Novel multi-atlas segmentation by matrix completion

Gerard Sanroma, Guorong Wu, Kim Thung, Yanrong Guo, Dinggang Shen

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

The goal of multi-atlas segmentation is to estimate the anatomical labels on each target image point by combining the labels from a set of registered atlas images via label fusion. Typically, label fusion can be formulated either as a reconstruction or as a classification problem. Reconstruction-based methods compute the target labels as a weighted average of the atlas labels. Such weights are derived from the representation of the target image patches as a linear combination of the atlas image patches. However, the related issue is that the optimal weights in the image domain are not necessarily corresponding to those in the label domain. Classification-based methods can avoid this issue by directly learning the relationship between image and label domains. However, the learned relationships, describing the common characteristics of all the training atlas patches, might not be representative for a particular target image patch, and thus undermine the labeling results. In order to overcome the limitations of both types of methods, we innovatively formulate the patch-based label fusion problem as a matrix completion problem. By doing so, we can jointly utilize (1) the relationships between atlas and target image patches (thus taking the advantage of the reconstruction-based methods), and (2) the relationships between image and label domains (taking the advantage of the classification-based methods). In this way, our generalized paradigm can improve the label fusion accuracy in segmenting the challenging structures, e.g., hippocampus, compared to the state-of-the-art methods.

Original languageEnglish
Pages (from-to)207-214
Number of pages8
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8679
Publication statusPublished - 2014 Jan 1
Externally publishedYes

Fingerprint

Matrix Completion
Atlas
Labels
Segmentation
Patch
Fusion
Target
Fusion reactions
Matrix Completion Problem
Hippocampus
Weighted Average
Classification Problems
Labeling
Linear Combination
Paradigm

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Novel multi-atlas segmentation by matrix completion. / Sanroma, Gerard; Wu, Guorong; Thung, Kim; Guo, Yanrong; Shen, Dinggang.

In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 8679, 01.01.2014, p. 207-214.

Research output: Contribution to journalArticle

@article{d0819dce10a44ac3a914b390c3bb5eb5,
title = "Novel multi-atlas segmentation by matrix completion",
abstract = "The goal of multi-atlas segmentation is to estimate the anatomical labels on each target image point by combining the labels from a set of registered atlas images via label fusion. Typically, label fusion can be formulated either as a reconstruction or as a classification problem. Reconstruction-based methods compute the target labels as a weighted average of the atlas labels. Such weights are derived from the representation of the target image patches as a linear combination of the atlas image patches. However, the related issue is that the optimal weights in the image domain are not necessarily corresponding to those in the label domain. Classification-based methods can avoid this issue by directly learning the relationship between image and label domains. However, the learned relationships, describing the common characteristics of all the training atlas patches, might not be representative for a particular target image patch, and thus undermine the labeling results. In order to overcome the limitations of both types of methods, we innovatively formulate the patch-based label fusion problem as a matrix completion problem. By doing so, we can jointly utilize (1) the relationships between atlas and target image patches (thus taking the advantage of the reconstruction-based methods), and (2) the relationships between image and label domains (taking the advantage of the classification-based methods). In this way, our generalized paradigm can improve the label fusion accuracy in segmenting the challenging structures, e.g., hippocampus, compared to the state-of-the-art methods.",
author = "Gerard Sanroma and Guorong Wu and Kim Thung and Yanrong Guo and Dinggang Shen",
year = "2014",
month = "1",
day = "1",
language = "English",
volume = "8679",
pages = "207--214",
journal = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
issn = "0302-9743",
publisher = "Springer Verlag",

}

TY - JOUR

T1 - Novel multi-atlas segmentation by matrix completion

AU - Sanroma, Gerard

AU - Wu, Guorong

AU - Thung, Kim

AU - Guo, Yanrong

AU - Shen, Dinggang

PY - 2014/1/1

Y1 - 2014/1/1

N2 - The goal of multi-atlas segmentation is to estimate the anatomical labels on each target image point by combining the labels from a set of registered atlas images via label fusion. Typically, label fusion can be formulated either as a reconstruction or as a classification problem. Reconstruction-based methods compute the target labels as a weighted average of the atlas labels. Such weights are derived from the representation of the target image patches as a linear combination of the atlas image patches. However, the related issue is that the optimal weights in the image domain are not necessarily corresponding to those in the label domain. Classification-based methods can avoid this issue by directly learning the relationship between image and label domains. However, the learned relationships, describing the common characteristics of all the training atlas patches, might not be representative for a particular target image patch, and thus undermine the labeling results. In order to overcome the limitations of both types of methods, we innovatively formulate the patch-based label fusion problem as a matrix completion problem. By doing so, we can jointly utilize (1) the relationships between atlas and target image patches (thus taking the advantage of the reconstruction-based methods), and (2) the relationships between image and label domains (taking the advantage of the classification-based methods). In this way, our generalized paradigm can improve the label fusion accuracy in segmenting the challenging structures, e.g., hippocampus, compared to the state-of-the-art methods.

AB - The goal of multi-atlas segmentation is to estimate the anatomical labels on each target image point by combining the labels from a set of registered atlas images via label fusion. Typically, label fusion can be formulated either as a reconstruction or as a classification problem. Reconstruction-based methods compute the target labels as a weighted average of the atlas labels. Such weights are derived from the representation of the target image patches as a linear combination of the atlas image patches. However, the related issue is that the optimal weights in the image domain are not necessarily corresponding to those in the label domain. Classification-based methods can avoid this issue by directly learning the relationship between image and label domains. However, the learned relationships, describing the common characteristics of all the training atlas patches, might not be representative for a particular target image patch, and thus undermine the labeling results. In order to overcome the limitations of both types of methods, we innovatively formulate the patch-based label fusion problem as a matrix completion problem. By doing so, we can jointly utilize (1) the relationships between atlas and target image patches (thus taking the advantage of the reconstruction-based methods), and (2) the relationships between image and label domains (taking the advantage of the classification-based methods). In this way, our generalized paradigm can improve the label fusion accuracy in segmenting the challenging structures, e.g., hippocampus, compared to the state-of-the-art methods.

UR - http://www.scopus.com/inward/record.url?scp=84921819798&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84921819798&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84921819798

VL - 8679

SP - 207

EP - 214

JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SN - 0302-9743

ER -