Structured sparsity regularized multiple kernel learning for Alzheimer's disease diagnosis

Jialin Peng, Xiaofeng Zhu, Ye Wang, Le An, Dinggang Shen

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Multimodal data fusion has shown great advantages in uncovering information that could be overlooked by using single modality. In this paper, we consider the integration of high-dimensional multi-modality imaging and genetic data for Alzheimer's disease (AD) diagnosis. With a focus on taking advantage of both phenotype and genotype information, a novel structured sparsity, defined by ℓ1, p-norm (p > 1), regularized multiple kernel learning method is designed. Specifically, to facilitate structured feature selection and fusion from heterogeneous modalities and also capture feature-wise importance, we represent each feature with a distinct kernel as a basis, followed by grouping the kernels according to modalities. Then, an optimally combined kernel presentation of multimodal features is learned in a data-driven approach. Contrary to the Group Lasso (i.e., ℓ2, 1-norm penalty) which performs sparse group selection, the proposed regularizer enforced on kernel weights is to sparselyselect concise feature set within each homogenous group and fuse the heterogeneous feature groups by taking advantage of dense norms. We have evaluated our method using data of subjects from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The effectiveness of the method is demonstrated by the clearly improved prediction diagnosis and also the discovered brain regions and SNPs relevant to AD.

Original languageEnglish
Pages (from-to)370-382
Number of pages13
JournalPattern Recognition
Volume88
DOIs
Publication statusPublished - 2019 Apr 1

Fingerprint

Neuroimaging
Data fusion
Electric fuses
Feature extraction
Brain
Imaging techniques

Keywords

  • Alzheimer's disease diagnosis
  • Feature selection
  • Multimodal features
  • Multiple kernel learning
  • Structured sparsity

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Computer Vision and Pattern Recognition
  • Artificial Intelligence

Cite this

Structured sparsity regularized multiple kernel learning for Alzheimer's disease diagnosis. / Peng, Jialin; Zhu, Xiaofeng; Wang, Ye; An, Le; Shen, Dinggang.

In: Pattern Recognition, Vol. 88, 01.04.2019, p. 370-382.

Research output: Contribution to journalArticle

Peng, Jialin ; Zhu, Xiaofeng ; Wang, Ye ; An, Le ; Shen, Dinggang. / Structured sparsity regularized multiple kernel learning for Alzheimer's disease diagnosis. In: Pattern Recognition. 2019 ; Vol. 88. pp. 370-382.
@article{ff26f998dc984f7f80f599f2ce11a1ca,
title = "Structured sparsity regularized multiple kernel learning for Alzheimer's disease diagnosis",
abstract = "Multimodal data fusion has shown great advantages in uncovering information that could be overlooked by using single modality. In this paper, we consider the integration of high-dimensional multi-modality imaging and genetic data for Alzheimer's disease (AD) diagnosis. With a focus on taking advantage of both phenotype and genotype information, a novel structured sparsity, defined by ℓ1, p-norm (p > 1), regularized multiple kernel learning method is designed. Specifically, to facilitate structured feature selection and fusion from heterogeneous modalities and also capture feature-wise importance, we represent each feature with a distinct kernel as a basis, followed by grouping the kernels according to modalities. Then, an optimally combined kernel presentation of multimodal features is learned in a data-driven approach. Contrary to the Group Lasso (i.e., ℓ2, 1-norm penalty) which performs sparse group selection, the proposed regularizer enforced on kernel weights is to sparselyselect concise feature set within each homogenous group and fuse the heterogeneous feature groups by taking advantage of dense norms. We have evaluated our method using data of subjects from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The effectiveness of the method is demonstrated by the clearly improved prediction diagnosis and also the discovered brain regions and SNPs relevant to AD.",
keywords = "Alzheimer's disease diagnosis, Feature selection, Multimodal features, Multiple kernel learning, Structured sparsity",
author = "Jialin Peng and Xiaofeng Zhu and Ye Wang and Le An and Dinggang Shen",
year = "2019",
month = "4",
day = "1",
doi = "10.1016/j.patcog.2018.11.027",
language = "English",
volume = "88",
pages = "370--382",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Structured sparsity regularized multiple kernel learning for Alzheimer's disease diagnosis

AU - Peng, Jialin

AU - Zhu, Xiaofeng

AU - Wang, Ye

AU - An, Le

AU - Shen, Dinggang

PY - 2019/4/1

Y1 - 2019/4/1

N2 - Multimodal data fusion has shown great advantages in uncovering information that could be overlooked by using single modality. In this paper, we consider the integration of high-dimensional multi-modality imaging and genetic data for Alzheimer's disease (AD) diagnosis. With a focus on taking advantage of both phenotype and genotype information, a novel structured sparsity, defined by ℓ1, p-norm (p > 1), regularized multiple kernel learning method is designed. Specifically, to facilitate structured feature selection and fusion from heterogeneous modalities and also capture feature-wise importance, we represent each feature with a distinct kernel as a basis, followed by grouping the kernels according to modalities. Then, an optimally combined kernel presentation of multimodal features is learned in a data-driven approach. Contrary to the Group Lasso (i.e., ℓ2, 1-norm penalty) which performs sparse group selection, the proposed regularizer enforced on kernel weights is to sparselyselect concise feature set within each homogenous group and fuse the heterogeneous feature groups by taking advantage of dense norms. We have evaluated our method using data of subjects from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The effectiveness of the method is demonstrated by the clearly improved prediction diagnosis and also the discovered brain regions and SNPs relevant to AD.

AB - Multimodal data fusion has shown great advantages in uncovering information that could be overlooked by using single modality. In this paper, we consider the integration of high-dimensional multi-modality imaging and genetic data for Alzheimer's disease (AD) diagnosis. With a focus on taking advantage of both phenotype and genotype information, a novel structured sparsity, defined by ℓ1, p-norm (p > 1), regularized multiple kernel learning method is designed. Specifically, to facilitate structured feature selection and fusion from heterogeneous modalities and also capture feature-wise importance, we represent each feature with a distinct kernel as a basis, followed by grouping the kernels according to modalities. Then, an optimally combined kernel presentation of multimodal features is learned in a data-driven approach. Contrary to the Group Lasso (i.e., ℓ2, 1-norm penalty) which performs sparse group selection, the proposed regularizer enforced on kernel weights is to sparselyselect concise feature set within each homogenous group and fuse the heterogeneous feature groups by taking advantage of dense norms. We have evaluated our method using data of subjects from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The effectiveness of the method is demonstrated by the clearly improved prediction diagnosis and also the discovered brain regions and SNPs relevant to AD.

KW - Alzheimer's disease diagnosis

KW - Feature selection

KW - Multimodal features

KW - Multiple kernel learning

KW - Structured sparsity

UR - http://www.scopus.com/inward/record.url?scp=85057534760&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85057534760&partnerID=8YFLogxK

U2 - 10.1016/j.patcog.2018.11.027

DO - 10.1016/j.patcog.2018.11.027

M3 - Article

AN - SCOPUS:85057534760

VL - 88

SP - 370

EP - 382

JO - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

ER -