagnosis.

AB - Feature selection has been commonly regarded as an effective method to lessen the problem of high dimension and low sample size in medical image analysis. In this paper, we propose a novel multi-modality canonical feature selection method. Unlike the conventional sparse Multi-Task Learning (MTL) based feature selection method that mostly considered only the relationship between target response variables, we further consider the correlations between features of different modalities by projecting them into a canonical space determined by canonical correlation analysis. We call the projections as canonical representations. By setting the canonical representations as regressors in a sparse least square regression framework and by further penalizing the objective function with a new canonical regularizer on the weight coefficient matrix, we formulate a multi-modality canonical feature selection method. With the help of the canonical information of canonical representations and also a canonical regularizer, the proposed method selects canonical-cross-modality features that are useful for the tasks of clinical scores regression and multi-class disease identification. In our experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, we combine Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) images to jointly predict clinical scores of Alzheimer's Disease Assessment Scale-Cognitive subscale (ADAS-Cog) and Mini-Mental State Examination (MMSE) and also identify multi-class disease status for Alzheimer's disease diagnosis.

UR - http://www.scopus.com/inward/record.url?scp=84906985073&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84906985073&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-10470-6_21

DO - 10.1007/978-3-319-10470-6_21

M3 - Conference contribution

AN - SCOPUS:84906985073

SN - 9783319104690

VL - 8674 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 162

EP - 169

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -