Automatic labeling of MR brain images by hierarchical learning of atlas forests

Lichi Zhang, Qian Wang, Yaozong Gao, Guorong Wu, Dinggang Shen

Research output: Contribution to journalArticle

19 Citations (Scopus)

Abstract

Purpose: Automatic brain image labeling is highly demanded in the field of medical image analysis. Multiatlas-based approaches are widely used due to their simplicity and robustness in applications. Also, random forest technique is recognized as an efficient method for labeling, although there are several existing limitations. In this paper, the authors intend to address those limitations by proposing a novel framework based on the hierarchical learning of atlas forests. Methods: Their proposed framework aims to train a hierarchy of forests to better correlate voxels in the MR images with their corresponding labels. There are two specific novel strategies for improving brain image labeling. First, different from the conventional ways of using a single level of random forests for brain labeling, the authors design a hierarchical structure to incorporate multiple levels of forests. In particular, each atlas forest in the bottom level is trained in accordance with each individual atlas, and then the bottom-level forests are clustered based on their capabilities in labeling. For each clustered group, the authors retrain a new representative forest in the higher level by using all atlases associated with the lower-level atlas forests in the current group, as well as the tentative label maps yielded from the lower level. This clustering and retraining procedure is conducted iteratively to yield a hierarchical structure of forests. Second, in the testing stage, the authors also present a novel atlas forest selection method to determine an optimal set of atlas forests from the constructed hierarchical structure (by disabling those nonoptimal forests) for accurately labeling the test image. Results: For validating their proposed framework, the authors evaluate it on the public datasets, including Alzheimer's disease neuroimaging initiative, Internet brain segmentation repository, and LONI LPBA40. The authors compare the results with the conventional approaches. The experiments show that the use of the two novel strategies can significantly improve the labeling performance. Note that when more levels are constructed in the hierarchy, the labeling performance can be further improved, but more computational time will be also required. Conclusions: The authors have proposed a novel multiatlas-based framework for automatic and accurate labeling of brain anatomies, which can achieve accurate labeling results for MR brain images.

Original languageEnglish
Article number023103
JournalMedical Physics
Volume43
Issue number3
DOIs
Publication statusPublished - 2016 Mar 1

Fingerprint

Atlases
Learning
Brain
Forests
Neuroimaging
Internet
Cluster Analysis
Anatomy
Alzheimer Disease

Keywords

  • atlas selection
  • brain MR images
  • clustering
  • image segmentation
  • random forest

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Cite this

Automatic labeling of MR brain images by hierarchical learning of atlas forests. / Zhang, Lichi; Wang, Qian; Gao, Yaozong; Wu, Guorong; Shen, Dinggang.

In: Medical Physics, Vol. 43, No. 3, 023103, 01.03.2016.

Research output: Contribution to journalArticle

Zhang, Lichi ; Wang, Qian ; Gao, Yaozong ; Wu, Guorong ; Shen, Dinggang. / Automatic labeling of MR brain images by hierarchical learning of atlas forests. In: Medical Physics. 2016 ; Vol. 43, No. 3.
@article{ccb609e9e9864b93a3a5285f07307af9,
title = "Automatic labeling of MR brain images by hierarchical learning of atlas forests",
abstract = "Purpose: Automatic brain image labeling is highly demanded in the field of medical image analysis. Multiatlas-based approaches are widely used due to their simplicity and robustness in applications. Also, random forest technique is recognized as an efficient method for labeling, although there are several existing limitations. In this paper, the authors intend to address those limitations by proposing a novel framework based on the hierarchical learning of atlas forests. Methods: Their proposed framework aims to train a hierarchy of forests to better correlate voxels in the MR images with their corresponding labels. There are two specific novel strategies for improving brain image labeling. First, different from the conventional ways of using a single level of random forests for brain labeling, the authors design a hierarchical structure to incorporate multiple levels of forests. In particular, each atlas forest in the bottom level is trained in accordance with each individual atlas, and then the bottom-level forests are clustered based on their capabilities in labeling. For each clustered group, the authors retrain a new representative forest in the higher level by using all atlases associated with the lower-level atlas forests in the current group, as well as the tentative label maps yielded from the lower level. This clustering and retraining procedure is conducted iteratively to yield a hierarchical structure of forests. Second, in the testing stage, the authors also present a novel atlas forest selection method to determine an optimal set of atlas forests from the constructed hierarchical structure (by disabling those nonoptimal forests) for accurately labeling the test image. Results: For validating their proposed framework, the authors evaluate it on the public datasets, including Alzheimer's disease neuroimaging initiative, Internet brain segmentation repository, and LONI LPBA40. The authors compare the results with the conventional approaches. The experiments show that the use of the two novel strategies can significantly improve the labeling performance. Note that when more levels are constructed in the hierarchy, the labeling performance can be further improved, but more computational time will be also required. Conclusions: The authors have proposed a novel multiatlas-based framework for automatic and accurate labeling of brain anatomies, which can achieve accurate labeling results for MR brain images.",
keywords = "atlas selection, brain MR images, clustering, image segmentation, random forest",
author = "Lichi Zhang and Qian Wang and Yaozong Gao and Guorong Wu and Dinggang Shen",
year = "2016",
month = "3",
day = "1",
doi = "10.1118/1.4941011",
language = "English",
volume = "43",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "AAPM - American Association of Physicists in Medicine",
number = "3",

}

TY - JOUR

T1 - Automatic labeling of MR brain images by hierarchical learning of atlas forests

AU - Zhang, Lichi

AU - Wang, Qian

AU - Gao, Yaozong

AU - Wu, Guorong

AU - Shen, Dinggang

PY - 2016/3/1

Y1 - 2016/3/1

N2 - Purpose: Automatic brain image labeling is highly demanded in the field of medical image analysis. Multiatlas-based approaches are widely used due to their simplicity and robustness in applications. Also, random forest technique is recognized as an efficient method for labeling, although there are several existing limitations. In this paper, the authors intend to address those limitations by proposing a novel framework based on the hierarchical learning of atlas forests. Methods: Their proposed framework aims to train a hierarchy of forests to better correlate voxels in the MR images with their corresponding labels. There are two specific novel strategies for improving brain image labeling. First, different from the conventional ways of using a single level of random forests for brain labeling, the authors design a hierarchical structure to incorporate multiple levels of forests. In particular, each atlas forest in the bottom level is trained in accordance with each individual atlas, and then the bottom-level forests are clustered based on their capabilities in labeling. For each clustered group, the authors retrain a new representative forest in the higher level by using all atlases associated with the lower-level atlas forests in the current group, as well as the tentative label maps yielded from the lower level. This clustering and retraining procedure is conducted iteratively to yield a hierarchical structure of forests. Second, in the testing stage, the authors also present a novel atlas forest selection method to determine an optimal set of atlas forests from the constructed hierarchical structure (by disabling those nonoptimal forests) for accurately labeling the test image. Results: For validating their proposed framework, the authors evaluate it on the public datasets, including Alzheimer's disease neuroimaging initiative, Internet brain segmentation repository, and LONI LPBA40. The authors compare the results with the conventional approaches. The experiments show that the use of the two novel strategies can significantly improve the labeling performance. Note that when more levels are constructed in the hierarchy, the labeling performance can be further improved, but more computational time will be also required. Conclusions: The authors have proposed a novel multiatlas-based framework for automatic and accurate labeling of brain anatomies, which can achieve accurate labeling results for MR brain images.

AB - Purpose: Automatic brain image labeling is highly demanded in the field of medical image analysis. Multiatlas-based approaches are widely used due to their simplicity and robustness in applications. Also, random forest technique is recognized as an efficient method for labeling, although there are several existing limitations. In this paper, the authors intend to address those limitations by proposing a novel framework based on the hierarchical learning of atlas forests. Methods: Their proposed framework aims to train a hierarchy of forests to better correlate voxels in the MR images with their corresponding labels. There are two specific novel strategies for improving brain image labeling. First, different from the conventional ways of using a single level of random forests for brain labeling, the authors design a hierarchical structure to incorporate multiple levels of forests. In particular, each atlas forest in the bottom level is trained in accordance with each individual atlas, and then the bottom-level forests are clustered based on their capabilities in labeling. For each clustered group, the authors retrain a new representative forest in the higher level by using all atlases associated with the lower-level atlas forests in the current group, as well as the tentative label maps yielded from the lower level. This clustering and retraining procedure is conducted iteratively to yield a hierarchical structure of forests. Second, in the testing stage, the authors also present a novel atlas forest selection method to determine an optimal set of atlas forests from the constructed hierarchical structure (by disabling those nonoptimal forests) for accurately labeling the test image. Results: For validating their proposed framework, the authors evaluate it on the public datasets, including Alzheimer's disease neuroimaging initiative, Internet brain segmentation repository, and LONI LPBA40. The authors compare the results with the conventional approaches. The experiments show that the use of the two novel strategies can significantly improve the labeling performance. Note that when more levels are constructed in the hierarchy, the labeling performance can be further improved, but more computational time will be also required. Conclusions: The authors have proposed a novel multiatlas-based framework for automatic and accurate labeling of brain anatomies, which can achieve accurate labeling results for MR brain images.

KW - atlas selection

KW - brain MR images

KW - clustering

KW - image segmentation

KW - random forest

UR - http://www.scopus.com/inward/record.url?scp=84958582011&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84958582011&partnerID=8YFLogxK

U2 - 10.1118/1.4941011

DO - 10.1118/1.4941011

M3 - Article

C2 - 26936703

AN - SCOPUS:84958582011

VL - 43

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 3

M1 - 023103

ER -