Validation of a fully automated 3D hippocampal segmentation method using subjects with Alzheimer's disease mild cognitive impairment, and elderly controls

Jonathan H. Morra, Zhuowen Tu, Liana G. Apostolova, Amity E. Green, Christina Avedissian, Sarah K. Madsen, Neelroop Parikshak, Xue Hua, Arthur W. Toga, Clifford R Jr. Jack, Michael W. Weiner, Paul M. Thompson

Research output: Contribution to journalArticle

149 Citations (Scopus)

Abstract

We introduce a new method for brain MRI segmentation, called the auto context model (ACM), to segment the hippocampus automatically in 3D T1-weighted structural brain MRI scans of subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). In a training phase, our algorithm used 21 hand-labeled segmentations to learn a classification rule for hippocampal versus non-hippocampal regions using a modified AdaBoost method, based on ∼ 18,000 features (image intensity, position, image curvatures, image gradients, tissue classification maps of gray/white matter and CSF, and mean, standard deviation, and Haar filters of size 1 × 1 × 1 to 7 × 7 × 7). We linearly registered all brains to a standard template to devise a basic shape prior to capture the global shape of the hippocampus, defined as the pointwise summation of all the training masks. We also included curvature, gradient, mean, standard deviation, and Haar filters of the shape prior and the tissue classified images as features. During each iteration of ACM - our extension of AdaBoost - the Bayesian posterior distribution of the labeling was fed back in as an input, along with its neighborhood features as new features for AdaBoost to use. In validation studies, we compared our results with hand-labeled segmentations by two experts. Using a leave-one-out approach and standard overlap and distance error metrics, our automated segmentations agreed well with human raters; any differences were comparable to differences between trained human raters. Our error metrics compare favorably with those previously reported for other automated hippocampal segmentations, suggesting the utility of the approach for large-scale studies.

Original languageEnglish (US)
Pages (from-to)59-68
Number of pages10
JournalNeuroImage
Volume43
Issue number1
DOIs
StatePublished - Oct 15 2008

Fingerprint

Alzheimer Disease
Hippocampus
Brain
Hand
Validation Studies
Masks
Neuroimaging
Magnetic Resonance Imaging
Cognitive Dysfunction
Gray Matter
White Matter

ASJC Scopus subject areas

  • Cognitive Neuroscience
  • Neurology

Cite this

Validation of a fully automated 3D hippocampal segmentation method using subjects with Alzheimer's disease mild cognitive impairment, and elderly controls. / Morra, Jonathan H.; Tu, Zhuowen; Apostolova, Liana G.; Green, Amity E.; Avedissian, Christina; Madsen, Sarah K.; Parikshak, Neelroop; Hua, Xue; Toga, Arthur W.; Jack, Clifford R Jr.; Weiner, Michael W.; Thompson, Paul M.

In: NeuroImage, Vol. 43, No. 1, 15.10.2008, p. 59-68.

Research output: Contribution to journalArticle

Morra, JH, Tu, Z, Apostolova, LG, Green, AE, Avedissian, C, Madsen, SK, Parikshak, N, Hua, X, Toga, AW, Jack, CRJ, Weiner, MW & Thompson, PM 2008, 'Validation of a fully automated 3D hippocampal segmentation method using subjects with Alzheimer's disease mild cognitive impairment, and elderly controls', NeuroImage, vol. 43, no. 1, pp. 59-68. https://doi.org/10.1016/j.neuroimage.2008.07.003
Morra, Jonathan H. ; Tu, Zhuowen ; Apostolova, Liana G. ; Green, Amity E. ; Avedissian, Christina ; Madsen, Sarah K. ; Parikshak, Neelroop ; Hua, Xue ; Toga, Arthur W. ; Jack, Clifford R Jr. ; Weiner, Michael W. ; Thompson, Paul M. / Validation of a fully automated 3D hippocampal segmentation method using subjects with Alzheimer's disease mild cognitive impairment, and elderly controls. In: NeuroImage. 2008 ; Vol. 43, No. 1. pp. 59-68.
@article{e35336e7a1d54d4cbebbef3bac3adeaa,
title = "Validation of a fully automated 3D hippocampal segmentation method using subjects with Alzheimer's disease mild cognitive impairment, and elderly controls",
abstract = "We introduce a new method for brain MRI segmentation, called the auto context model (ACM), to segment the hippocampus automatically in 3D T1-weighted structural brain MRI scans of subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). In a training phase, our algorithm used 21 hand-labeled segmentations to learn a classification rule for hippocampal versus non-hippocampal regions using a modified AdaBoost method, based on ∼ 18,000 features (image intensity, position, image curvatures, image gradients, tissue classification maps of gray/white matter and CSF, and mean, standard deviation, and Haar filters of size 1 × 1 × 1 to 7 × 7 × 7). We linearly registered all brains to a standard template to devise a basic shape prior to capture the global shape of the hippocampus, defined as the pointwise summation of all the training masks. We also included curvature, gradient, mean, standard deviation, and Haar filters of the shape prior and the tissue classified images as features. During each iteration of ACM - our extension of AdaBoost - the Bayesian posterior distribution of the labeling was fed back in as an input, along with its neighborhood features as new features for AdaBoost to use. In validation studies, we compared our results with hand-labeled segmentations by two experts. Using a leave-one-out approach and standard overlap and distance error metrics, our automated segmentations agreed well with human raters; any differences were comparable to differences between trained human raters. Our error metrics compare favorably with those previously reported for other automated hippocampal segmentations, suggesting the utility of the approach for large-scale studies.",
author = "Morra, {Jonathan H.} and Zhuowen Tu and Apostolova, {Liana G.} and Green, {Amity E.} and Christina Avedissian and Madsen, {Sarah K.} and Neelroop Parikshak and Xue Hua and Toga, {Arthur W.} and Jack, {Clifford R Jr.} and Weiner, {Michael W.} and Thompson, {Paul M.}",
year = "2008",
month = "10",
day = "15",
doi = "10.1016/j.neuroimage.2008.07.003",
language = "English (US)",
volume = "43",
pages = "59--68",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",
number = "1",

}

TY - JOUR

T1 - Validation of a fully automated 3D hippocampal segmentation method using subjects with Alzheimer's disease mild cognitive impairment, and elderly controls

AU - Morra, Jonathan H.

AU - Tu, Zhuowen

AU - Apostolova, Liana G.

AU - Green, Amity E.

AU - Avedissian, Christina

AU - Madsen, Sarah K.

AU - Parikshak, Neelroop

AU - Hua, Xue

AU - Toga, Arthur W.

AU - Jack, Clifford R Jr.

AU - Weiner, Michael W.

AU - Thompson, Paul M.

PY - 2008/10/15

Y1 - 2008/10/15

N2 - We introduce a new method for brain MRI segmentation, called the auto context model (ACM), to segment the hippocampus automatically in 3D T1-weighted structural brain MRI scans of subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). In a training phase, our algorithm used 21 hand-labeled segmentations to learn a classification rule for hippocampal versus non-hippocampal regions using a modified AdaBoost method, based on ∼ 18,000 features (image intensity, position, image curvatures, image gradients, tissue classification maps of gray/white matter and CSF, and mean, standard deviation, and Haar filters of size 1 × 1 × 1 to 7 × 7 × 7). We linearly registered all brains to a standard template to devise a basic shape prior to capture the global shape of the hippocampus, defined as the pointwise summation of all the training masks. We also included curvature, gradient, mean, standard deviation, and Haar filters of the shape prior and the tissue classified images as features. During each iteration of ACM - our extension of AdaBoost - the Bayesian posterior distribution of the labeling was fed back in as an input, along with its neighborhood features as new features for AdaBoost to use. In validation studies, we compared our results with hand-labeled segmentations by two experts. Using a leave-one-out approach and standard overlap and distance error metrics, our automated segmentations agreed well with human raters; any differences were comparable to differences between trained human raters. Our error metrics compare favorably with those previously reported for other automated hippocampal segmentations, suggesting the utility of the approach for large-scale studies.

AB - We introduce a new method for brain MRI segmentation, called the auto context model (ACM), to segment the hippocampus automatically in 3D T1-weighted structural brain MRI scans of subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI). In a training phase, our algorithm used 21 hand-labeled segmentations to learn a classification rule for hippocampal versus non-hippocampal regions using a modified AdaBoost method, based on ∼ 18,000 features (image intensity, position, image curvatures, image gradients, tissue classification maps of gray/white matter and CSF, and mean, standard deviation, and Haar filters of size 1 × 1 × 1 to 7 × 7 × 7). We linearly registered all brains to a standard template to devise a basic shape prior to capture the global shape of the hippocampus, defined as the pointwise summation of all the training masks. We also included curvature, gradient, mean, standard deviation, and Haar filters of the shape prior and the tissue classified images as features. During each iteration of ACM - our extension of AdaBoost - the Bayesian posterior distribution of the labeling was fed back in as an input, along with its neighborhood features as new features for AdaBoost to use. In validation studies, we compared our results with hand-labeled segmentations by two experts. Using a leave-one-out approach and standard overlap and distance error metrics, our automated segmentations agreed well with human raters; any differences were comparable to differences between trained human raters. Our error metrics compare favorably with those previously reported for other automated hippocampal segmentations, suggesting the utility of the approach for large-scale studies.

UR - http://www.scopus.com/inward/record.url?scp=52049108022&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=52049108022&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2008.07.003

DO - 10.1016/j.neuroimage.2008.07.003

M3 - Article

C2 - 18675918

AN - SCOPUS:52049108022

VL - 43

SP - 59

EP - 68

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

IS - 1

ER -