Toward an interpretable Alzheimer's disease diagnostic model with regional abnormality representation via deep learning

Eunho Lee, Jun Sik Choi, Minjeong Kim, Heung-Il Suk

Research output: Contribution to journalArticle

Abstract

In this paper, we propose a novel method for magnetic resonance imaging based Alzheimer's disease (AD) or mild cognitive impairment (MCI) diagnosis that systematically integrates voxel-based, region-based, and patch-based approaches into a unified framework. Specifically, we parcellate the brain into predefined regions based on anatomical knowledge (i.e., templates) and derive complex nonlinear relationships among voxels, whose intensities denote volumetric measurements, within each region. Unlike existing methods that use cubical or rectangular shapes, we consider the anatomical shapes of regions as atypical patches. Using complex nonlinear relationships among voxels in each region learned by deep neural networks, we extract a “regional abnormality representation.” We then make a final clinical decision by integrating the regional abnormality representations over the entire brain. It is noteworthy that the regional abnormality representations allow us to interpret and understand the symptomatic observations of a subject with AD or MCI by mapping and visualizing these observations in the brain space. On the baseline MRI dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, our method achieves state-of-the-art performance for four binary classification tasks and one three-class classification task. Additionally, we conducted exhaustive experiments and analysis to validate the efficacy and potential of our method.

Original languageEnglish
Article number116113
JournalNeuroImage
Volume202
DOIs
Publication statusPublished - 2019 Nov 15

Fingerprint

Alzheimer Disease
Learning
Brain
Neuroimaging
Magnetic Resonance Imaging
Cognitive Dysfunction

Keywords

  • Abnormality representation
  • Alzheimer's disease
  • Deep neural network
  • Interpretable diagnostic model
  • Magnetic resonance imaging
  • Mild cognitive impairment

ASJC Scopus subject areas

  • Neurology
  • Cognitive Neuroscience

Cite this

Toward an interpretable Alzheimer's disease diagnostic model with regional abnormality representation via deep learning. / Lee, Eunho; Choi, Jun Sik; Kim, Minjeong; Suk, Heung-Il.

In: NeuroImage, Vol. 202, 116113, 15.11.2019.

Research output: Contribution to journalArticle

@article{e558bee6fb5541f685a4b9618a69b233,
title = "Toward an interpretable Alzheimer's disease diagnostic model with regional abnormality representation via deep learning",
abstract = "In this paper, we propose a novel method for magnetic resonance imaging based Alzheimer's disease (AD) or mild cognitive impairment (MCI) diagnosis that systematically integrates voxel-based, region-based, and patch-based approaches into a unified framework. Specifically, we parcellate the brain into predefined regions based on anatomical knowledge (i.e., templates) and derive complex nonlinear relationships among voxels, whose intensities denote volumetric measurements, within each region. Unlike existing methods that use cubical or rectangular shapes, we consider the anatomical shapes of regions as atypical patches. Using complex nonlinear relationships among voxels in each region learned by deep neural networks, we extract a “regional abnormality representation.” We then make a final clinical decision by integrating the regional abnormality representations over the entire brain. It is noteworthy that the regional abnormality representations allow us to interpret and understand the symptomatic observations of a subject with AD or MCI by mapping and visualizing these observations in the brain space. On the baseline MRI dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, our method achieves state-of-the-art performance for four binary classification tasks and one three-class classification task. Additionally, we conducted exhaustive experiments and analysis to validate the efficacy and potential of our method.",
keywords = "Abnormality representation, Alzheimer's disease, Deep neural network, Interpretable diagnostic model, Magnetic resonance imaging, Mild cognitive impairment",
author = "Eunho Lee and Choi, {Jun Sik} and Minjeong Kim and Heung-Il Suk",
year = "2019",
month = "11",
day = "15",
doi = "10.1016/j.neuroimage.2019.116113",
language = "English",
volume = "202",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Toward an interpretable Alzheimer's disease diagnostic model with regional abnormality representation via deep learning

AU - Lee, Eunho

AU - Choi, Jun Sik

AU - Kim, Minjeong

AU - Suk, Heung-Il

PY - 2019/11/15

Y1 - 2019/11/15

N2 - In this paper, we propose a novel method for magnetic resonance imaging based Alzheimer's disease (AD) or mild cognitive impairment (MCI) diagnosis that systematically integrates voxel-based, region-based, and patch-based approaches into a unified framework. Specifically, we parcellate the brain into predefined regions based on anatomical knowledge (i.e., templates) and derive complex nonlinear relationships among voxels, whose intensities denote volumetric measurements, within each region. Unlike existing methods that use cubical or rectangular shapes, we consider the anatomical shapes of regions as atypical patches. Using complex nonlinear relationships among voxels in each region learned by deep neural networks, we extract a “regional abnormality representation.” We then make a final clinical decision by integrating the regional abnormality representations over the entire brain. It is noteworthy that the regional abnormality representations allow us to interpret and understand the symptomatic observations of a subject with AD or MCI by mapping and visualizing these observations in the brain space. On the baseline MRI dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, our method achieves state-of-the-art performance for four binary classification tasks and one three-class classification task. Additionally, we conducted exhaustive experiments and analysis to validate the efficacy and potential of our method.

AB - In this paper, we propose a novel method for magnetic resonance imaging based Alzheimer's disease (AD) or mild cognitive impairment (MCI) diagnosis that systematically integrates voxel-based, region-based, and patch-based approaches into a unified framework. Specifically, we parcellate the brain into predefined regions based on anatomical knowledge (i.e., templates) and derive complex nonlinear relationships among voxels, whose intensities denote volumetric measurements, within each region. Unlike existing methods that use cubical or rectangular shapes, we consider the anatomical shapes of regions as atypical patches. Using complex nonlinear relationships among voxels in each region learned by deep neural networks, we extract a “regional abnormality representation.” We then make a final clinical decision by integrating the regional abnormality representations over the entire brain. It is noteworthy that the regional abnormality representations allow us to interpret and understand the symptomatic observations of a subject with AD or MCI by mapping and visualizing these observations in the brain space. On the baseline MRI dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort, our method achieves state-of-the-art performance for four binary classification tasks and one three-class classification task. Additionally, we conducted exhaustive experiments and analysis to validate the efficacy and potential of our method.

KW - Abnormality representation

KW - Alzheimer's disease

KW - Deep neural network

KW - Interpretable diagnostic model

KW - Magnetic resonance imaging

KW - Mild cognitive impairment

UR - http://www.scopus.com/inward/record.url?scp=85071510128&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071510128&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2019.116113

DO - 10.1016/j.neuroimage.2019.116113

M3 - Article

AN - SCOPUS:85071510128

VL - 202

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

M1 - 116113

ER -