Fast and robust parameter estimation for statistical partial volume models in brain MRI.

Digital Media Institute/Signal Processing, Tampere University of Technology, FIN-33101, Finland.
NeuroImage (Impact Factor: 6.13). 10/2004; 23(1):84-97. DOI: 10.1016/j.neuroimage.2004.05.007
Source: PubMed

ABSTRACT Due to the finite spatial resolution of imaging devices, a single voxel in a medical image may be composed of mixture of tissue types, an effect known as partial volume effect (PVE). Partial volume estimation, that is, the estimation of the amount of each tissue type within each voxel, has received considerable interest in recent years. Much of this work has been focused on the mixel model, a statistical model of PVE. We propose a novel trimmed minimum covariance determinant (TMCD) method for the estimation of the parameters of the mixel PVE model. In this method, each voxel is first labeled according to the most dominant tissue type. Voxels that are prone to PVE are removed from this labeled set, following which robust location estimators with high breakdown points are used to estimate the mean and the covariance of each tissue class. Comparisons between different methods for parameter estimation based on classified images as well as expectation--maximization-like (EM-like) procedure for simultaneous parameter and partial volume estimation are reported. The robust estimators based on a pruned classification as presented here are shown to perform well even if the initial classification is of poor quality. The results obtained are comparable to those obtained using the EM-like procedure, but require considerably less computation time. Segmentation results of real data based on partial volume estimation are also reported. In addition to considering the parameter estimation problem, we discuss differences between different approximations to the complete mixel model. In summary, the proposed TMCD method allows for the accurate, robust, and efficient estimation of partial volume model parameters, which is crucial to a variety of brain MRI data analysis procedures such as the accurate estimation of tissue volumes and the accurate delineation of the cortical surface.


Available from: Jussi Tohka, Jun 03, 2015
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The magnetization transfer ratio (MTR) and diffusion tensor imaging (DTI) correlates of early brain development were examined in cohort of 18 very preterm neonates (27-31 gestational weeks) presenting with normal radiological findings scanned within 2 weeks after birth (28-32 gestational weeks). A combination of non-linear image registration, tissue segmentation, and voxel-wise regression was used to map the age dependent changes in MTR and DTI-derived parameters in 3D across the brain based on the cross-sectional in vivo preterm data. The regression coefficient maps obtained differed between brain regions and between the different quantitative MRI indices. Significant linear increases as well as decreases in MTR and DTI-derived parameters were observed throughout the preterm brain. In particular, the lamination pattern in the cerebral wall was evident on parametric and regression coefficient maps. The frontal white matter area (subplate and intermediate zone) demonstrated a linear decrease in MTR. While the intermediate zone showed an unexpected decrease in fractional anisotropy (FA) with age, with this decrease (and the increase in mean diffusivity (MD)) driven primarily by an increase in radial diffusivity (RD) values, the subplate showed no change in FA (and an increase in MD). The latter was the result of a concomitant similar increase in axial diffusivity (AD) and RD values. Interpreting the in vivo results in terms of available histological data, we present a biophysical model that describes the relation between various microstructural changes measured by complementary quantitative methods available on clinical scanners and a range of maturational processes in brain tissue. Copyright © 2015. Published by Elsevier Inc.
    NeuroImage 02/2015; 112. DOI:10.1016/j.neuroimage.2015.02.051 · 6.13 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: People differ in their cognitive functioning. This variability has been exhaustively examined at the behavioral, neural and genetic level to uncover the mechanisms by which some individuals are more cognitively efficient than others. Studies investigating the neural underpinnings of interindividual differences in cognition aim to establish a reliable nexus between functional/structural properties of a given brain network and higher order cognitive performance. However, these studies have produced inconsistent results, which might be partly attributed to methodological variations. In the current study, 82 healthy young participants underwent MRI scanning and completed a comprehensive cognitive battery including measurements of fluid, crystallized, and spatial intelligence, along with working memory capacity/executive updating, controlled attention, and processing speed. The cognitive scores were obtained by confirmatory factor analyses. T1 -weighted images were processed using three different surface-based morphometry (SBM) pipelines, varying in their degree of user intervention, for obtaining measures of cortical thickness (CT) across the brain surface. Distribution and variability of CT and CT-cognition relationships were systematically compared across pipelines and between two cognitively/demographically matched samples to overcome potential sources of variability affecting the reproducibility of findings. We demonstrated that estimation of CT was not consistent across methods. In addition, among SBM methods, there was considerable variation in the spatial pattern of CT-cognition relationships. Finally, within each SBM method, results did not replicate in matched subsamples. Hum Brain Mapp, 2015. © 2015 Wiley Periodicals, Inc. © 2015 Wiley Periodicals, Inc.
    Human Brain Mapping 05/2015; DOI:10.1002/hbm.22843 · 6.92 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Despite potential implications for the early detection of impending Alzheimer's disease (AD), very little is known about the differences of large-scale brain networks between amnestic mild cognitive impairment (aMCI) with high cerebral amyloid-beta protein (Aβ) deposition (i.e., aMCI+) and aMCI with no or very little Aβ deposition (i.e., aMCI-). We first aimed to extend the current literature on altering intrinsic functional connectivity (FC) of the default mode network (DMN) and salience network (SN) from cognitively normal (CN) to AD dementia. Second, we further examined the differences of the DMN and the SN between aMCI-, aMCI+, and CN. Forty-three older adult (12 CN, 10 aMCI+, 10 aMCI-, and 11 AD dementia) subjects were included. All participants received comprehensive clinical and neuropsychological assessment, resting-state functional magnetic resonance imaging, structural MRI, and Pittsburgh compound-B-PET scans. FC data were preprocessed using multivariate exploratory linear optimized decomposition into independent components of FMRIB's Software Library. Group comparisons were carried out using the "dual-regression" approach. In addition, to verify presence of gray matter volume changes with intrinsic functional network alterations, voxel-based morphometry was performed on the acquired T1-weighted data. As expected, AD dementia participants exhibited decreased FC in the DMN compared to CN (particularly in the precuneus and cingulate gyrus). The degree of alteration in the DMN in aMCI+ compared to CN was intermediate to that of AD. In contrast, aMCI- exhibited increased FC in the DMN compared to CN (primarily in the precuneus) as well as aMCI+. In terms of the SN, aMCI- exhibited decreased FC compared to both CN and aMCI+ particularly in the inferior frontal gyrus. FC within the SN in aMCI+ and AD did not differ from CN. Compared to CN, aMCI- showed atrophy in bilateral superior temporal gyri whereas aMCI+ showed atrophy in right precuneus. The results indicate that despite the similarity in cross-sectional cognitive features, aMCI- has quite different functional brain connectivity compared to aMCI+.
    Frontiers in Aging Neuroscience 02/2015; 7:15. DOI:10.3389/fnagi.2015.00015 · 2.84 Impact Factor