Functional magnetic resonance imaging (fMRI) as it exists, in its many forms and vari- ants, has revolutionized the fields of neurology and psychology by revealing functional differences non-invasively. Although blood oxygenation level dependent (BOLD) fMRI is used interchangeably with fMRI, it measures one single difference in a phys- iological measurement using a set sequence. As such, there are other established changes in the brain that relate to blood movement and capacity that can also be measured using MRI. One measure, exogenous steady state cerebral blood volume, uses a bolus routine contrast agent administered intravenously alongside a pair of high resolution ‘structural-like’ MRI images to provide detailed information within small cortical and subcortical structures. In this thesis I design a semi-automated algorithm to generate maps of steady state exogenous cerebral blood volume magnetic resonance imaging datasets. To do this I developed an algorithm and tested it on existing MRI scanning protocols. A series of automated pre-processing steps are developed and tested, including automated scan flagging for artifacts and requisite vascular segmentation. Then, a methodology is developed to create cerebral blood volume (CBV) region of interest (ROI) masks that can then be applied on an existing database to test known CBV dysfunction in a group of patients at high risk for psychosis. Finally, we develop an experiment to see if template based cerebral blood alterations co-registered with class segmentation maps have any positive predictive value in determining disease state in a well characterized cohort of five age-matched groups in an Alzheimer’s disease neuroimaging study.

Includes vita. / Title from PDF of title page. / Document formatted into pages; contains 222 pages. / Includes bibliographical references. / Text (Electronic thesis) in PDF format. / ABSTRACT: This dissertation presents a knowledge-guided expert system that is capable of applying routinesfor multispectral analysis, (un)supervised clustering, and basic image processing to automatically detect and segment brain tissue abnormalities, and then label glioblastoma-multiforme brain tumors in magnetic resonance volumes of the human brain. The magnetic resonance images used here consist of three feature images (T1-weighted, proton density, T2-weighted) and the system is designed to be independent of a particular scanning protocol. Separate, but contiguous 2D slices in the transaxial plane form a brain volume. This allows complete tumor volumes to be measured and if repeat scans are taken over time, the system may be used to monitor tumor response to past treatments and aid in the planning of future treatment. Furthermore, once processing begins, the system is completely unsupervised, thus avoiding the problems of human variability found in supervised segmentation efforts.Each slice is initially segmented by an unsupervised fuzzy c-means algorithm. The segmented image, along with its respective cluster centers, is then analyzed by a rule-based expert system which iteratively locates tissues of interest based on the hierarchy of cluster centers in feature space. Model-based recognition techniques analyze tissues of interest by searching for expected characteristics and comparing those found with previously defined qualitative models. Normal/abnormal classification is performed through a default reasoning method: if a significant model deviation is found, the slice is considered abnormal. Otherwise, the slice is considered normal. Tumor segmentation in abnormal slices begins with multispectral histogram analysis and thresholding to separate suspected tumor from the rest of the intra-cranial region. The tumor is then refined with a variant of seed growing, followed by spatial component analysis and a final thresholding step to remove non-tumor pixels.The knowledge used in this system was extracted from general principles of magnetic resonance imaging, the distributions of individual voxels and cluster centers in feature space, and anatomical information. Knowledge is used both for single slice processing and information propagation between slices. A standard rule-based expert system shell (CLIPS) was modified to include the multispectral analysis, clustering, and image processing tools.A total of sixty-three volume data sets from eight patients and seventeen volunteers (four with and thirteen without gadolinium enhancement) were acquired from a single magnetic resonance imaging system with slightly varying scanning protocols were available for processing. All volumes were processed for normal/abnormal classification. Tumor segmentation was performed on the abnormal slices and the results were compared with a radiologist-labeled ground truth' tumor volume and tumor segmentations created by applying supervised k-nearest neighbors, a partially supervised variant of the fuzzy c-means clustering algorithm, and a commercially available seed growing package. The results of the developed automatic system generally correspond well to ground truth, both on a per slice basis and more importantly in tracking total tumor volume during treatment over time. / System requirements: World Wide Web browser and PDF reader. / Mode of access: World Wide Web.

Diffusion-weighted (DW) magnetic resonance imaging is an important neuroimaging technique that has successful applications in diagnosis of ischemic stroke and methods based on diffusion tensor imaging (DTI). Tensor measures have been used for detecting changes in tissue microstructure and for non-invasively tracing white matter connections in vivo. The most common image acquistion strategy is to use a DW single-shot echo-planar imaging (ss-EPI) pulse sequence, which is attractive due to its robustness to motion artefacts and high imaging speed. However, this sequence has limited achievable spatial resolution and suffers from geometric distortion and blurring artefacts. Readout-segmented echo-planar imaging (rs-EPI) is a DW sequence that is capable of acquiring high-resolution images by segmenting the acquisition of k- space into multiple shots. The fast, short readouts reduce distortion and blurring and the problem of artefacts due to motion-induced phase changes between shots can be overcome with navigator techniques. The rs-EPI sequence has two main shortcomings. (i) The method is slow to produce image volumes, which is limiting for clinical scans due to patient welfare and prevents us from acquiring very many directions in DTI. (ii) The sequence (like other diffusion techniques) is far from the optimum repetition time (TR) for acquiring data with the highest possible signal-to-noise ratio (SNR) in a given time. The work in this thesis seeks to address both of these important issues using a range of approaches. In Chapter 4 a partial Fourier extension is presented, which addresses point (i) by reducing the number of readout segments acquired and estimating the missing data. This allows reductions in scan time by approximately 40% and the reliability of the images is demonstrated in comparisons with the original images. The application of a simultaneous multi-slice scheme to rs-EPI, to address points (i) and (ii), is described in Chapter 5. Using the slice-accelerated rs-EPI sequence, tractography data were compared to ss-EPI data and high-resolution trace-weighted data were acquired in clinically relevant scan times. Finally, a 3D multi-slab extension that addresses point (i) is presented in Chapter 6. A 3D sequence could also allow higher resolution in the slice direction than 2D multi-slice methods, which are limited by the difficulties in exciting thin, accurate slices. A 3D version of rs-EPI was simulated and implemented and a k-space acquisition synchronised to the cardiac cycle showed substantial improvements in image artefacts compared to a conventional k-space acquisition.

The field of cognitive neuroscience has benefited greatly from multimodal investigations of the human brain, which integrate various tools and neuroimaging data to understand brain functions and guide treatments for brain disorders. In this dissertation, we present a series of studies that illustrate the use of multimodal approaches to investigate brain structure and function, brain connectivity, and neuromodulation effects. Firstly, we propose a novel landmark-guided region-based spatial normalization technique to accurately quantify brain morphology, which can improve the sensitivity and specificity of functional imaging studies. Subsequently, we shift the investigation to the characteristics of functional brain activity due to visual stimulations. Our findings reveal that the task-evoked positive blood-oxygen-level dependent (BOLD) response is accompanied by sustained negative BOLD responses in the visual cortex. These negative BOLD responses are likely generated through subcortical neuromodulatory systems with distributed ascending projections to the cortex. To further explore the cortico-subcortical relationship, we conduct a multimodal investigation that involves simultaneous data acquisition of pupillometry, electroencephalography (EEG), and functional magnetic resonance imaging (fMRI). This investigation aims to examine the connectivity of brain circuits involved in the cognitive processes of salient stimuli. Using pupillary response as a surrogate measure of activity in the locus coeruleus-norepinephrine system, we find that the pupillary response is associated with the reorganization of functional brain networks during salience processing. In addition, we propose a cortico-subcortical integrated network reorganization model with potential implications for understanding attentional processing and network switching. Lastly, we employ a multimodal investigation that involves concurrent transcranial magnetic stimulation (TMS), EEG, and fMRI to explore network perturbations and measurements of the propagation effects. In a preliminary exploration on brain-state dependency of TMS-induced effects, we find that the propagation of left dorsolateral prefrontal cortex TMS to regions in the lateral frontoparietal network might depend on the brain-state, as indexed by the EEG prefrontal alpha phase. Overall, the studies in this dissertation contribute to the understanding of the structural and functional characteristics of brain network systems, and may inform future investigations that use multimodal methodological approaches, such as pupillometry, brain connectivity, and neuromodulation tools. The work presented in this dissertation has potential implications for the development of efficient and personalized treatments for major depressive disorder, attention deficit hyperactivity disorder, and Alzheimer's disease.

<p align="justify">L'Imagerie fonctionnelle par Résonance Magnétique (IRMf) a connu un développement important depuis sa découverte au début des années 1990. Basée le plus souvent sur l'effet BOLD (Blood Oxygenation Level Dependent), cette technique permet d'obtenir de façon totalement non-invasive des cartes d'activation cérébrale, avec de meilleures résolutions spatiale et temporelle que les méthodes préexistantes telles que la tomographie par émission de positrons (TEP). Facilement praticable au moyen des imageurs par RMN disponible dans les hôpitaux, elle a mené à de nombreuses applications dans le domaine des neurosciences et de l'étude des pathologies cérébrales.</p><p><p align="justify">Il est maintenant bien établi que l'effet BOLD est dû à une augmentation de l'oxygénation du sang veineux dans les régions du cerveau où se produit l'activation neuronale, impliquant une diminution de la différence de susceptibilité magnétique entre le sang et les tissus environnants (la déoxyhémoglobine étant paramagnétique et l'oxyhémoglobine diamagnétique), et par conséquent un augmentation du signal si la méthode d'acquisition est sensible aux inhomogénéités de champ magnétique. Cependant, il reste encore de nombreuses inconnues quant aux mécanismes liant les variations d'oxygénation, de flux et de volume sanguin à l'augmentation de signal observée, et la dépendance du phénomène en des paramètres tels que l'intensité du champ, la résolution spatiale, et le type de séquence de RMN utilisée. La première partie de la thèse est donc consacrée à l'étude de l'effet BOLD, dans le cas particulier des contributions dues aux veines de drainage dans les séquences de type écho de gradient rendues sensibles au mouvement par l'ajout de gradients de champ. Le modèle développé montre que, contrairement au comportement suggéré par de précédentes publications, l'effet de ces gradients n'est pas une diminution monotone de la différence de signal lorsque l'intensité des gradients augmente. D'importantes oscillations sont produites par l'effet de phase dû au déplacement des spins du sang dans les gradients additionnels, et par la variation de cette phase suite à l'augmentation du flux sanguin. La validation expérimentale du modèle est réalisée au moyen de la séquence PRESTO (Principles of Echo-Shifting combined with a Train of Observations), c'est-à-dire une séquence en écho de gradient où des gradients supplémentaires permettent d'augmenter la sensibilité aux inhomogénéités de champ, et donc à l'effet BOLD. Un accord qualitatif avec la théorie est établi en montrant que la variation de signal observée peut augmenter lorsqu'on intensifie les gradients additionnels.</p><p><p align="justify">Un autre source de débat continuel dans le domaine de l'IRMf réside dans l'optimalisation des méthodes d'acquisition, au point de vue notamment de leur sensibilité à l'effet BOLD, leurs résolutions spatiale et temporelle, leur sensibilité à divers artefacts tels que la perte de signal dans les zones présentant des inhomogénéités de champ à grande échelle, et la contamination des cartes d'activation par les contributions des grosses veines, qui peuvent être distantes du lieu d'activation réel. Les séquences en écho de spin sont connues pour être moins sensibles à ces deux derniers problèmes, c'est pourquoi la deuxième partie de la thèse est consacrée à une nouvelle technique permettant de donner une pondération T2 plutôt que T2* aux images. Le principe de base de la méthode n'est pas neuf, puisqu'il s'agit de la « Préparation T2 » (T2prep), qui consiste à atténuer l'aimantation longitudinale différemment selon la valeur du temps de relaxation T2, mais il n’avait jamais été appliqué à l’IRMf. Ses avantages par rapport à d’autres méthodes hybrides T2 et T2* sont principalement le gain en résolution temporelle et en dissipation d’énergie électromagnétique dans les tissus. Le contraste généré par ces séquences est étudié au moyen de solutions stationnaires des équations de Bloch. Des prédictions sont faites quant au contraste BOLD, sur base de ces solutions stationnaires et d’une description simplifiée de l’effet BOLD en termes de variations de T2 et T2*. Une méthode est proposée pour rendre le signal constant au travers du train d’impulsions en faisant varier l’angle de bascule d’une impulsion à l’autre, ce qui permet de diminuer le flou dans les images. Des expériences in vitro montrent un accord quantitatif excellent avec les prédictions théoriques quant à l’intensité des signaux mesurés, aussi bien dans le cas de l’angle constant que pour la série d’angles variables. Des expériences d’activation du cortex visuel démontrent la faisabilité de l’IRMf au moyen de séquences T2prep, et confirment les prédictions théoriques quant à la variation de signal causée par l’activation.</p><p><p align="justify"> La troisième partie de la thèse constitue la suite logique des deux premières, puisqu’elle est consacrée à une extension du principe de déplacement d’écho (echo-shifting) aux séquences en écho de spin à l’état stationnaire, ce qui permet d’obtenir une pondération T2 et T2* importante tout en maintenant un temps de répétition court, et donc une bonne résolution temporelle. Une analyse théorique approfondie de la formation du signal dans de telles séquences est présentée. Elle est basée en partie sur la technique de résolution des équations de Bloch utilisée dans la deuxième partie, qui consiste à calculer l’aimantation d’état stationnaire en fonction des angles de précession dans le plan transverse, puis à intégrer sur les isochromats pour obtenir le signal résultant d’un voxel (volume element). Le problème est aussi envisagé sous l’angle des « trajectoires de cohérence », c’est-à-dire la subdivision du signal en composantes plus ou moins déphasées, par l’effet combiné des impulsions RF, des gradients appliqués et des inhomogénéités du champ magnétique principal. Cette approche permet d’interpréter l’intensité du signal dans les séquences à écho déplacé comme le résultat d’interférences destructives entre diverses composantes physiquement interprétables. Elle permet de comprendre comment la variation de la phase de l’impulsion d’excitation (RF-spoiling) élimine ces interférences. Des expériences in vitro montrent un accord quantitatif excellent avec les calculs théoriques, et la faisabilité de la méthode in vivo est établie. Il n’est pas encore possible de conclure quant à l’applicabilité de la nouvelle méthode dans le cadre de l’IRMf, mais l’approche théorique proposée a en tout cas permis de revoir en profondeur les mécanismes de formation du signal pour l’ensemble des méthodes à écho déplacé, puisque le cas de l’écho de gradient s’avère complètement similaire au cas de l’écho de spin.</p><p><p align="justify">La thèse évolue donc progressivement de la modélisation de l’effet BOLD vers la conception de séquences, permettant ainsi d’aborder deux aspects fondamentaux de la physique de l’IRMf.</p><p> / Doctorat en sciences appliquées / info:eu-repo/semantics/nonPublished

Indiana University-Purdue University Indianapolis (IUPUI) / Magnetic resonance (MR) spectroscopy is a neuroimaging technique. It is widely used to quantify the concentration of important metabolites in a brain tissue. Imbalance in concentration of brain metabolites has been found to be associated with development of neurological impairment. There has been increasing trend of using MR spectroscopy as a diagnosis tool for neurological disorders. We established statistical methodology to analyze data obtained from the MR spectroscopy in the context of the HIV associated neurological disorder. First, we have developed novel methodology to study the association of marker of neurological disorder with MR spectrum from brain and how this association evolves with time. The entire problem fits into the framework of scalar-on-function regression model with individual spectrum being the functional predictor. We have extended one of the existing cross-sectional scalar-on-function regression techniques to longitudinal set-up. Advantage of proposed method includes: 1) ability to model flexible time-varying association between response and functional predictor and (2) ability to incorporate prior information. Second part of research attempts to study the influence of the clinical and demographic factors on the progression of brain metabolites over time. In order to understand the influence of these factors in fully non-parametric way, we proposed LongCART algorithm to construct regression tree with longitudinal data. Such a regression tree helps to identify smaller subpopulations (characterized by baseline factors) with differential longitudinal profile and hence helps us to identify influence of baseline factors. Advantage of LongCART algorithm includes: (1) it maintains of type-I error in determining best split, (2) substantially reduces computation time and (2) applicable even observations are taken at subject-specific time-points. Finally, we carried out an in-depth analysis of longitudinal changes in the brain metabolite concentrations in three brain regions, namely, white matter, gray matter and basal ganglia in chronically infected HIV patients enrolled in HIV Neuroimaging Consortium study. We studied the influence of important baseline factors (clinical and demographic) on these longitudinal profiles of brain metabolites using LongCART algorithm in order to identify subgroup of patients at higher risk of neurological impairment. / Partial research support was provided by the National Institutes of Health grants U01-MH083545, R01-CA126205 and U01-CA086368