3D structure tensor analysis of light microscopy data for validating diffusion MRI

Ahmad Raza Khan, Anda Cornea, Lindsey A. Leigland, Steven Kohama, Sune Nørhøj Jespersen, Christopher (Chris) Kroenke

Research output: Contribution to journalArticle

28 Citations (Scopus)

Abstract

Diffusion magnetic resonance imaging (d-MRI) is a powerful non-invasive and non-destructive technique for characterizing brain tissue on the microscopic scale. However, the lack of validation of d-MRI by independent experimental means poses an obstacle to accurate interpretation of data acquired using this method. Recently, structure tensor analysis has been applied to light microscopy images, and this technique holds promise to be a powerful validation strategy for d-MRI. Advantages of this approach include its similarity to d-MRI in terms of averaging the effects of a large number of cellular structures, and its simplicity, which enables it to be implemented in a high-throughput manner. However, a drawback of previous implementations of this technique arises from it being restricted to 2D. As a result, structure tensor analyses have been limited to tissue sectioned in a direction orthogonal to the direction of interest. Here we describe the analytical framework for extending structure tensor analysis to 3D, and utilize the results to analyze serial image "stacks" acquired with confocal microscopy of rhesus macaque hippocampal tissue. Implementation of 3D structure tensor procedures requires removal of sources of anisotropy introduced in tissue preparation and confocal imaging. This is accomplished with image processing steps to mitigate the effects of anisotropic tissue shrinkage, and the effects of anisotropy in the point spread function (PSF). In order to address the latter confound, we describe procedures for measuring the dependence of PSF anisotropy on distance from the microscope objective within tissue. Prior to microscopy, ex vivo d-MRI measurements performed on the hippocampal tissue revealed three regions of tissue with mutually orthogonal directions of least restricted diffusion that correspond to CA1, alveus and inferior longitudinal fasciculus. We demonstrate the ability of 3D structure tensor analysis to identify structure tensor orientations that are parallel to d-MRI derived diffusion tensors in each of these three regions. It is concluded that the 3D generalization of structure tensor analysis will further improve the utility of this method for validation of d-MRI by making it a more flexible experimental technique that closer resembles the inherently 3D nature of d-MRI measurements.

Original languageEnglish (US)
Pages (from-to)192-203
Number of pages12
JournalNeuroImage
Volume111
DOIs
StatePublished - May 1 2015

Fingerprint

Diffusion Magnetic Resonance Imaging
Microscopy
Light
Anisotropy
Cellular Structures
Macaca mulatta
Confocal Microscopy
Brain

Keywords

  • Anisotropy
  • Diffusion
  • Hippocampus
  • Structure tensor

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

3D structure tensor analysis of light microscopy data for validating diffusion MRI. / Khan, Ahmad Raza; Cornea, Anda; Leigland, Lindsey A.; Kohama, Steven; Jespersen, Sune Nørhøj; Kroenke, Christopher (Chris).

In: NeuroImage, Vol. 111, 01.05.2015, p. 192-203.

Research output: Contribution to journalArticle

Khan, Ahmad Raza ; Cornea, Anda ; Leigland, Lindsey A. ; Kohama, Steven ; Jespersen, Sune Nørhøj ; Kroenke, Christopher (Chris). / 3D structure tensor analysis of light microscopy data for validating diffusion MRI. In: NeuroImage. 2015 ; Vol. 111. pp. 192-203.
@article{198f8e4872784d77aad402249ee117d1,
title = "3D structure tensor analysis of light microscopy data for validating diffusion MRI",
abstract = "Diffusion magnetic resonance imaging (d-MRI) is a powerful non-invasive and non-destructive technique for characterizing brain tissue on the microscopic scale. However, the lack of validation of d-MRI by independent experimental means poses an obstacle to accurate interpretation of data acquired using this method. Recently, structure tensor analysis has been applied to light microscopy images, and this technique holds promise to be a powerful validation strategy for d-MRI. Advantages of this approach include its similarity to d-MRI in terms of averaging the effects of a large number of cellular structures, and its simplicity, which enables it to be implemented in a high-throughput manner. However, a drawback of previous implementations of this technique arises from it being restricted to 2D. As a result, structure tensor analyses have been limited to tissue sectioned in a direction orthogonal to the direction of interest. Here we describe the analytical framework for extending structure tensor analysis to 3D, and utilize the results to analyze serial image {"}stacks{"} acquired with confocal microscopy of rhesus macaque hippocampal tissue. Implementation of 3D structure tensor procedures requires removal of sources of anisotropy introduced in tissue preparation and confocal imaging. This is accomplished with image processing steps to mitigate the effects of anisotropic tissue shrinkage, and the effects of anisotropy in the point spread function (PSF). In order to address the latter confound, we describe procedures for measuring the dependence of PSF anisotropy on distance from the microscope objective within tissue. Prior to microscopy, ex vivo d-MRI measurements performed on the hippocampal tissue revealed three regions of tissue with mutually orthogonal directions of least restricted diffusion that correspond to CA1, alveus and inferior longitudinal fasciculus. We demonstrate the ability of 3D structure tensor analysis to identify structure tensor orientations that are parallel to d-MRI derived diffusion tensors in each of these three regions. It is concluded that the 3D generalization of structure tensor analysis will further improve the utility of this method for validation of d-MRI by making it a more flexible experimental technique that closer resembles the inherently 3D nature of d-MRI measurements.",
keywords = "Anisotropy, Diffusion, Hippocampus, Structure tensor",
author = "Khan, {Ahmad Raza} and Anda Cornea and Leigland, {Lindsey A.} and Steven Kohama and Jespersen, {Sune N{\o}rh{\o}j} and Kroenke, {Christopher (Chris)}",
year = "2015",
month = "5",
day = "1",
doi = "10.1016/j.neuroimage.2015.01.061",
language = "English (US)",
volume = "111",
pages = "192--203",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - 3D structure tensor analysis of light microscopy data for validating diffusion MRI

AU - Khan, Ahmad Raza

AU - Cornea, Anda

AU - Leigland, Lindsey A.

AU - Kohama, Steven

AU - Jespersen, Sune Nørhøj

AU - Kroenke, Christopher (Chris)

PY - 2015/5/1

Y1 - 2015/5/1

N2 - Diffusion magnetic resonance imaging (d-MRI) is a powerful non-invasive and non-destructive technique for characterizing brain tissue on the microscopic scale. However, the lack of validation of d-MRI by independent experimental means poses an obstacle to accurate interpretation of data acquired using this method. Recently, structure tensor analysis has been applied to light microscopy images, and this technique holds promise to be a powerful validation strategy for d-MRI. Advantages of this approach include its similarity to d-MRI in terms of averaging the effects of a large number of cellular structures, and its simplicity, which enables it to be implemented in a high-throughput manner. However, a drawback of previous implementations of this technique arises from it being restricted to 2D. As a result, structure tensor analyses have been limited to tissue sectioned in a direction orthogonal to the direction of interest. Here we describe the analytical framework for extending structure tensor analysis to 3D, and utilize the results to analyze serial image "stacks" acquired with confocal microscopy of rhesus macaque hippocampal tissue. Implementation of 3D structure tensor procedures requires removal of sources of anisotropy introduced in tissue preparation and confocal imaging. This is accomplished with image processing steps to mitigate the effects of anisotropic tissue shrinkage, and the effects of anisotropy in the point spread function (PSF). In order to address the latter confound, we describe procedures for measuring the dependence of PSF anisotropy on distance from the microscope objective within tissue. Prior to microscopy, ex vivo d-MRI measurements performed on the hippocampal tissue revealed three regions of tissue with mutually orthogonal directions of least restricted diffusion that correspond to CA1, alveus and inferior longitudinal fasciculus. We demonstrate the ability of 3D structure tensor analysis to identify structure tensor orientations that are parallel to d-MRI derived diffusion tensors in each of these three regions. It is concluded that the 3D generalization of structure tensor analysis will further improve the utility of this method for validation of d-MRI by making it a more flexible experimental technique that closer resembles the inherently 3D nature of d-MRI measurements.

AB - Diffusion magnetic resonance imaging (d-MRI) is a powerful non-invasive and non-destructive technique for characterizing brain tissue on the microscopic scale. However, the lack of validation of d-MRI by independent experimental means poses an obstacle to accurate interpretation of data acquired using this method. Recently, structure tensor analysis has been applied to light microscopy images, and this technique holds promise to be a powerful validation strategy for d-MRI. Advantages of this approach include its similarity to d-MRI in terms of averaging the effects of a large number of cellular structures, and its simplicity, which enables it to be implemented in a high-throughput manner. However, a drawback of previous implementations of this technique arises from it being restricted to 2D. As a result, structure tensor analyses have been limited to tissue sectioned in a direction orthogonal to the direction of interest. Here we describe the analytical framework for extending structure tensor analysis to 3D, and utilize the results to analyze serial image "stacks" acquired with confocal microscopy of rhesus macaque hippocampal tissue. Implementation of 3D structure tensor procedures requires removal of sources of anisotropy introduced in tissue preparation and confocal imaging. This is accomplished with image processing steps to mitigate the effects of anisotropic tissue shrinkage, and the effects of anisotropy in the point spread function (PSF). In order to address the latter confound, we describe procedures for measuring the dependence of PSF anisotropy on distance from the microscope objective within tissue. Prior to microscopy, ex vivo d-MRI measurements performed on the hippocampal tissue revealed three regions of tissue with mutually orthogonal directions of least restricted diffusion that correspond to CA1, alveus and inferior longitudinal fasciculus. We demonstrate the ability of 3D structure tensor analysis to identify structure tensor orientations that are parallel to d-MRI derived diffusion tensors in each of these three regions. It is concluded that the 3D generalization of structure tensor analysis will further improve the utility of this method for validation of d-MRI by making it a more flexible experimental technique that closer resembles the inherently 3D nature of d-MRI measurements.

KW - Anisotropy

KW - Diffusion

KW - Hippocampus

KW - Structure tensor

UR - http://www.scopus.com/inward/record.url?scp=84923596100&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84923596100&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2015.01.061

DO - 10.1016/j.neuroimage.2015.01.061

M3 - Article

VL - 111

SP - 192

EP - 203

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

ER -