Morphology-driven automatic segmentation of MR images of the neonatal brain

The segmentation of MR images of the neonatal brain is an essential step in the study and evaluation of infant brain development. State-of-the-art methods for adult brain MRI segmentation are not applicable to the neonatal brain, due to large differences in structure and tissue properties between newborn and adult brains. Existing newborn brain MRI segmentation methods either rely on manual interaction or require the use of atlases or templates, which unavoidably introduces a bias of the results towards the population that was used to derive the atlases. We propose a different approach for the segmentation of neonatal brain MRI, based on the infusion of high-level brain morphology knowledge, regarding relative tissue location, connectivity and structure. Our method does not require manual interaction, or the use of an atlas, and the generality of its priors makes it applicable to different neonatal populations, while avoiding atlas-related bias. The proposed algorithm segments the brain both globally (intracranial cavity, cerebellum, brainstem and the two hemispheres) and at tissue level (cortical and subcortical gray matter, myelinated and unmyelinated white matter, and cerebrospinal fluid). We validate our algorithm through visual inspection by medical experts, as well as by quantitative comparisons that demonstrate good agreement with expert manual segmentations. The algorithm's robustness is verified by testing on variable quality images acquired on different machines, and on subjects with variable anatomy (enlarged ventricles, preterm- vs. term-born). (C) 2012 Elsevier B.V. All rights reserved.

Global tractography with embedded anatomical priors for quantitative connectivity analysis

Alessandro Daducci, Alia Lemkaddem, Didrik Hans Malcolm Skiöldebrand, Jean-Philippe Thiran

Tractography algorithms provide us with the ability to non-invasively reconstruct fiber pathways in the white matter (WM) by exploiting the directional information described with diffusion magnetic resonance. These methods could be divided into two major classes, local and global. Local methods reconstruct each fiber tract iteratively by considering only directional information at the voxel level and its neighborhood. Global methods, on the other hand, reconstruct all the fiber tracts of the whole brain simultaneously by solving a global energy minimization problem. The latter have shown improvements compared to previous techniques but these algorithms still suffer from an important shortcoming that is crucial in the context of brain connectivity analyses. As no anatomical priors are usually considered during the reconstruction process, the recovered fiber tracts are not guaranteed to connect cortical regions and, as a matter of fact, most of them stop prematurely in the WM; this violates important properties of neural connections, which are known to originate in the gray matter (GM) and develop in the WM. Hence, this shortcoming poses serious limitations for the use of these techniques for the assessment of the structural connectivity between brain regions and, de facto, it can potentially bias any subsequent analysis. Moreover, the estimated tracts are not quantitative, every fiber contributes with the same weight toward the predicted diffusion signal. In this work, we propose a novel approach for global tractography that is specifically designed for connectivity analysis applications which: (i) explicitly enforces anatomical priors of the tracts in the optimization and (ii) considers the effective contribution of each of them, i.e., volume, to the acquired diffusion magnetic resonance imaging (MRI) image. We evaluated our approach on both a realistic diffusion MRI phantom and in vivo data, and also compared its performance to existing tractography algorithms.

2014

Quantitative Susceptibility Mapping in the Human Brain

Diana Khabipova

Magnetic resonance imaging (MRI) offers a good tissue contrast and the ability to visualize many disease related morphologies. The work presented in this thesis investigates the study of underlying structure of the brain using quantitative methods with a special emphasis on quantitative susceptibility mapping (QSM). Magnetic susceptibility reflects the interaction of a material to the magnetic field and measures in biological tissues the magnetic susceptibility of inclusions. The reconstruction of QSM requires further processing steps as the magnetic field produced by the sources needs to be disentangled from the orders of magnitude bigger background field. The produced field also depends not only on the shape and the orientation, but also on the anisotropy of susceptibility and the microstructural compartmentalization of the biological source. For this reason, reconstruction methods need to be capable to calculate accurate values for different brain regions as well as applicable in the everyday clinical diagnosis. Within the framework of the thesis a data acquisition protocol based on a multiple-echo gradient echo sequence as well as a post-processing protocol was implemented. One of the processing steps, the background removal method, was applied to preserve the brain regions close to the cerebrospinal fluid (CSF). This method outperforms state of the art methods in this regions but is computationally intensive. Different brain regions were studied using quantitative methods with special emphasis on the QSM. A new method, modulated closed form solution, with extremely fast computational time is proposed. The comparison with other single orientation methods revealed similar results and the highest correlation to the state-of-the-art method (COSMOS) in the deep gray matter. The R2* maps calculated from the same dataset are also able to distinguish the deep gray matter structures with a similar quality. However, QSM shows a higher sensitivity in early stage multiple sclerosis lesions as well as white matter-gray matter structures. In the human cortex the obtained cortical maps show enhancement of primary sensory cortex, which is known to be highly myelinated, on three evaluated quantitative contrasts R1,R2* and susceptibility. The contrasts based on the relaxation rates, R1 and R2*, show a monotonically decrease from the white matter to the CSF imitating the decrease in iron and myelin. The susceptibility behaviour is more complex as iron and myelin content introduce an opposing sensitivity, allowing to study iron and myelin content when combining the three contrasts. The microstructural organization of white matter influences the R2*, R2 as well as field map from which QSM is calculated. This structure leads to an orientation dependence of the studied contrasts and for QSM the spherical assumption is not valid anymore. Therefore a new QSM method is introduced, which includes the Lorentzian correction in white matter. Main fibres such as forceps major and minor were analysed for the three different quantitative contrasts. The anisotropic component associated with susceptibility is similar for the relaxation rates whereas the isotropic component of R2* shows a higher variability. The resulting deep gray matter structure of the new QSM method remained similar to the state-of-the-art method when comparing the isotropic component but calculates physically meaningful susceptibility maps with improved contrast between known fibre bundles.

EPFL2016

AxTract: Toward microstructure informed tractography

Alessandro Daducci, Maxime Descoteaux, Gabriel Girard, Jean-Philippe Thiran

Diffusion-weighted (DW) magnetic resonance imaging (MRI) tractography has become the tool of choice to probe the human brain's white matter in vivo. However, tractography algorithms produce a large number of erroneous streamlines (false positives), largely due to complex ambiguous tissue configurations. Moreover, the relationship between the resulting streamlines and the underlying white matter microstructure characteristics remains poorly understood. In this work, we introduce a new approach to simultaneously reconstruct white matter fascicles and characterize the apparent distribution of axon diameters within fascicles. To achieve this, our method, AxTract, takes full advantage of the recent development DW-MRI microstructure acquisition, modeling, and reconstruction techniques. This enables AxTract to separate parallel fascicles with different microstructure characteristics, hence reducing ambiguities in areas of complex tissue configuration. We report a decrease in the incidence of erroneous streamlines compared to the conventional deterministic tractography algorithms on simulated data. We also report an average increase in streamline density over 15 known fascicles of the 34 healthy subjects. Our results suggest that microstructure information improves tractography in crossing areas of the white matter. Moreover, AxTract provides additional microstructure information along the fascicle that can be studied alongside other streamline-based indices. Overall, AxTract provides the means to distinguish and follow white matter fascicles using their microstructure characteristics, bringing new insights into the white matter organization. This is a step forward in microstructure informed tractography, paving the way to a new generation of algorithms able to deal with intricate configurations of white matter fibers and providing quantitative brain connectivity analysis. Hum Brain Mapp 38:5485-5500, 2017. (C) 2017 Wiley Periodicals, Inc.

Wiley-Blackwell2017