From volumetric graphical objects to surface registration

Relatively recent developments in medical imaging allow the anatomical and functional noninvasive study of internal organs of the body. One of the most complex and interesting organs in our body is the brain. Many efforts are being carried out to try to understand how it works, its internal connections, how different zones interact with each other, why, in some cases, a zone takes in charge a functionality of another damaged one, etc.; the questions are innumerable and the answers, if any, are sparse. Recently, magnetic resonance imaging (MRI) equipments have been developed to capture the functional activity of the brain. Unfortunately, functional images suffer from a lack of spatial resolution. Thus, the need of combining them with anatomical ones, which have a much higher quality. Nowadays these images provide precious information for diagnosis purposes and this is the reason for the need of imaging tools to extract and analyze the information which is present in those images, and which sometimes is not appreciable for the human eye. Acquired MRI images are presented as volumetric sets of data with which one can reconstruct a three dimensional volumetric model of the scanned organ. Nevertheless, sometimes physicians are just interested in the surface of the organ, for example in brain functional activation studies. In addition, many times it is important to be able to carry out group studies, or to locate a lesion with respect to a standard reference. To do such studies it is necessary to perform a transformation of the surfaces in order to put them into correspondence. Such technique is called registration. Unfortunately, volumetric registration does not give satisfactory results for surface registration purposes. The solution is to work directly with surfaces instead of processing the volumetric data. Unfortunately we do not have information about the surfaces, hence the need of computer graphics and computer vision techniques. The present work proposes a number of methods to help in the task of providing information from volumetric objects to surface registration. To achieve that some stages are necessary. The proposed stages are independent one from another so they can be used, or not, depending on the particular need. They can even be used for other purposes than those presented in the present work. The first two stages address the problem of image quality. This is a very important "preprocessing" step because posterior processing will depend on the image quality. Usually MR images are corrupted with different artifacts; among them a spatial-dependent bias field and a Riciandistributed noise. For the former a supervised correction is proposed, and for the later a directional spatial filter using a novel diffusivity function. After both stages the volumetric MRI data is ready to be classified. The purpose of the proposed classification process is to extract the white matter of the brain, given that sulci are better extracted from white matter than from the gray one. So far we have handled volumetric data, the next stage is concerned with the registration of the reconstructed white matter surface. This is achieved by means of a hierarchical representation of brain's sulci using level sets. The hierarchical representation and surface registration are carried out using the theory of partial differential equations (PDE) and level set methods. Validation studies are carried out independently in order to show the performance of the proposed approaches. Also a full scheme is proposed with the aim of presenting a complete approach from volumetric objects to surface registration.

Hybrid Representation and Visualization Methods Applied to Neuronal Microcircuits

Sébastien Lasserre

For the past twenty years, scientific visualization has assisted simulation-based research to explore problems of organized complexity. Environmental research, social sciences, astronomy, and modern physics have exploited the increasing power of computer graphics to interactively visualize three-dimensional models. Of late, neuroscience has witnessed the use of large-scale simulation with biologically detailed neuron models, and aspires to benefit from scientific visualization too. This thesis is the result of a close collaboration with the Blue Brain Project, a neuroscience endeavor working on biologically detailed large-scale simulation of the brain, and tries to answer the following question: How can scientific visualization improve neuroscience insights in the context of simulation-based research? Scientific visualization of a model typically preserves its spatial layout. In continuity with this axiom, we identified in more depth the requirements for a scientific visualization solution in neuroscience. Thus, the preservation of the domain specificity, the hybrid visualization, the validation, and the adaptation to the limitations of the human-visual perception enhance the readability of the generated results. We applied these requirements to elaborate an hybrid representation of neuronal structures, and proposed a unified framework of visualization methods which preserves the domain specificities of neuroscience. This framework was finally assessed in case studies, where neuroscientific investigations about structure, composition, and dynamics of neuronal microcircuit challenge the developed visualization solutions to derive insight. We believe that these representations and visualization methods form a solid ground for the elaboration of virtual instruments dedicated to a new approach of neuroscience, where in silico landscapes of the brain are complementary to the classical in vitro and in vivo investigations.

EPFL2010

Quantitative morphometry analysis of the fetal brain using clinical MR imaging

Meritxell Bach Cuadra, Gabriele Vincenzo Bonanno, Marie Schaer, Jean-Philippe Thiran

In vivo fetal magnetic resonance imaging provides a unique approach for the study of early human brain development [1]. In utero cerebral morphometry could potentially be used as a marker of the cerebral maturation and help to distinguish between normal and abnormal development in ambiguous situations. However, this quantitative approach is a major challenge because of the movement of the fetus inside the amniotic cavity, the poor spatial resolution provided by very fast MRI sequences and the partial volume effect. Extensive efforts are made to deal with the reconstruction of high-resolution 3D fetal volumes based on several acquisitions with lower resolution [2,3,4]. Frameworks were developed for the segmentation of specific regions of the fetal brain such as posterior fossa, brainstem or germinal matrix [5,6], or for the entire brain tissue [7,8], applying the Expectation-Maximization Markov Random Field (EM-MRF) framework. However, many of these previous works focused on the young fetus (i.e. before 24 weeks) and use anatomical atlas priors to segment the different tissue or regions. As most of the gyral development takes place after the 24th week, a comprehensive and clinically meaningful study of the fetal brain should not dismiss the third trimester of gestation. To cope with the rapidly changing appearance of the developing brain, some authors proposed a dynamic atlas [8]. To our opinion, this approach however faces a risk of circularity: each brain will be analyzed / deformed using the template of its biological age, potentially biasing the effective developmental delay. Here, we expand our previous work [9] to propose post-processing pipeline without prior that allow a comprehensive set of morphometric measurement devoted to clinical application. Data set & Methods: Prenatal MR imaging was performed with a 1-T system (GE Medical Systems, Milwaukee) using single shot fast spin echo (ssFSE) sequences (TR 7000 ms, TE 180 ms, FOV 40 x 40 cm, slice thickness 5.4mm, in plane spatial resolution 1.09mm). For each fetus, 6 axial volumes shifted by 1 mm were acquired under mother’s sedation (about 1min per volume). First, each volume is segmented semi-automatically using region-growing algorithms to extract fetal brain from surrounding maternal tissues. Inhomogeneity intensity correction [10] and linear intensity normalization are then performed. Brain tissues (CSF, GM and WM) are then segmented based on the low-resolution volumes as presented in [9]. A high-resolution image with isotropic voxel size of 1.09 mm is created as proposed in [2] and using B-splines for the scattered data interpolation [11]. Basal ganglia segmentation is performed using a levet set implementation on the high-resolution volume [12]. The resulting white matter image is then binarized and given as an input in FreeSurfer software (http://surfer.nmr.mgh.harvard.edu) to provide topologically accurate three-dimensional reconstructions of the fetal brain according to the local intensity gradient. References: [1] Guibaud, Prenatal Diagnosis 29(4) (2009). [2] Rousseau, Acad. Rad. 13(9), 2006. [3] Jiang, IEEE TMI 2007. [4] Warfield IADB, MICCAI 2009. [5] Claude, IEEE Trans. Bio. Eng. 51(4) 2004. [6] Habas, MICCAI 2008. [7] Bertelsen, ISMRM 2009. [8] Habas, Neuroimage 53(2) 2010. [9] Bach Cuadra, IADB, MICCAI 2009. [10] Styner, IEEE TMI 19(39 (2000). [11] Lee, IEEE Trans. Visual. And Comp. Graph. 3(3), 1997. [12] Bach Cuadra, ISMRM 2010.

2011

Development and implementation of ¹H magnetic resonance techniques for an improved in vivo assessment of models of cerebral metabolic disorders

Mélanie Craveiro

Nuclear magnetic resonance (NMR) is a physical phenomenon that is widely used in the biomedical field due to its non-invasive and non-destructive properties, which make it an optimal tool for the in vivo investigation of living organs such as the brain. This thesis focused on the development of 1H magnetic resonance (MR) techniques at ultra-high magnetic field strength to improve the measurement and quantification of the 1H MR signal in the rodent brain, and to accurately assess the alterations of the metabolite and water concentrations in several cerebral metabolic disorders. In rodents, 1H MR spectroscopy (MRS) allows the assessment of the concentration of up to 20 metabolites that take part in many cerebral metabolic processes such as neurotransmission, cell energy metabolism, cell growth and osmosis. 1H MRS thus provides a unique tool to non-invasively investigate the cerebral metabolism in healthy and pathological conditions in vivo. However, it is essential for reliable quantification that systematic errors and overlap of the measured metabolite signals are minimized. To that aim, the impact of a potential additional short T2 relaxation time component, which might affect the glutamine quantification at long echo times, was assessed in this thesis. The J-difference editing technique MEGA-SPECIAL was optimized to obtain the unequivocal detection of the glutamine signal at moderate echo time. As a control, the glutamine concentration obtained with this method was then compared to short echo time 1H MR spectroscopy measurements. Since the two measurements of the glutamine concentration at short and moderate echo time did not result in significant differences, this study concluded that there is a low probability of an effect of a short T2 relaxation time component on the glutamine concentration measurement. Since a reliable quantification of 1H spectra partly relies on the accurate assessment of the overlapping macromolecule contribution to the metabolites, an optimized method for the post-processing of the measured macromolecule signal was developed to ensure an accurate assessment of its contribution. This method was applied to investigate potential regional differences in the mouse brain macromolecule signals that may affect metabolite quantification when not taken into account, as well as for the assessment of macromolecule alterations in a human-glioma mouse model. No regional macro- molecule variation was found to significantly affect the metabolite quantification in the healthy mouse brain, which supports the common use of a general macromolecule spec- trum for healthy rodent brain 1H spectra quantification. However, several alterations of the macromolecule spectrum, some of which were reported for the first time, were observed in glioma tissues, and their accurate assessment was shown to be necessary for reliable metabolite quantification. Localized 1H MR spectroscopy techniques that are designed to acquire the1H MR signal from a 3D volume can also be combined with MR imaging (MRI) to acquire information on the spatial distribution of the individual metabolites, which is often of interest when heterogeneous tissues are investigated (such as in pathological conditions). This MR spectroscopic imaging (MRSI) technique is however limited in its conventional implementation by the long measurement time that is necessary to acquire the spatial information. A fast MRSI technique, spectroscopic RARE (spRARE), was therefore implemented at high magnetic field strength in this thesis in order to benefit from the advantage of high magnetic fields in addition to the rapid spatial encoding scheme offered by spRARE. Several optimizations of the metabolite signal detection were implemented in spRARE, before the efficiency of the technique was validated at 9.4 T through a comparison of the acquired in vitro metabolite signals with numerical simulations using the density matrix formalism. The in vivo quantification of the water content is also of interest as it offers perspec- tives for investigating cerebral disease progression that is associated with brain water content alterations. In addition, when the water signal is used as an internal reference for metabolite concentration quantification in MRS, such knowledge can also contribute to a more reliable quantification. A method that is based on the multi-spin-echo MRI technique was therefore developed for the absolute quantification of the water content in the rodent brain, with the aim of assessing alterations of its content in pathological conditions. It was then applied on a bile-duct-ligation rat model of chronic hepatic encephalopathy (HE) to assess the brain water content changes associated with the disease as well as its metabolic alterations. The absence of significant brain water content changes detected in the investigated HE rat model, which was supported by post-mortem brain water content measurements using gravimetry and the dry/wet weight-ratio technique, potentially opens the way for investigating cerebral metabolic processes in chronic HE in absence of edema formation. Localized 1H MRS and conventional MRSI were also applied for the investigation of rodent models of brain disorders in order to assess the metabolic profile associated with those disorders. These studies for the first time allowed the assessment of GABA as a potential early Parkinson’s disease marker, while they also provided new insights into the role of the PPAR-β brain receptor in the early neuroprotective effects following ischemia.

EPFL2013