A Bayesian Model of Shape and Appearance for Subcortical Brain Segmentation

FMRIB Centre, Department of Clinical Neurology, University of Oxford, Oxford, UK.
NeuroImage (Impact Factor: 6.36). 02/2011; 56(3):907-22. DOI: 10.1016/j.neuroimage.2011.02.046
Source: PubMed


Automatic segmentation of subcortical structures in human brain MR images is an important but difficult task due to poor and variable intensity contrast. Clear, well-defined intensity features are absent in many places along typical structure boundaries and so extra information is required to achieve successful segmentation. A method is proposed here that uses manually labelled image data to provide anatomical training information. It utilises the principles of the Active Shape and Appearance Models but places them within a Bayesian framework, allowing probabilistic relationships between shape and intensity to be fully exploited. The model is trained for 15 different subcortical structures using 336 manually-labelled T1-weighted MR images. Using the Bayesian approach, conditional probabilities can be calculated easily and efficiently, avoiding technical problems of ill-conditioned covariance matrices, even with weak priors, and eliminating the need for fitting extra empirical scaling parameters, as is required in standard Active Appearance Models. Furthermore, differences in boundary vertex locations provide a direct, purely local measure of geometric change in structure between groups that, unlike voxel-based morphometry, is not dependent on tissue classification methods or arbitrary smoothing. In this paper the fully-automated segmentation method is presented and assessed both quantitatively, using Leave-One-Out testing on the 336 training images, and qualitatively, using an independent clinical dataset involving Alzheimer's disease. Median Dice overlaps between 0.7 and 0.9 are obtained with this method, which is comparable or better than other automated methods. An implementation of this method, called FIRST, is currently distributed with the freely-available FSL package.

    • "Participants with failed MRI quality were excluded. For subcortical volumes, T1-weighted MRI data were processed with the FMRIB Software library (FSL) v5.0.1 (Jenkinson et al., 2012); and 7 subcortical ROIs per hemisphere were generated using FMRIB's Integrated Registration and Segmentation Tool (Patenaude et al., 2011). Values beyond 1.96 "
    [Show abstract] [Hide abstract]
    ABSTRACT: Successful brain aging in the oldest old (≥90 years) is under-explored. This study examined cross-sectional brain morphological differences from eighth to eleventh decades of life in non-demented individuals by high-resolution magnetic resonance imaging. 277 non-demented community dwelling participants (71-103 years) from Sydney Memory and Ageing Study and Sydney Centenarian Study comprised the sample, including a subsample of 160 cognitively high-functioning elders. Relationships between age and MRI-derived measurements were studied using general linear models; and structural profiles of the ≥90 years were delineated. In full sample and the sub-sample, significant linear negative relationship of grey matter with age was found, with the greatest age effects in the medial temporal lobe and parietal and occipital cortices. This pattern was further confirmed by comparing directly the ≥90 years to the 71-89 years groups. Significant quadratic age effects on total white matter and white matter hyperintensities were observed. Our study demonstrated heterogeneous differences across brain regions between the oldest old and young old, with an emphasis on hippocampus, temporo-posterior cortex and white matter hyperintensities.
    No preview · Article · Jan 2016
    • "ICV was estimated based on the determinant of the transformation matrix used when transforming the MR volume to Talairach space. FSL FIRST was utilised to segment subcortical structures for the Sydney MAS and OATS datasets as previously reported (Patenaude, et al., 2011). Input images were registered to MNI M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 7 space through two-stage linear transformation. "
    [Show abstract] [Hide abstract]
    ABSTRACT: INTRODUCTION: Reduction in hippocampal and amygdala volume measured via structural MRI is an early marker of Alzheimer’s disease (AD). Whether genetic risk factors for AD exert an effect on these subcortical structures independent of clinical status has not been fully investigated. METHODS: We examine whether increased genetic risk for AD influences hippocampal and amygdala volumes in case-control and population cohorts at different ages, in 1674 Older (aged >53yrs; 17% AD, 39% MCI) and 467 young (16-30 yrs) adults. RESULTS: An AD polygenic risk score (PRS) combining common risk variants excluding APOE, and a SNP in TREM2, were both associated with reduced hippocampal volume in healthy older adults and those with mild cognitive impairment (MCI). APOE ε4 was associated with hippocampal and amygdala volume in those with AD and MCI, but was not associated in healthy older adults. No associations were found in young adults. DISCUSSION: Genetic risk for AD affects the hippocampus before the clinical symptoms of AD, reflecting a neurodegenerative effect prior to clinical manifestations in older adults.
    No preview · Article · Jan 2016
    • "The quality of MRI output after manual editing was then classified into " pass " – all regions usable, " partial " – more than 30 of the 34 cortical ROIs usable, and " fail " – excluded from our sample pool. Volumes of subcortical ROIs were generated using FMRIB's Integrated Registration and Segmentation Tool (FIRST) [34]. Values beyond 1.96 "
    [Show abstract] [Hide abstract]
    ABSTRACT: Underpinnings of mild cognitive impairment (MCI) change with increasing age. We hypothesize that MRI signatures of mild cognitive impairment (MCI) would be different at a higher age compared to younger elders. Methods - 244 participants (71-103 years) from the Sydney Memory and Ageing Study and the Sydney Centenarian Study were categorized as amnestic MCI (aMCI), non-amnestic MCI (naMCI) or cognitively normal (CN). Brain "atrophy" and white matter hyper-intensities (WMHs) associated with MCI subtypes and age effects were examined by general linear models, controlling for confounding factors. Reduced logistic regressions were performed to determine structures that best discriminated aMCI from CN in individuals <85 and those ≥85 years. Results - aMCI was associated with smaller volumes of overall cortex, medial temporal structures, anterior corpus callosum, and select frontal and parietal regions compared to CN; such associations did not significantly change with age. Structures that best discriminated aMCI from CN differed however in the <85 and ≥85 age groups: cortex, putamen, parahippocampal, precuneus and superior frontal cortex in <85 years, and the hippocampus, pars triangularis and temporal pole in ≥85 years. Differences between naMCI and CN were small and non-significant in the sample. WMHs were not significantly associated with MCI subtypes. Conclusions - Structural MRI distinguishes aMCI, but not naMCI, from CN in elderly individuals. The structures that best distinguish aMCI from CN differ in those <85 from those ≥85, suggesting different neuropathological underpinnings of cognitive impairment in the very old.
    No preview · Article · Dec 2015 · Current Alzheimer research
Show more