Manifold regularized multitask feature learning for multimodality disease classification

Alzheimer's Disease Neuroimaging Initiative

Research output: Contribution to journalArticle

58 Citations (Scopus)

Abstract

Multimodality based methods have shown great advantages in classification of Alzheimer's disease (AD) and its prodromal stage, that is, mild cognitive impairment (MCI). Recently, multitask feature selection methods are typically used for joint selection of common features across multiple modalities. However, one disadvantage of existing multimodality based methods is that they ignore the useful data distribution information in each modality, which is essential for subsequent classification. Accordingly, in this paper we propose a manifold regularized multitask feature learning method to preserve both the intrinsic relatedness among multiple modalities of data and the data distribution information in each modality. Specifically, we denote the feature learning on each modality as a single task, and use group-sparsity regularizer to capture the intrinsic relatedness among multiple tasks (i.e., modalities) and jointly select the common features from multiple tasks. Furthermore, we introduce a new manifold-based Laplacian regularizer to preserve the data distribution information from each task. Finally, we use the multikernel support vector machine method to fuse multimodality data for eventual classification. Conversely, we also extend our method to the semisupervised setting, where only partial data are labeled. We evaluate our method using the baseline magnetic resonance imaging (MRI), fluorodeoxyglucose positron emission tomography (FDG-PET), and cerebrospinal fluid (CSF) data of subjects from AD neuroimaging initiative database. The experimental results demonstrate that our proposed method can not only achieve improved classification performance, but also help to discover the disease-related brain regions useful for disease diagnosis.

Original languageEnglish
Pages (from-to)489-507
Number of pages19
JournalHuman Brain Mapping
Volume36
Issue number2
DOIs
Publication statusPublished - 2015 Jan 1

Fingerprint

Learning
Information Dissemination
Alzheimer Disease
Prodromal Symptoms
Brain Diseases
Neuroimaging
Positron-Emission Tomography
Cerebrospinal Fluid
Magnetic Resonance Imaging
Databases

Keywords

  • Alzheimer's disease
  • Feature selection
  • Group-sparsity regularizer
  • Manifold regularization
  • Multimodality classification
  • Multitask learning

ASJC Scopus subject areas

  • Clinical Neurology
  • Anatomy
  • Neurology
  • Radiology Nuclear Medicine and imaging
  • Radiological and Ultrasound Technology

Cite this

Manifold regularized multitask feature learning for multimodality disease classification. / Alzheimer's Disease Neuroimaging Initiative.

In: Human Brain Mapping, Vol. 36, No. 2, 01.01.2015, p. 489-507.

Research output: Contribution to journalArticle

Alzheimer's Disease Neuroimaging Initiative. / Manifold regularized multitask feature learning for multimodality disease classification. In: Human Brain Mapping. 2015 ; Vol. 36, No. 2. pp. 489-507.
@article{ba38f9c6c5d8434c8b2747cd07da6789,
title = "Manifold regularized multitask feature learning for multimodality disease classification",
abstract = "Multimodality based methods have shown great advantages in classification of Alzheimer's disease (AD) and its prodromal stage, that is, mild cognitive impairment (MCI). Recently, multitask feature selection methods are typically used for joint selection of common features across multiple modalities. However, one disadvantage of existing multimodality based methods is that they ignore the useful data distribution information in each modality, which is essential for subsequent classification. Accordingly, in this paper we propose a manifold regularized multitask feature learning method to preserve both the intrinsic relatedness among multiple modalities of data and the data distribution information in each modality. Specifically, we denote the feature learning on each modality as a single task, and use group-sparsity regularizer to capture the intrinsic relatedness among multiple tasks (i.e., modalities) and jointly select the common features from multiple tasks. Furthermore, we introduce a new manifold-based Laplacian regularizer to preserve the data distribution information from each task. Finally, we use the multikernel support vector machine method to fuse multimodality data for eventual classification. Conversely, we also extend our method to the semisupervised setting, where only partial data are labeled. We evaluate our method using the baseline magnetic resonance imaging (MRI), fluorodeoxyglucose positron emission tomography (FDG-PET), and cerebrospinal fluid (CSF) data of subjects from AD neuroimaging initiative database. The experimental results demonstrate that our proposed method can not only achieve improved classification performance, but also help to discover the disease-related brain regions useful for disease diagnosis.",
keywords = "Alzheimer's disease, Feature selection, Group-sparsity regularizer, Manifold regularization, Multimodality classification, Multitask learning",
author = "{Alzheimer's Disease Neuroimaging Initiative} and Biao Jie and Daoqiang Zhang and Bo Cheng and Dinggang Shen",
year = "2015",
month = "1",
day = "1",
doi = "10.1002/hbm.22642",
language = "English",
volume = "36",
pages = "489--507",
journal = "Human Brain Mapping",
issn = "1065-9471",
publisher = "Wiley-Liss Inc.",
number = "2",

}

TY - JOUR

T1 - Manifold regularized multitask feature learning for multimodality disease classification

AU - Alzheimer's Disease Neuroimaging Initiative

AU - Jie, Biao

AU - Zhang, Daoqiang

AU - Cheng, Bo

AU - Shen, Dinggang

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Multimodality based methods have shown great advantages in classification of Alzheimer's disease (AD) and its prodromal stage, that is, mild cognitive impairment (MCI). Recently, multitask feature selection methods are typically used for joint selection of common features across multiple modalities. However, one disadvantage of existing multimodality based methods is that they ignore the useful data distribution information in each modality, which is essential for subsequent classification. Accordingly, in this paper we propose a manifold regularized multitask feature learning method to preserve both the intrinsic relatedness among multiple modalities of data and the data distribution information in each modality. Specifically, we denote the feature learning on each modality as a single task, and use group-sparsity regularizer to capture the intrinsic relatedness among multiple tasks (i.e., modalities) and jointly select the common features from multiple tasks. Furthermore, we introduce a new manifold-based Laplacian regularizer to preserve the data distribution information from each task. Finally, we use the multikernel support vector machine method to fuse multimodality data for eventual classification. Conversely, we also extend our method to the semisupervised setting, where only partial data are labeled. We evaluate our method using the baseline magnetic resonance imaging (MRI), fluorodeoxyglucose positron emission tomography (FDG-PET), and cerebrospinal fluid (CSF) data of subjects from AD neuroimaging initiative database. The experimental results demonstrate that our proposed method can not only achieve improved classification performance, but also help to discover the disease-related brain regions useful for disease diagnosis.

AB - Multimodality based methods have shown great advantages in classification of Alzheimer's disease (AD) and its prodromal stage, that is, mild cognitive impairment (MCI). Recently, multitask feature selection methods are typically used for joint selection of common features across multiple modalities. However, one disadvantage of existing multimodality based methods is that they ignore the useful data distribution information in each modality, which is essential for subsequent classification. Accordingly, in this paper we propose a manifold regularized multitask feature learning method to preserve both the intrinsic relatedness among multiple modalities of data and the data distribution information in each modality. Specifically, we denote the feature learning on each modality as a single task, and use group-sparsity regularizer to capture the intrinsic relatedness among multiple tasks (i.e., modalities) and jointly select the common features from multiple tasks. Furthermore, we introduce a new manifold-based Laplacian regularizer to preserve the data distribution information from each task. Finally, we use the multikernel support vector machine method to fuse multimodality data for eventual classification. Conversely, we also extend our method to the semisupervised setting, where only partial data are labeled. We evaluate our method using the baseline magnetic resonance imaging (MRI), fluorodeoxyglucose positron emission tomography (FDG-PET), and cerebrospinal fluid (CSF) data of subjects from AD neuroimaging initiative database. The experimental results demonstrate that our proposed method can not only achieve improved classification performance, but also help to discover the disease-related brain regions useful for disease diagnosis.

KW - Alzheimer's disease

KW - Feature selection

KW - Group-sparsity regularizer

KW - Manifold regularization

KW - Multimodality classification

KW - Multitask learning

UR - http://www.scopus.com/inward/record.url?scp=84922322664&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84922322664&partnerID=8YFLogxK

U2 - 10.1002/hbm.22642

DO - 10.1002/hbm.22642

M3 - Article

VL - 36

SP - 489

EP - 507

JO - Human Brain Mapping

JF - Human Brain Mapping

SN - 1065-9471

IS - 2

ER -