Measuring temporal morphological changes robustly in brain MR images via 4-dimensional template warping

Dinggang Shen, Christos Davatzikos

Research output: Contribution to journalArticle

88 Citations (Scopus)

Abstract

Robustly measuring subtle longitudinal brain changes is a major challenge in computational neuroanatomy. This paper describes a method for measuring temporal morphological brain changes, by means of a 4-dimensional image warping mechanism. Longitudinal stability is achieved by considering all temporal MR images of an individual simultaneously, rather than by individually warping a template to an individual, or by warping the images of one time-point to those of another time-point. Following earlier work in 3D, a local morphological signature is attached to each voxel of a sequence of images, and it includes a set of image attributes reflecting morphological characteristics of the spatiotemporal structure around the respective voxel at different scales. This attribute vector forms the basis for searching in the 4-dimensional space for a counterpart that has similar morphological signature, thereby leading to automated detection of anatomical correspondence. Ambiguities in this process are reduced by constructing attribute vectors that are highly distinctive of respective voxels, and by using a hierarchical matching procedure in which reliable and easily distinguishable voxels are used to guide the 4D deformation process. The resultant deformations are smooth both in the spatial and temporal dimensions, and are shown to significantly improve warping accuracy over a series of independent 3D warpings, in longitudinal measurements.

Original languageEnglish
Pages (from-to)1508-1517
Number of pages10
JournalNeuroImage
Volume21
Issue number4
DOIs
Publication statusPublished - 2004 Apr 1
Externally publishedYes

Fingerprint

Brain
Neuroanatomy

Keywords

  • 4D registration
  • Atlas warping
  • Attribute vectors
  • Brain atrophy
  • Brain morphology
  • Deformable registration
  • Hierarchical deformation
  • Longitudinal change
  • Temporal change

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

Measuring temporal morphological changes robustly in brain MR images via 4-dimensional template warping. / Shen, Dinggang; Davatzikos, Christos.

In: NeuroImage, Vol. 21, No. 4, 01.04.2004, p. 1508-1517.

Research output: Contribution to journalArticle

@article{82a643985d63499994f12c4ba2a4c4ef,
title = "Measuring temporal morphological changes robustly in brain MR images via 4-dimensional template warping",
abstract = "Robustly measuring subtle longitudinal brain changes is a major challenge in computational neuroanatomy. This paper describes a method for measuring temporal morphological brain changes, by means of a 4-dimensional image warping mechanism. Longitudinal stability is achieved by considering all temporal MR images of an individual simultaneously, rather than by individually warping a template to an individual, or by warping the images of one time-point to those of another time-point. Following earlier work in 3D, a local morphological signature is attached to each voxel of a sequence of images, and it includes a set of image attributes reflecting morphological characteristics of the spatiotemporal structure around the respective voxel at different scales. This attribute vector forms the basis for searching in the 4-dimensional space for a counterpart that has similar morphological signature, thereby leading to automated detection of anatomical correspondence. Ambiguities in this process are reduced by constructing attribute vectors that are highly distinctive of respective voxels, and by using a hierarchical matching procedure in which reliable and easily distinguishable voxels are used to guide the 4D deformation process. The resultant deformations are smooth both in the spatial and temporal dimensions, and are shown to significantly improve warping accuracy over a series of independent 3D warpings, in longitudinal measurements.",
keywords = "4D registration, Atlas warping, Attribute vectors, Brain atrophy, Brain morphology, Deformable registration, Hierarchical deformation, Longitudinal change, Temporal change",
author = "Dinggang Shen and Christos Davatzikos",
year = "2004",
month = "4",
day = "1",
doi = "10.1016/j.neuroimage.2003.12.015",
language = "English",
volume = "21",
pages = "1508--1517",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",
number = "4",

}

TY - JOUR

T1 - Measuring temporal morphological changes robustly in brain MR images via 4-dimensional template warping

AU - Shen, Dinggang

AU - Davatzikos, Christos

PY - 2004/4/1

Y1 - 2004/4/1

N2 - Robustly measuring subtle longitudinal brain changes is a major challenge in computational neuroanatomy. This paper describes a method for measuring temporal morphological brain changes, by means of a 4-dimensional image warping mechanism. Longitudinal stability is achieved by considering all temporal MR images of an individual simultaneously, rather than by individually warping a template to an individual, or by warping the images of one time-point to those of another time-point. Following earlier work in 3D, a local morphological signature is attached to each voxel of a sequence of images, and it includes a set of image attributes reflecting morphological characteristics of the spatiotemporal structure around the respective voxel at different scales. This attribute vector forms the basis for searching in the 4-dimensional space for a counterpart that has similar morphological signature, thereby leading to automated detection of anatomical correspondence. Ambiguities in this process are reduced by constructing attribute vectors that are highly distinctive of respective voxels, and by using a hierarchical matching procedure in which reliable and easily distinguishable voxels are used to guide the 4D deformation process. The resultant deformations are smooth both in the spatial and temporal dimensions, and are shown to significantly improve warping accuracy over a series of independent 3D warpings, in longitudinal measurements.

AB - Robustly measuring subtle longitudinal brain changes is a major challenge in computational neuroanatomy. This paper describes a method for measuring temporal morphological brain changes, by means of a 4-dimensional image warping mechanism. Longitudinal stability is achieved by considering all temporal MR images of an individual simultaneously, rather than by individually warping a template to an individual, or by warping the images of one time-point to those of another time-point. Following earlier work in 3D, a local morphological signature is attached to each voxel of a sequence of images, and it includes a set of image attributes reflecting morphological characteristics of the spatiotemporal structure around the respective voxel at different scales. This attribute vector forms the basis for searching in the 4-dimensional space for a counterpart that has similar morphological signature, thereby leading to automated detection of anatomical correspondence. Ambiguities in this process are reduced by constructing attribute vectors that are highly distinctive of respective voxels, and by using a hierarchical matching procedure in which reliable and easily distinguishable voxels are used to guide the 4D deformation process. The resultant deformations are smooth both in the spatial and temporal dimensions, and are shown to significantly improve warping accuracy over a series of independent 3D warpings, in longitudinal measurements.

KW - 4D registration

KW - Atlas warping

KW - Attribute vectors

KW - Brain atrophy

KW - Brain morphology

KW - Deformable registration

KW - Hierarchical deformation

KW - Longitudinal change

KW - Temporal change

UR - http://www.scopus.com/inward/record.url?scp=1842452878&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=1842452878&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2003.12.015

DO - 10.1016/j.neuroimage.2003.12.015

M3 - Article

C2 - 15050575

AN - SCOPUS:1842452878

VL - 21

SP - 1508

EP - 1517

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

IS - 4

ER -