Segmentation of perivascular spaces in 7 T MR image using auto-context model with orientation-normalized features

Sang Hyun Park, Xiaopeng Zong, Yaozong Gao, Weili Lin, Dinggang Shen

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

Quantitative study of perivascular spaces (PVSs) in brain magnetic resonance (MR) images is important for understanding the brain lymphatic system and its relationship with neurological diseases. One of the major challenges is the accurate extraction of PVSs that have very thin tubular structures with various directions in three-dimensional (3D) MR images. In this paper, we propose a learning-based PVS segmentation method to address this challenge. Specifically, we first determine a region of interest (ROI) by using the anatomical brain structure and the vesselness information derived from eigenvalues of image derivatives. Then, in the ROI, we extract a number of randomized Haar features which are normalized with respect to the principal directions of the underlying image derivatives. The classifier is trained by the random forest model that can effectively learn both discriminative features and classifier parameters to maximize the information gain. Finally, a sequential learning strategy is used to further enforce various contextual patterns around the thin tubular structures into the classifier. For evaluation, we apply our proposed method to the 7 T brain MR images scanned from 17 healthy subjects aged from 25 to 37. The performance is measured by voxel-wise segmentation accuracy, cluster-wise classification accuracy, and similarity of geometric properties, such as volume, length, and diameter distributions between the predicted and the true PVSs. Moreover, the accuracies are also evaluated on the simulation images with motion artifacts and lacunes to demonstrate the potential of our method in segmenting PVSs from elderly and patient populations. The experimental results show that our proposed method outperforms all existing PVS segmentation methods.

Original languageEnglish
Pages (from-to)223-235
Number of pages13
JournalNeuroImage
Volume134
DOIs
Publication statusPublished - 2016 Jul 1

Fingerprint

Magnetic Resonance Spectroscopy
Brain
Learning
Lymphatic System
Artifacts
Healthy Volunteers
Population
Direction compound

Keywords

  • 7 T MR image
  • Orientation-normalized Haar feature
  • Perivascular spaces
  • Random forest model
  • Sequential classifiers

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

Segmentation of perivascular spaces in 7 T MR image using auto-context model with orientation-normalized features. / Park, Sang Hyun; Zong, Xiaopeng; Gao, Yaozong; Lin, Weili; Shen, Dinggang.

In: NeuroImage, Vol. 134, 01.07.2016, p. 223-235.

Research output: Contribution to journalArticle

Park, Sang Hyun ; Zong, Xiaopeng ; Gao, Yaozong ; Lin, Weili ; Shen, Dinggang. / Segmentation of perivascular spaces in 7 T MR image using auto-context model with orientation-normalized features. In: NeuroImage. 2016 ; Vol. 134. pp. 223-235.
@article{e986163113544a4b89508f6fdf11356c,
title = "Segmentation of perivascular spaces in 7 T MR image using auto-context model with orientation-normalized features",
abstract = "Quantitative study of perivascular spaces (PVSs) in brain magnetic resonance (MR) images is important for understanding the brain lymphatic system and its relationship with neurological diseases. One of the major challenges is the accurate extraction of PVSs that have very thin tubular structures with various directions in three-dimensional (3D) MR images. In this paper, we propose a learning-based PVS segmentation method to address this challenge. Specifically, we first determine a region of interest (ROI) by using the anatomical brain structure and the vesselness information derived from eigenvalues of image derivatives. Then, in the ROI, we extract a number of randomized Haar features which are normalized with respect to the principal directions of the underlying image derivatives. The classifier is trained by the random forest model that can effectively learn both discriminative features and classifier parameters to maximize the information gain. Finally, a sequential learning strategy is used to further enforce various contextual patterns around the thin tubular structures into the classifier. For evaluation, we apply our proposed method to the 7 T brain MR images scanned from 17 healthy subjects aged from 25 to 37. The performance is measured by voxel-wise segmentation accuracy, cluster-wise classification accuracy, and similarity of geometric properties, such as volume, length, and diameter distributions between the predicted and the true PVSs. Moreover, the accuracies are also evaluated on the simulation images with motion artifacts and lacunes to demonstrate the potential of our method in segmenting PVSs from elderly and patient populations. The experimental results show that our proposed method outperforms all existing PVS segmentation methods.",
keywords = "7 T MR image, Orientation-normalized Haar feature, Perivascular spaces, Random forest model, Sequential classifiers",
author = "Park, {Sang Hyun} and Xiaopeng Zong and Yaozong Gao and Weili Lin and Dinggang Shen",
year = "2016",
month = "7",
day = "1",
doi = "10.1016/j.neuroimage.2016.03.076",
language = "English",
volume = "134",
pages = "223--235",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Segmentation of perivascular spaces in 7 T MR image using auto-context model with orientation-normalized features

AU - Park, Sang Hyun

AU - Zong, Xiaopeng

AU - Gao, Yaozong

AU - Lin, Weili

AU - Shen, Dinggang

PY - 2016/7/1

Y1 - 2016/7/1

N2 - Quantitative study of perivascular spaces (PVSs) in brain magnetic resonance (MR) images is important for understanding the brain lymphatic system and its relationship with neurological diseases. One of the major challenges is the accurate extraction of PVSs that have very thin tubular structures with various directions in three-dimensional (3D) MR images. In this paper, we propose a learning-based PVS segmentation method to address this challenge. Specifically, we first determine a region of interest (ROI) by using the anatomical brain structure and the vesselness information derived from eigenvalues of image derivatives. Then, in the ROI, we extract a number of randomized Haar features which are normalized with respect to the principal directions of the underlying image derivatives. The classifier is trained by the random forest model that can effectively learn both discriminative features and classifier parameters to maximize the information gain. Finally, a sequential learning strategy is used to further enforce various contextual patterns around the thin tubular structures into the classifier. For evaluation, we apply our proposed method to the 7 T brain MR images scanned from 17 healthy subjects aged from 25 to 37. The performance is measured by voxel-wise segmentation accuracy, cluster-wise classification accuracy, and similarity of geometric properties, such as volume, length, and diameter distributions between the predicted and the true PVSs. Moreover, the accuracies are also evaluated on the simulation images with motion artifacts and lacunes to demonstrate the potential of our method in segmenting PVSs from elderly and patient populations. The experimental results show that our proposed method outperforms all existing PVS segmentation methods.

AB - Quantitative study of perivascular spaces (PVSs) in brain magnetic resonance (MR) images is important for understanding the brain lymphatic system and its relationship with neurological diseases. One of the major challenges is the accurate extraction of PVSs that have very thin tubular structures with various directions in three-dimensional (3D) MR images. In this paper, we propose a learning-based PVS segmentation method to address this challenge. Specifically, we first determine a region of interest (ROI) by using the anatomical brain structure and the vesselness information derived from eigenvalues of image derivatives. Then, in the ROI, we extract a number of randomized Haar features which are normalized with respect to the principal directions of the underlying image derivatives. The classifier is trained by the random forest model that can effectively learn both discriminative features and classifier parameters to maximize the information gain. Finally, a sequential learning strategy is used to further enforce various contextual patterns around the thin tubular structures into the classifier. For evaluation, we apply our proposed method to the 7 T brain MR images scanned from 17 healthy subjects aged from 25 to 37. The performance is measured by voxel-wise segmentation accuracy, cluster-wise classification accuracy, and similarity of geometric properties, such as volume, length, and diameter distributions between the predicted and the true PVSs. Moreover, the accuracies are also evaluated on the simulation images with motion artifacts and lacunes to demonstrate the potential of our method in segmenting PVSs from elderly and patient populations. The experimental results show that our proposed method outperforms all existing PVS segmentation methods.

KW - 7 T MR image

KW - Orientation-normalized Haar feature

KW - Perivascular spaces

KW - Random forest model

KW - Sequential classifiers

UR - http://www.scopus.com/inward/record.url?scp=84963800692&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84963800692&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2016.03.076

DO - 10.1016/j.neuroimage.2016.03.076

M3 - Article

C2 - 27046107

AN - SCOPUS:84963800692

VL - 134

SP - 223

EP - 235

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

ER -