A method for simultaneously segmenting multiple anatomical brain structures from multi-modal MR images has been developed. An artificial neural network (ANN) was trained from a set of feature vectors created by a combination of high-resolution registration methods, atlas based spatial probability distributions, and a training set of 16 expert traced data sets. A set of feature vectors were adapted to increase performance of ANN segmentation; 1) a modified spatial location for structural symmetry of human brain, 2) neighbors along the priors' descent for directional consistency, and 3) candidate vectors based on the priors for the segmentation of multiple structures. The trained neural network was then applied to 8 data sets, and the results were compared with expertly traced structures for validation purposes. Comparing several reliability metrics, including a relative overlap, similarity index, and intraclass correlation of the ANN generated segmentations to a manual trace are similar or higher to those measures previously developed methods. The ANN provides a level of consistency between subjects and time efficiency comparing human labor that allows it to be used for very large studies.

Segmentation of sub-cortical Gray Matter (GM) structures in magnetic resonance brain images is crucial in clinic and research for many purposes such as early diagnosis of neurological diseases, guidance of surgical operations and longitudinal volumetric studies. Unfortunately, the algorithms that segment the brain into 3 tissues usually suffer from poor performance in the sub-cortical region. In order to increase the detection of sub-cortical GM structures, an evolutionary hierarchical region merging approach, abbreviated as EHRM, is proposed in this study. Through EHRM, an intensity based region merging is utilized while merging is allowed to proceed among disconnected regions. Texture information is also incorporated into the scheme to prevent the region merging between tissues with similar intensity but different texture properties. The proposed algorithm is tested on real and simulated datasets. The performance is compared with a popular segmentation algorithm, which is also intensity driven: the FAST algorithm [1] in the widely used FSL suite. EHRM is shown to make a significant improvement the detection of sub-cortical GM structures. Average improvements of 10%, 36% and 22% are achieved for caudate, putamen and thalamus respectively. The accuracy of volumetric estimations also increased for GM and WM. Performance of EHRM is robust in presence of bias field. In addition, EHRM operates in O(N) complexity. Furthermore, the algorithm proposed here is simple, because it does not incorporate spatial priors such as an atlas image or intensity priors. With these features, EHRM may become a favorable alternative to the existing brain segmentation tools.

Multiple Sclerosis (MS) is an autoimmune disease of central nervous system. It may result in a variety of symptoms from blurred vision to severe muscle weakness and degradation, depending on the affected regions in brain. To better understand this disease and to quantify its evolution, magnetic resonance imaging (MRI) is increasingly used nowadays. Manual delineation of MS lesions in MR images by human expert is time-consuming, subjective, and prone to inter-expert variability. Therefore, automatic segmentation is needed as an alternative to manual segmentation. However, the progression of the MS lesions shows considerable variability and MS lesions present temporal changes in shape, location, and area between patients and even for the same patient, which renders the automatic segmentation of MS lesions a challenging problem. In this dissertation, a set of segmentation pipelines are proposed for automatic segmentation of multiple sclerosis (MS) lesions from brain magnetic resonance imaging (MRI) data. These techniques use a trained support vector machine (SVM) to discriminate between the blocks in regions of MS lesions and the blocks in non-MS lesion regions mainly based on the textural features with aid of the other features. The main contribution of this set of frameworks is the use of textural features to detect MS lesions in a fully automated approach that does not rely on manually delineating the MS lesions. In addition, the technique introduces the concept of the multi-sectional views segmentation to produce verified segmentation. The multi-sectional views pipeline is customized to provide better segmentation performance and to benefit from the properties and the nature of MS lesion in MRI. These customization and enhancement leads to development of the customized MV-T-SVM. The MRI datasets that were used in the evaluation of the proposed pipelines are simulated MRI datasets (3 subjects) generated using the McGill University BrainWeb MRI Simulator, real datasets (51 subjects) publicly available at the workshop of MS Lesion Segmentation Challenge 2008 and real MRI datasets (10 subjects) for MS subjects acquired at the University of Miami. The obtained results indicate that the proposed method would be viable for use in clinical practice for the detection of MS lesions in MRI.

In this thesis, we aim to investigate the correlation between myelination and the gestational age for preterm infants, with the former being an important developmental process during human brain maturation. Quantification of myelin requires dedicated imaging, but the conventional magnetic resonance images routinely acquired during clinical imaging of neonates carry signatures that are thought to be associated with myelination. This thesis thus focuses on structural segmentation and spatio-temporal modelling of the so-called myelin-like signals on T2-weighted scans for early prognostic evaluation of the preterm brain. The segmentation part poses the major challenges of this task: insufficient spatial prior information of myelination and the presence of substantial partial volume voxels in clinical data. Specific spatial priors for the developing brain are obtained from either probabilistic atlases or manually annotated training images, but none of them currently include myelin as an individual tissue type. This causes further difficulties in partial volume estimation which depends on the probabilistic atlases of the composing pure tissues. Our key contribution is the development of an expectation-maximisation framework that incorporates an explicit partial volume class whose locations are configured in relation to the composing pure tissues in a predefined region of interest via second-order Markov random fields. This approach resolves the above challenges without requiring any probabilistic atlas of myelin. We also investigate atlas-based whole brain segmentation that generates the binary mask for the region of interest. We then construct a spatio-temporal growth model for myelin-like signals using logistic regression based on the automatic segmentations of 114 preterm infants aged between 29 and 44 gestational weeks. Lastly, we demonstrate the ability of age estimation using the normal growth model in a leave-one-out procedure.

Segmentation of the brain into sub-volumes has many clinical applications. Manyneurological diseases are connected with brain atrophy (tissue loss). By dividingthe brain into smaller compartments, volume comparison between the compartmentscan be made, as well as monitoring local volume changes over time. Theformer is especially interesting for the left and right cerebral hemispheres, dueto their symmetric appearance. By using automatic segmentation, the time consumingstep of manually labelling the brain is removed, allowing for larger scaleresearch.In this thesis, three automatic methods for segmenting the brain from magneticresonance (MR) images are implemented and evaluated. Since neither ofthe evaluated methods resulted in sufficiently good segmentations to be clinicallyrelevant, a novel segmentation method, called SB-GC (shape bottleneck detectionincorporated in graph cuts), is also presented. SB-GC utilizes quantitative MRIdata as input data, together with shape bottleneck detection and graph cuts tosegment the brain into the left and right cerebral hemispheres, the cerebellumand the brain stem. SB-GC shows promises of highly accurate and repeatable resultsfor both healthy, adult brains and more challenging cases such as childrenand brains containing pathologies.

L’essor de l’imagerie médicale par résonance magnétique (IRM) permet une exploration de plus en plus précise du cerveau en période néonatale. Comment interpréter le plus objectivement possible des images dont les particularités compliquent l’analyse ? La controverse autour des hyperintensités diffuses de la substance blanche (diffuse excessive high signal intensity, DEHSI) en est une illustration. Le premier objectif est d’étudier la variabilité des appréciations des radiologues. Il existe une bonne reproductibilité des mesures bidimensionnelles des structures cérébrales, mais une reproductibilité intra et inter-observateurs moyenne de l’analyse visuelle de l’intensité de signal de la substance blanche néonatale. Le second objectif est le développement d’une méthode de segmentation utilisant des outils de traitement d’images, essentiellement morphologiques, en particulier des opérateurs connexes. Elle permet de segmenter la substance grise, la substance blanche et le liquide cérébro-spinal à l’étage sus-tentoriel et détecter automatiquement la présence d’hyperintensités de la substance blanche. Une mesure normalisée de la sévérité de celles-ci par rapport à la substance blanche adjacente est calculée, ce qui constitue une contribution originale de la thèse. La validation des résultats sur des images acquises avec des champs magnétiques de 1,5 et 3 T par comparaison à des segmentations manuelles autorise l’utilisation de ce logiciel. La confrontation des résultats obtenus au suivi clinique à long terme de nouveau-nés permettra de mieux connaître et interpréter le développement cérébral visualisé en IRM et d’apporter une réponse face au défi que constituent les DEHSI. / Progress in magnetic resonance imaging (MRI) has allowed more detailed exploration of the development and maturation of the neonatal brain. Among the challenges facing radiologists are determining how best to objectively analyze images with very different characteristics when compared to older children. One issue is the “diffuse excessive high signal intensity” (DEHSI) of the white matter in premature newborns, whose definition, classification and prognosis have been vigorously debated. The role played in this analysis by the subjectivity of the radiological interpretation is not well understood. Our primary objective was to study the variability of this subjective analysis by the radiologist. Although reproducibility is acceptable for bi-dimensional measurement of brain structures, it is only fair for the analysis of signal intensity of brain white matter. The secondary objective was the design of a robust and reliable semi-automatic method to segment the gray matter, the white matter, and the cerebrospinal fluid and detect potential high signal intensity regions (it calculates a normalized mean value, and compares it to the normal surrounding white matter.). The algorithm is composed of an isotropic diffusion filter, morphological tools and connected operators, all implemented in a software interface. The results of this algorithm have been validated on MRI images acquired on 1.5 and 3 T devices by comparing them with segmentation results. This new tool could be employed in routine MRI. Correlation of the results with clinical outcomes in infants would permit a better understanding of cerebral development and, particularly, elucidate the significance of DEHSI.

The Pulse Couple Neural Network (PCNN) was developed by Eckhorn to model the observed synchronization of neural assemblies in the visual cortex of small mammals such as a cat. In this dissertation, three novel PCNN based automatic segmentation algorithms were developed to segment Magnetic Resonance Imaging (MRI) data: (a) PCNN image 'signature' based single region cropping; (b) PCNN - Kittler Illingworth minimum error thresholding and (c) PCNN -Gaussian Mixture Model - Expectation Maximization (GMM-EM) based multiple material segmentation. Among other control tests, the proposed algorithms were tested on three T2 weighted acquisition configurations comprising a total of 42 rat brain volumes, 20 T1 weighted MR human brain volumes from Harvard's Internet Brain Segmentation Repository and 5 human MR breast volumes. The results were compared against manually segmented gold standards, Brain Extraction Tool (BET) V2.1 results, published results and single threshold methods. The Jaccard similarity index was used for numerical evaluation of the proposed algorithms. Our quantitative results demonstrate conclusively that PCNN based multiple material segmentation strategies can approach a human eye's intensity delineation capability in grayscale image segmentation tasks.

Comment obtenir des données anatomiques pendant une neurochirurgie ? a été ce qui a guidé le travail développé dans le cadre de cette thèse. Les IRM sont actuellement utilisées en amont de l'opération pour fournir cette information, que ce soit pour le diagnostique ou pour définir le plan de traitement. De même, ces images pre-opératoires peuvent aussi être utilisées pendant l'opération, pour pallier la difficulté et le coût des images per-opératoires. Pour les rendre utilisables en salle d'opération, un recalage doit être effectué avec la position du patient. Cependant, le cerveau subit des déformations pendant la chirurgie, phénomène appelé Brain Shift, ce qui altère la qualité du recalage. Pour corriger cela, d'autres données pré-opératoires peuvent être acquises, comme la localisation de la surface corticale, ou encore des images US localisées en 3D. Ce nouveau recalage permet de compenser ce problème, mais en partie seulement. Ainsi, des modèles mécaniques ont été développés, entre autres pour apporter des solutions à l'amélioration de ce recalage. Ils permettent ainsi d'estimer les déformations du cerveau. De nombreuses méthodes existent pour implémenter ces modèles, selon différentes lois de comportement et différents paramètres physiologiques. Dans tous les cas, cela requiert un modèle anatomique patient-spécifique. Actuellement, ce modèle est obtenu par contourage manuel, ou quelquefois semi-manuel. Le but de ce travail de thèse est donc de proposer une méthode automatique pour obtenir un modèle du cerveau adapté sur l'anatomie du patient, et utilisable pour une simulation mécanique. La méthode implémentée se base sur les modèles déformables pour segmenter les structures anatomiques les plus pertinentes dans une modélisation bio-mécanique. En effet, les membranes internes du cerveau sont intégrées: falx cerebri and tentorium cerebelli. Et bien qu'il ait été démontré que ces structures jouent un rôle primordial, peu d'études les prennent en compte. Par ailleurs, la segmentation résultante de notre travail est validée par comparaison avec des données disponibles en ligne. De plus, nous construisons un modèle 3D, dont les déformations seront simulées en utilisant une méthode de résolution par Éléments Finis. Ainsi, nous vérifions par des expériences l'importance des membranes, ainsi que celle des paramètres physiologiques. / The general problem that motivates the work developed in this thesis is: how to obtain anatomical information during a neurosurgery?. Magnetic Resonance (MR) images are usually acquired before the surgery to provide anatomical information for diagnosis and planning. Also, the same images are commonly used during the surgery, because to acquire MRI images in the operating room is complex and expensive. To make these images useful inside the operating room, a registration between them and the patient's position has to be processed. The problem is that the brain suffers deformations during the surgery, in a process called brain shift, degrading the quality of registration. To correct this, intra-operative information may be used, for example, the position of the brain surface or US images localized in 3D. The new registration will compensate this problem, but only to a certain extent. Mechanical models of the brain have been developed as a solution to improve this registration. They allow to estimate brain deformation under certain boundary conditions. In the literature, there are a variety of methods for implementing these models, different equation laws used for continuum mechanic, and different reported mechanical properties of the tissues. However, a patient specific anatomical model is always required. Currently, most mechanical models obtain the associated anatomical model by manual or semi-manual segmentation. The aim of this thesis is to propose and implement an automatic method to obtain a model of the brain fitted to the patient's anatomy and suitable for mechanical modeling. The implemented method uses deformable model techniques to segment the most relevant anatomical structures for mechanical modeling. Indeed, the internal membranes of the brain are included: falx cerebri and tentorium cerebelli. Even though the importance of these structures is stated in the literature, only a few of publications include them in the model. The segmentation obtained by our method is assessed using the most used online databases. In addition, a 3D model is constructed to validate the usability of the anatomical model in a Finite Element Method (FEM). And the importance of the internal membranes and the variation of the mechanical parameters is studied.

No description available.