Dissertations / Theses on the topic 'Brain Magnetic resonance imaging Statistical methods'

In this thesis, we have focused on the image pre-processing in order to enhance the image information. The main aspects of this enhancement rely on removing any image noise and correcting any intensity bias induced by the scanner. Besides, we also contributed with a new technique based on a multispectral, adaptive, region growing algorithm in order to segment the brain from the rest of the head. We include, as a pre-processing step, the image registration process, in which we proposed a novel pipeline by using information from multiple modalities to improve the results of this process. Furthermore, we have also studied the current techniques for the detection and segmentation of WML, proposing a new method based on a previous proposal. Therefore, we presented a tool able to automatically detect and segment WML of Multiple sclerosis and Lupus patients.
En aquesta tesi ens centrem, per una part, en el pre-processat de la imatge per tal d'eliminar el soroll i corregir les inhomogeneïtats en les intensitats, ambdós errors introduïts per l'escàner. A més hem contribuït també amb una nova tècnica basada en un algoritme de “región growing” per tal de segmentar el cervell de dins de tota la imatge del cap. Incloem com a pre-processat el registre d'imatges, on hem proposat una “pipeline" mitjançant la informació de múltiples modalitats per tal de millorar els resultats d'aquest procés. Per altra banda, hem estudiat també les tècniques actuals de detecció i segmentació de lesions en la matèria blanca, proposant un mètode nou basat en anteriors propostes. Així doncs, presentem una eina automàtica capaç de detectar i segmentar lesions en la matèria blanca de pacients d'Esclerosi Múltiple i Lupus.

The number of people suffering from neurodegenerative diseases, such as Alzheimer`s disease, increased dramatically over the past centuries and is expected to increase even further within the next years. Based on predictions of the World Health Organization and Alzheimer`s Disease International, 115 million people will suffer from dementia by the year 2050. An additionally increase in other age related neurodegenerative diseases is also forecasted. Quite naturally, neurodegenerative diseases became a focus of attention of governments and health insurances, trying to control the uprising financial burden. Early detection and treatment of neurodegenerative diseases could be an important component in containing this problem. In particular, researchers focused on automatic methods to analyze patients’ imaging data. One way to detect structural changes in magnetic resonance images (MRI) is the voxel-based method approach. It was specifically implemented for various imaging modalities, e.g. T1-weighted images or diffusion tensor imaging (DTI). Voxel-based morphometry (VBM), a method specifically designed to analyze T1-weighted images, has become very popular over the last decade. Investigations using VBM revealed numerous structural brain changes related to, e.g. neurodegeneration, learning induced structural changes or aging. Although voxel-based methods are designed to be robust and reliable structural change detection methods, it is known that they can be influenced by physical and physiological factors. Dehydration, for example, can affect the volume of brain structures and possibly induce a confound in morphometric studies. Therefore, three-dimensional T1-weighted images were acquired of six young and healthy subjects during different states of hydration. Measurements during normal hydration, hyperhydration, and dehydration made it possible to assess consequential volume changes in gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). The datasets were analyzed using VBM, FreeSurfer and SIENA. A significant decrease of GM and WM volume, associated with dehydration, was found in various brain regions. The most prominent effects were found in temporal and parietal areas, in the left inferior orbito-frontal region, and in the extra-nuclear region. Moreover, we found consistent increases in CSF, an expansion around 6% of the ventricular system affecting both lateral ventricles, i.e. the third and fourth ventricle. Similar degrees of shrinkage in WM volume and increase of the ventricular system have been reported in studies of Alzheimer’s disease during disease progression and in its prestage mild cognitive impairment. Based on these findings, a potential confound in GM and WM or CSF studies due to the subjects’ hydration state cannot be excluded and should be appropriately addressed. These results underline the sensitivity of VBM and might also concern other voxel-based methods, such as Tract-Based Spatial Statistics (TBSS). TBSS was specifically designed for WM analyses and its sensitivity might be helpful for revealing the spatial relation of structural WM changes and related blood serum biomarkers. Two common brain related biomarkers are the glial protein S100B, a plasticity inducing neuro- and gliotrophin, and neuron-specific enolase (NSE), a marker for neuronal damage. However, the spatial specificity of these biomarkers for brain region has not been investigated in vivo until now. Therefore, we acquired two MRI parameters – T1- weighted and DTI - sensitive to changes in GM and WM, and obtained serum S100B and NSE levels of 41 healthy subjects. Additionally, the gene expression of S100B on the whole brain level in a male cohort of three subjects from the Allen Brain Database was analyzed. Furthermore, a female post mortal brain was investigated using double immunofluorescence labeling with oligodendrocyte markers. It could be shown that S100B is specifically related to white matter structures, namely the corpus callosum, anterior forceps and superior longitudinal fasciculus in female subjects. This effect was observed in fractional anisotropy and radial diffusivity – the latest an indicator of myelin changes. Histological data confirmed a co-localization of S100B with oligodendrocyte markers in the human corpus callosum. S100B was most abundantly expressed in the corpus callosum according to the whole genome Allen Human Brain Atlas. In addition, NSE was related to gray matter structures, namely the amygdala. This effect was detected across sexes. The data demonstrates a very high S100B expression in white matter tracts, in particular in human corpus callosum. This was the first in vivo study validating the specificity of the glial marker S100B for the human brain, and supporting the assumption that radial diffusivity represents a myelin marker. The results open a new perspective for future studies investigating major neuropsychiatric disorders. All above mentioned studies are mainly dependent on the sensitivity and accuracy of soft and hardware parameters. In particular, technical developments have improved acquisition accuracy in the field of MRI. Interestingly, very little is known about the confounding effects of variations due to hardware parameters and their possible impact on reliability and sensitivity of VBM. Recent studies have shown that different acquisition parameters may influence VBM results. Therefore age-related GM changes were investigated with VBM in 36 healthy volunteers grouped into 12 young, 12 middle-aged and 12 elderly subject. Six T1-weighted datasets were acquired per subject with a 12-channel matrix coil, as well as a 32-channel array, MP-RAGE and MP2RAGE, and with isotropic resolutions of 0.8 and 1 mm. DARTEL-VBM was applied on all images and GM, WM and CSF segments were statistically analyzed.. Paired t-tests and statistical interaction tests revealed significant effects of acquisition parameters on the estimated gray-matter-density (GMD) in various cortical and subcortical brain regions. MP2RAGE seemed slightly less prone to false positive results when comparing data acquired with different RF coils and yielded superior segmentation of deep GM structures. With the 12-channel coil, MP-RAGE was superior in detecting age-related changes, especially in cortical structures. Most differences between both sequences became insignificant with the 32-channel coil, indicating that the MP2RAGE images benefited more from the improved signal-to-noise ratio and improved parallel-imaging reconstruction). A possible explanation might be an overestimation of the GM compartment on the MP-RAGE images. In view of substantial effects obtained for all parameters, careful standardization of the acquisition protocol is advocated. While the current investigation focused on aging effects, similar results are expected for other VBM studies, like on plasticity or neurodegenerative diseases. This work has shown that voxel-based methods are sensitive to subtle structural brain changes, independent of imaging modality and scanning parameters. In particular, the studies investigated and discussed the analysis of T1- and diffusion weighted images with VBM and TBSS in the context of dehydration, blood serum sensitive biomarkers and aging were discussed. The major goal of these studies was the investigation of the sensitivity of voxel-based methods. In conclusion, sensitivity and accuracy of voxelbased methods is already high, but it can be increased significantly, using optimal hardand software parameters. It is of note, though, that these optimizations and the concomitant increase of detection sensitivity could also introduce additional confounding factors in the imaging data and interfere with the latter preprocessing and statistical computations. To avoid an interference e.g. originating from physiological parameters, a very careful selection and monitoring of biological parameters of each volunteer throughout the whole study is recommended. A potential impact of scanning parameters can be minimized by strict adherence to the imaging protocol for each study subjectwithin a study. A general increase in detection sensitivity due to optimized parameters selection in hard- and/or can not be concluded by the above mentioned studies. Although the present work addressed some of those issues, the topic of optimal selection of parameters for morphometric studies is still very complex and controversial and has to be individually decided. Further investigations are needed to define more general scanning and preprocessing standards to increase detection sensitivity without the concomitant amplification of confounding factors.

Structural brain networks can be constructed at a macroscopic scale using diffusion magnetic resonance imaging (dMRI) and whole-brain tractography. Under this approach, grey matter regions, such as Brodmann areas, form the nodes of a network and tractography is used to construct a set of white matter fibre tracts which form the connections. Graph-theoretic measures may then be used to characterise patterns of connectivity. In this study, we measured the test-retest properties of such networks by varying several factors affecting network construction using ten healthy volunteers who underwent a dMRI protocol at 1.5 T on two separate occasions. High resolution T1-weighted brains were parcellated into regions-of-interest and network connections were identified using dMRI and two alternative tractography algorithms, two alternative seeding strategies, constraints on anatomical plausibility and three alternative network weightings. Test-retest performance was found to improve when: 1) seeding from white matter, rather than grey; and 2) using probabilistic tractography, rather than deterministic. In terms of network weighting, a measure of streamline density produced better test-retest performance than tract-averaged diffusion anisotropy, although it remains unclear which is most representative of the underlying axonal connections. These findings were then used to inform network construction for two further cohorts: a casecontrol analysis of 30 patients with amyotrophic lateral sclerosis (ALS) compared with 30 age-matched healthy controls; and a cross-sectional analysis of 80 healthy volunteers aged 25– 64 years. In both cases, networks were constructed using a weighting reflecting tract-averaged fractional anisotropy (FA). A mass-univariate statistical technique called network-based statistics, identified an impaired motor-frontal-subcortical subnetwork (10 nodes and 12 bidirectional connections), consistent with upper motor neuron pathology, in the ALS group compared with the controls. Reduced FA for three of the impaired network connections, which involved fibres of the cortico-spinal tract, were significantly correlated with the rate of disease progression. Cross-sectional analysis of the 80 healthy volunteers was intended to provide supporting evidence for the widely reported age-related decline in white matter integrity. However, no meaningful relationships were found between increasing age and impaired connectivity based on global, lobar and nodal network properties – findings which were confirmed with a conventional voxel-based analysis of the dMRI data. In conclusion, whilst current acquisition protocols and methods can produce networks capable of characterising the genuine between-subject differences in connectivity, it is challenging to measure subtle white matter changes, for example, due to normal ageing. We conclude that future work should be undertaken to address these concerns.

An ever-increasing number of functional magnetic resonance imaging (fMRI) studies are now using information-based multi-voxel pattern analysis (MVPA) techniques to decode mental states. In doing so, they achieve a significantly greater sensitivity compared to when they use univariate analysis frameworks. Two most prominent MVPA methods for information mapping are searchlight decoding and classifier weight mapping. The new MVPA brain mapping methods, however, have also posed new challenges for analysis and statistical inference on the group level. In this thesis, I discuss why the usual procedure of performing t-tests on MVPA derived information maps across subjects in order to produce a group statistic is inappropriate. I propose a fully nonparametric solution to this problem, which achieves higher sensitivity than the most commonly used t-based procedure. The proposed method is based on resampling methods and preserves the spatial dependencies in the MVPA-derived information maps. This enables to incorporate a cluster size control for the multiple testing problem. Using a volumetric searchlight decoding procedure and classifier weight maps, I demonstrate the validity and sensitivity of the new approach using both simulated and real fMRI data sets. In comparison to the standard t-test procedure implemented in SPM8, the new results showed a higher sensitivity and spatial specificity. The second goal of this thesis is the comparison of the two widely used information mapping approaches -- the searchlight technique and classifier weight mapping. Both methods take into account the spatially distributed patterns of activation in order to predict stimulus conditions, however the searchlight method solely operates on the local scale. The searchlight decoding technique has furthermore been found to be prone to spatial inaccuracies. For instance, the spatial extent of informative areas is generally exaggerated, and their spatial configuration is distorted. In this thesis, I compare searchlight decoding with linear classifier weight mapping, both using the formerly proposed non-parametric statistical framework using a simulation and ultra-high-field 7T experimental data. It was found that the searchlight method led to spatial inaccuracies that are especially noticeable in high-resolution fMRI data. In contrast, the weight mapping method was more spatially precise, revealing both informative anatomical structures as well as the direction by which voxels contribute to the classification. By maximizing the spatial accuracy of ultra-high-field fMRI results, such global multivariate methods provide a substantial improvement for characterizing structure-function relationships.

Abstract Multimodal measurements are increasingly being employed in the study of human physiology. Brain studies in particular can draw advantage of simultaneous measurements using different modalities to analyse correlations, mechanisms and relationships of physiological signals and their dynamics in relation to brain functions. Moreover, multimodal measurements help to identify components of physiological dynamics generated specifically by the brain. This thesis summarizes the study, design and development of non-invasive medical instruments that can be utilized in conjunction with magnetic resonance imaging (MRI). A key challenge in the development of measurement methods is posed by the extraordinary requirements that the MRI environment poses - all materials need to be MR-compatible and the selected instruments and devices must not be affected by the strong magnetic field generated by the MRI scanner nor the MRI by the instruments placed within its scanning volume. The presented methods allow simultaneous continuous measurement of heart rate (HR) and metabolism from the brain cortex as well as pulse wave velocity (PWV) and blood pressure measurements in synchrony with electroencephalography (EEG) and MRI. Furthermore, the thesis explored the reliability and accuracy of the responses gathered by the developed instruments and, using new experimental methods, estimated the propagation of near-infrared light in the human brain. The goal of the novel multimodal measurement environment is to provide more extensive tools for medical researchers, neurologists in particular, to acquire accurate information on the function of the brain and the human body. Measurements have been performed on more than 70 persons using the presented multimodal setup to study such factors as the correlation between blood oxygen level-dependent (BOLD) data and low-frequency oscillations (LFOs) during the resting state
Tiivistelmä Multimodaalisia kuvantamismenetelmiä käytetään enenevässä määrin ihmisen fysiologian ja elintoimintojen tutkimisessa. Erityisesti aivotutkimuksessa samanaikaisesti useammalla modaliteetilla mittaaminen mahdollistaa erilaisten fysiologisten mekanismien ja niiden korrelaatioiden tutkimisen kehon ja aivotoimintojen välillä. Lisäksi multimodaaliset mittaukset auttavat yksilöimään fysiologiset komponentit toisistaan ja identifioimaan aivojen tuottamia fysiologisia signaaleja. Tämä väitöskirja kokoaa tutkimustyön sekä laite- ja instrumentointisuunnittelun ja sen kehittämistyön ei-invasiivisesti toteutettujen lääketieteen mittausmenetelmien käyttämiseksi magneettikuvauksen aikana. Erityishaasteena työssä on ollut magneettikuvausympäristö, joka asettaa erityisvaatimuksia mm. mittalaitteissa käytettäville materiaaleille sekä laitteiden häiriönsiedolle magneettikuvauslaitteen aiheuttaman voimakkaan magneettikentän takia. Kehitettävät mittausmenetelmät eivät myöskään saa aiheuttaa häiriöitä magneettikuvauslaitteen tuottamalle kuvainformaatiolle. Väitöskirjassa esitettävät mittausmenetelmät tekevät mahdolliseksi mitata magneettikuvausympäristössä ihmisen sydämen sykettä, veren virtauksen kulkunopeutta ja verenpaineen vaihteluja sekä aivokuoren metaboliaa - kaikki synkronissa aivosähkökäyrän mittaamisen ja magneettikuvantamisen kanssa. Lisäksi väitöskirjassa tutkitaan kehitettyjen mittausmenetelmien antamaa mittaustarkkuutta sekä arvioidaan lähi-infrapunavalon etenemistä ihmisen aivoissa uudenlaisella menetelmällä. Kehitetyllä multimodaalisella mittausympäristöllä on tavoitteena antaa lääketieteen alan tutkijoille, erityisesti neurologeille, käyttöön mittausmenetelmiä, joiden avulla voidaan tutkia ihmisen aivojen ja kehon välisiä toimintoja aiempaa kattavammin. Laitekokonaisuudella on tutkittu jo yli 70:tä henkilöä. Näissä mittauksissa on tutkittu mm. veren happitasojen hitaita vaihteluja ihmisen aivojen ollessa lepotilassa, ns. resting state -tilassa

Diffusion magnetic resonance imaging (dmri) is a technique that can be used to examine the diffusion characteristics of water in the living brain. A recently developed application of this technique is tractography, in which information from brain images obtained using dmri is used to reconstruct the pathways which connect regions of the brain together. Proxy measures for the integrity, or coherence, of these pathways have also been defined using dmri-derived information. The disconnection hypothesis suggests that specific neurological impairments can arise from damage to these pathways as a consequence of the resulting interruption of information flow between relevant areas of cortex. The development of dmri and tractography have generated a considerable amount of renewed interest in the disconnectionist thesis, since they promise a means for testing the hypothesis in vivo in any number of pathological scenarios. However, in order to investigate the effects of pathology on particular pathways, it is necessary to be able to reliably locate them in three-dimensional dmri images. The aim of the work described in this thesis is to improve upon the robustness of existing methods for segmenting specific white matter tracts from image data, using tractography, and to demonstrate the utility of the novel methods for the comparative analysis of white matter integrity in groups of subjects. The thesis begins with an overview of probability theory, which will be a recurring theme throughout what follows, and its application to machine learning. After reviewing the principles of magnetic resonance in general, and dmri and tractography in particular, we then describe existing methods for segmenting particular tracts from group data, and introduce a novel approach. Our innovation is to use a reference tract to define the topological characteristics of the tract of interest, and then search a group of candidate tracts in the target brain volume for the best match to this reference. In order to assess how well two tracts match we define a heuristic but quantitative tract similarity measure. In later chapters we demonstrate that this method is capable of successfully segmenting tracts of interest in both young and old, healthy and unhealthy brains; and then describe a formalised version of the approach which uses machine learning methods to match tracts from different subjects. In this case the similarity between tracts is represented as a matching probability under an explicit model of topological variability between equivalent tracts in different brains. Finally, we examine the possibility of comparing the integrity of groups of white matter structures at a level more fine-grained than a whole tract.

This document is a critical synopsis of prior work by Michael Dwyer submitted in support of a PhD by published work. The selected work is focused on the application of quantitative magnet resonance imaging (MRI) analysis techniques to the study of multiple sclerosis (MS). MS is a debilitating disease with a multi-factorial pathology, progression, and clinical presentation. Its most salient feature is focal inflammatory lesions, but it also includes significant parenchymal atrophy and microstructural damage. As a powerful tool for in vivo investigation of tissue properties, MRI can provide important clinical and scientific information regarding these various aspects of the disease, but precise, accurate quantitative analysis techniques are needed to detect subtle changes and to cope with the vast amount of data produced in an MRI session. To address this, eight new techniques were developed by Michael Dwyer and his co-workers to better elucidate focal, atrophic, and occult/"invisible" pathology. These included: a method to better evaluate errors in lesion identification; a method to quantify differences in lesion distribution between scanner strengths; a method to measure optic nerve atrophy; a more precise method to quantify tissue-specific atrophy; a method sensitive to dynamic myelin changes; and a method to quantify iron in specific brain structures. Taken together, these new techniques are complementary and improve the ability of clinicians and researchers to reliably assess various key elements of MS pathology in vivo.

This thesis presents methods for quantitative MRI analysis of brain injury, repair and degeneration in multiple sclerosis (MS) that provide new insights into disease pathogenesis and evolution.
Demyelinated and inflammatory white-matter lesions are hallmark features of MS. A methodology is described to detect regions of acute white-matter lesions that undergo myelin destruction and repair based on analysis of magnetization transfer ratio (MTR) images. Validation is performed based on histopathology and error is assessed based on same-day scans. To quantify the spatial extent and temporal evolution of myelin destruction and repair, data from a 3-year clinical trial is analyzed using this method. Approximately 20% of acute lesion voxels show some repair over the initial 7 months. In subsequent months, there is little further repair, but some increases in the lesion volume undergoing demyelination.
Although less conspicuous on conventional MRI, there is considerable MS pathology in the brain tissue outside of white-matter lesions. An image-processing methodology was developed to obtain accurate metrics that quantify change over time in whole-brain MTR (associated with changes in myelin-density) and in T2 relaxation time (associated with changes in inflammatory edema). These metrics, in addition to metrics of brain atrophy and axonal integrity, were used to quantify brain injury and degeneration following immunoablation and autologous hematopoietic stem cell transplantation therapy for MS. Pronounced brain volume loss was detected immediately following therapy, associated with decreased myelin density and not resolution of edema.
Post-mortem histopathology has revealed abnormalities in the cortical grey-matter of MS patients that appear to be independent of white-matter lesions. A methodology to quantify neocortical injury and degeneration that yields cross-sectional and longitudinal metrics of cortical thickness and grey-matter/white-matter interface integrity both globally and regionally is presented and validated. MS patients with progressive disability showed greater decreases in cortical metrics compared to MS patients with stable disability.
The quantitative MRI analysis methods presented in this thesis are applicable to MRI data obtained in clinical trials of therapies for MS, have the necessary sensitivity and specificity to assess therapeutic efficacy, and provide new insights into disease pathogenesis and evolution.

Brain atrophy has been used as a marker for disease progression in Multiple Sclerosis (MS). SIENA, an automated tool for measuring brain volume change, was tested to see whether MRI slice thickness and gap presence affect longitudinal atrophy measures. Isotropic global scan-rescan images were used to simulate 3 mm and 5 mm axial slice thicknesses with 1 and 2mm gaps, respectively. SIENA remained accurate and precise with increasing slice thickness and gap presence. Furthermore, symmetric pre-registration was crucial for scans with larger slice-thickness and gaps.
SIENA was used to observe atrophy in children who have experienced a Clinically Isolated Syndrome (CIS) of the type leading to MS (CIS-MS). Brain atrophy was present within the first three months after a CIS event, and then subsided over the rest of the year. If the first acute episode was excluded, there was no significant difference in atrophy rates between the CIS-MS group and the CIS group, and no significant difference between those with T2-weighted brain lesions versus those who had none.

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.

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.

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.

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.

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.

Doctorat en sciences appliquées
info:eu-repo/semantics/nonPublished

This thesis is focused on developing novel and fully automated methods for the detection of new multiple sclerosis (MS) lesions in longitudinal brain magnetic resonance imaging (MRI). First, we proposed a fully automated logistic regression-based framework for the detection and segmentation of new T2-w lesions. The framework was based on intensity subtraction and deformation field (DF). Second, we proposed a fully convolutional neural network (FCNN) approach to detect new T2-w lesions in longitudinal brain MR images. The model was trained end-to-end and simultaneously learned both the DFs and the new T2-w lesions. Finally, we proposed a deep learning-based approach for MS lesion synthesis to improve the lesion detection and segmentation performance in both cross-sectional and longitudinal analysis
Esta tesis se centra en el desarrollo de métodos novedosos y totalmente automatizados para la detección de nuevas lesiones de esclerosis múltiple en la resonancia magnética longitudinal del cerebro. Primero, propusimos un marco totalmente automatizado basado en la regresión logística para la detección y segmentación de nuevas lesiones T2-w. El marco se basaba en la sustracción de intensidad y el campo de deformación (DF). En segundo lugar, propusimos un enfoque de red neuronal totalmente convolucional para detectar nuevas lesiones T2-w en imágenes de resonancia magnética del cerebro longitudinal. El modelo se entrenó de extremo a extremo y aprendió simultáneamente tanto los DF como las nuevas lesiones T2-w. Por último, propusimos un enfoque basado en el aprendizaje profundo para la síntesis de las lesiones de la EM, a fin de mejorar el rendimiento de la detección y la segmentación de las lesiones tanto en el análisis transversal como en el longitudinal

The interplay between sounds and movements is not only critical for music performance, but also for the acquisition of speech, and might underlie the success of using music as a therapeutic tool in the facilitation of movements. This dissertation is comprised of three functional magnetic resonance imaging studies that aim to elucidate the neural basis underlying interactions between the auditory and motor systems in the context of musical rhythm perception and production. Study 1 investigated the neural correlates that facilitate auditory-motor coupling while subjects tapped along with an isochronous rhythm. Auditory input was manipulated so that the metric saliency of the isochronous rhythm increased across five parametric levels in order to modulate subjects’ tapping behaviour. [...]
L’interaction entre le son et le mouvement n’est pas seulement essentielle lors de prestations musicales, mais aussi lors de l’ acquisition de la parole, et pourrait être à la base du succès de la musique lorsqu’elle est utilisée en tant qu’agent thérapeutique visant la facilitation du mouvement. Cette dissertation consiste en trois études d’imagerie par résonance magnétique fonctionnelle visant à élucider les fondements neuraux à la base de l’interaction entre le système auditif et le système moteur dans le contexte de la perception et de la production de rythmes musicaux. La première étude examina les corrélats neuraux facilitant le couplage auditif-moteur chez des sujets produisant des battements alors qu’ils étaient guidés par un rythme isochronique. L’information auditive fut manipulée pour que la proéminence métrique du rythme isochronique augmente à travers cinq niveaux paramétriques dans le but de moduler les battements produits par le sujet. [...]

Comment peut-on accumuler de la connaissance sur les fonctions cérébrales ? Comment peut-on bénéficier d'années de recherche en IRM fonctionnelle (IRMf) pour analyser des processus cognitifs plus fins et construire un modèle exhaustif du cerveau ? Les chercheurs se basent habituellement sur des études individuelles pour identifier des régions cérébrales recrutées par les processus cognitifs. La comparaison avec l'historique du domaine se fait généralement manuellement pas le biais de la littérature, qui permet de définir des régions d'intérêt dans le cerveau. Les méta-analyses permettent de définir des méthodes plus formelles et automatisables pour analyser la littérature. Cette thèse examine trois manières d'accumuler et d'organiser les connaissances sur le fonctionnement du cerveau en utilisant des cartes d'activation cérébrales d'un grand nombre d'études. Premièrement, nous présentons une approche qui utilise conjointement deux expériences d'IRMf similaires pour mieux conditionner une analyse statistique. Nous montrons que cette méthode est une alternative intéressante par rapport aux analyses qui utilisent des régions d'intérêts, mais demande cependant un travail manuel dans la sélection des études qui l'empêche de monter à l'échelle. A cause de la difficulté à sélectionner automatiquement les études, notre deuxième contribution se focalise sur l'analyse d'une unique étude présentant un grand nombre de conditions expérimentales. Cette méthode estime des réseaux fonctionnels (ensemble de régions cérébrales) et les associe à des profils fonctionnels (ensemble pondéré de descripteurs cognitifs). Les limitations de cette approche viennent du fait que nous n'utilisons qu'une seule étude, et qu'elle se base sur un modèle non supervisé qui est par conséquent plus difficile à valider. Ce travail nous a cependant apporté la notion de labels cognitifs, qui est centrale pour notre dernière contribution. Cette dernière contribution présente une méthode qui a pour objectif d'apprendre des atlas fonctionnels en combinant plusieurs jeux de données. [Henson2006] montre qu'une inférence directe, c.a.d. la probabilité d'une activation étant donné un processus cognitif, n'est souvent pas suffisante pour conclure sur l'engagement de régions cérébrales pour le processus cognitif en question. Réciproquement, [Poldrack 2006] présente l'inférence inverse qui est la probabilité qu'un processus cognitif soit impliqué étant donné qu'une région cérébrale est activée, et décrit le risque de raisonnements fallacieux qui peuvent en découler. Pour éviter ces problèmes, il ne faut utiliser l'inférence inverse que dans un contexte où l'on suffisamment bien échantillonné l'espace cognitif pour pouvoir faire une inférence pertinente. Nous présentons une méthode qui utilise un « meta-design » pour décrire des tâches cognitives avec un vocabulaire commun, et qui combine les inférences directe et inverse pour mettre en évidence des réseaux fonctionnels qui sont cohérents à travers les études. Nous utilisons un modèle prédictif pour l'inférence inverse, et effectuons les prédictions sur de nouvelles études pour s'assurer que la méthode n'apprend pas certaines idiosyncrasies des données d'entrées. Cette dernière contribution nous a permis d'apprendre des réseaux fonctionnels, et de les associer avec des concepts cognitifs. Nous avons exploré différentes approches pour analyser conjointement des études d'IRMf. L'une des difficultés principales était de trouver un cadre commun qui permette d'analyser ensemble ces études malgré leur diversité. Ce cadre s'est instancié sous la forme d'un vocabulaire commun pour décrire les tâches d'IRMf. et a permis d'établir un modèle statistique du cerveau à grande échelle et d'accumuler des connaissances à travers des études d'IRM fonctionnelle
How can we accumulate knowledge on brain functions? How can we leverage years of research in functional MRI to analyse finer-grained psychological constructs, and build a comprehensive model of the brain? Researchers usually rely on single studies to delineate brain regions recruited by mental processes. They relate their findings to previous works in an informal way by defining regions of interest from the literature. Meta-analysis approaches provide a more principled way to build upon the literature. This thesis investigates three ways to assemble knowledge using activation maps from a large amount of studies. First, we present an approach that uses jointly two similar fMRI experiments, to better condition an analysis from a statistical standpoint. We show that it is a valuable data-driven alternative to traditional regions of interest analyses, but fails to provide a systematic way to relate studies, and thus does not permit to integrate knowledge on a large scale. Because of the difficulty to associate multiple studies, we resort to using a single dataset sampling a large number of stimuli for our second contribution. This method estimates functional networks associated with functional profiles, where the functional networks are interacting brain regions and the functional profiles are a weighted set of cognitive descriptors. This work successfully yields known brain networks and automatically associates meaningful descriptions. Its limitations lie in the unsupervised nature of this method, which is more difficult to validate, and the use of a single dataset. It however brings the notion of cognitive labels, which is central to our last contribution. Our last contribution presents a method that learns functional atlases by combining several datasets. [Henson 2006] shows that forward inference, i.e. the probability of an activation given a cognitive process, is often not sufficient to conclude on the engagement of brain regions for a cognitive process. Conversely, [Poldrack 2006] describes reverse inference as the probability of a cognitive process given an activation, but warns of a logical fallacy in concluding on such inference from evoked activity. Avoiding this issue requires to perform reverse inference with a large coverage of the cognitive space. We present a framework that uses a "meta-design" to describe many different tasks with a common vocabulary, and use forward and reverse inference in conjunction to outline functional networks that are consistently represented across the studies. We use a predictive model for reverse inference, and perform prediction on unseen studies to guarantee that we do not learn studies' idiosyncrasies. This final contribution permits to learn functional atlases, i.e. functional networks associated with a cognitive concept. We explored different possibilities to jointly analyse multiple fMRI experiments. We have found that one of the main challenges is to be able to relate the experiments with one another. As a solution, we propose a common vocabulary to describe the tasks. [Henson 2006] advocates the use of forward and reverse inference in conjunction to associate cognitive functions to brain regions, which is only possible in the context of a large scale analysis to overcome the limitations of reverse inference. This framing of the problem therefore makes it possible to establish a large statistical model of the brain, and accumulate knowledge across functional neuroimaging studies

L'objectif de ce travail de thèse est l'exploration de l'imagerie par résonance magnétique (IRM) pour l'identification et la localisation des régions du cerveau impliquées dans l'épilepsie mésio-temporale. Précisément, les travaux visent 1) à optimiser un protocole d'IRM vasculaire sur un modèle animal d'épilepsie et 2) à concevoir une méthode de quantification de cartes IRM vasculaires basée sur la modélisation de la relation entre signaux IRM et paramètres biophysiques.Les acquisitions IRM sur un modèle expérimental murin d'épilepsie mésio-temporale avec sclérose de l'hippocampe ont été effectuées sur un scanner 9.4 T. Les données collectées ont permis de quantifier sept cartes IRM cellulaires et vasculaires quelques jours après l'état de mal épileptique puis plus tard, lorsque les crises spontanées sont apparues.Ces paramètres ont été employés pour l'identification automatique des régions épileptogènes et des régions de propagation des crises. Afin d'augmenter la détection de petites variations des paramètres IRM chez les individus épileptiques, une méthode de quantification basée sur la résonance magnétique fingerprinting est développée. Cette méthode consiste à identifier, parmi un ensemble de signaux simulés, le plus proche du signal IRM acquis et peut être vue comme un problème inverse qui présente les difficultés suivantes : le modèle direct est non-linéaire et provient d'une série d'équations sans expression analytique simple; les signaux en entrée sont de grandes dimensions; les vecteurs des paramètres en sortie sont multidimensionnels. Pour ces raisons, nous avons utilisé une méthode de régression inverse afin d'apprendre à partir de simulation la relation entre l'espace des paramètres et celui des signaux. Dans un domaine largement dominé par les approches d'apprentissage profond, la méthode proposée se révèle très compétitive fournissant des résultats plus précis. De plus, la méthode permet pour la première fois de produire un indice de confiance associé à chacune des estimations. En particulier, cet indice permet de réduire l'erreur de quantification en rejetant les estimations associées à une faible confiance.Actuellement, aucun protocole clinique permettant de localiser avec précision le foyer épileptique ne fait consensus. La possibilité d'une identification non-invasive de ces régions est donc un premier pas vers un potentiel transfert clinique
The objective of this thesis is the investigation of magnetic resonance imaging (MRI) for the identification and localization of brain regions involved in mesio-temporal lobe epilepsy (MTLE). Precisely, the work aims 1) at optimizing a vascular MRI protocol on an animal model of epilepsy and 2) at designing a method to quantify vascular MRI maps based on the modeling of the relationship between MRI signals and biophysical parameters.MRI acquisitions on an experimental mouse model of MTLE with hippocampal sclerosis were performed on a 9.4 T scanner. The data collected allowed the quantification of seven cellular and vascular MRI maps a few days after the epileptic condition and later when the spontaneous seizures emerged. These parameters were used for the automatic identification of epileptogenic regions and regions of seizure propagation. To enhance the detection of small variations in MRI parameters in epileptic subjects, a quantification method based on magnetic resonance fingerprinting has been developed. This method consists in identifying, among a set of simulated signals, the closest one to the acquired signal. It can be seen as an inverse problem that presents the following difficulties: the direct model is non-linear, as a complex series of equations or simulation tools; the inputs are high-dimensional signals; and the output is multidimensional. For these reasons, we used an appropriate inverse regression approach to learn a mapping between signal and biophysical parameter spaces. In a field widely dominated by deep learning approaches, the proposed method is very competitive and provides more accurate results. Moreover, the method allows for the first time to produce a confidence index associated with each estimate. In particular, this index allows to reduce the quantification error by discarding estimates associated with low confidence.So far no clinical protocol emerges as a consensus to accurately localize epileptic foci. The possibility of a non-invasive identification of these regions is therefore a first step towards a potential clinical transfer

A lesão do plexo braquial é considerada a alteração neural mais grave das extremidades. A principal causa é o trauma de alta energia, especialmente acidentes envolvendo veículos a motor. Por este motivo, as lesões traumáticas do plexo braquial são cada vez mais frequentes. O presente estudo avaliou a acurácia da ressonância magnética (RM) no diagnóstico das lesões traumáticas do plexo braquial no adulto, utilizando o achado intraoperatório como padrão-ouro. Também foi avaliada a acurácia da neurografia pesada em difusão (neurografia DW) em relação à RM convencional e a capacidade de diferenciação dos três tipos de lesão: avulsão, ruptura e lesão em continuidade. Trinta e três pacientes com história e diagnóstico clínico de lesão traumática do plexo braquial foram prospectivamente estudados por RM. Os achados obtidos pela RM sem e com o uso da neurografia DW, e os achados de exame clínico foram comparados com os achados intraoperatórios. A análise estatística foi feita com associação de significância de 5%. Observou-se alta correlação entre a RM com neurografia DW e a cirurgia (rs=0,79), e baixa correlação entre a RM convencional e a cirurgia (rs=0,41). A correlação interobservador foi maior para a RM com neurografia DW (rs = 0,94) do que para a RM sem neurografia DW (rs = 0,75). Os resultados de sensibilidade, acurácia e valor preditivo positivo foram acima de 95% para as RM com e sem neurografia DW no estudo de todo o plexo. As especificidades foram, em geral, maiores para a neurografia DW (p < 0,05). Em relação à diferenciação dos tipos de lesão, a RM com neurografia DW apresentou altas acurácias e sensibilidades no diagnóstico da avulsão/rotura, e alta especificidade no diagnóstico da lesão em continuidade. A acurácia da RM (93,9%) foi significativamente maior que a do exame clínico (76,5%) no diagnóstico das lesões de todo o plexo braquial (p < 0,05).
Brachial plexus injury is considered the most severe neural disorder in the extremities and in general resulting from high-energy trauma in young patients, usually involving motor vehicles. For this reason, traumatic brachial plexus injuries are becoming more frequent. This study evaluated the accuracy of magnetic resonance imaging (MRI) in the diagnosis of traumatic brachial plexus injuries in adults, using surgical findings as the gold standard method. We also evaluated the accuracy of diffusion weighted image neurography (DW neurography) compared to conventional MRI and the ability to differentiate the three types of injuries by MRI: avulsion, rupture and lesion-in-continuity. Thirty-three patients with clinical history and diagnosis of traumatic brachial plexus injury were prospectively studied by MRI. MRI findings (obtained with and without use of DW neurography) and clinical examination were compared with intraoperative findings. The statistical analysis was performed with 5% significance association. There was high correlation between MRI with DW neurography and surgery (rs = 0.79) and low correlation between conventional MRI and surgery (rs = 0.41). The interobserver correlation was higher for MRI with DW neurography (rs = 0.94) than for regular MRI (rs = 0.75). The sensitivities, accuracies and positive predictive values were above 95% for MRI (with and without DW neurography) in the evaluation of the entire plexus. The specificities were generally higher for DW neurography (p < 0.05). Regarding the differentiation between types of lesions, MRI with DW neurography demonstrated high accuracies and sensitivities in the diagnosis of avulsion / rupture and high specificity in the diagnosis of lesion-in-continuity. MRI accuracy (93.9%) was significantly higher than clinical examination (76.5%) in diagnosis of brachial plexus traumatic lesions (p < 0.05).

Em 2013, foi implantado no Brasil o Programa Nacional de Segurança do Paciente (PNSP), que através da Resolução RDC No 36, prevê que as instituições de saúde brasileiras devem apresentar um Plano de Segurança do Paciente (PSP) para as situações que possam proporcionar a ocorrência de eventos adversos (EAs), ou seja, lesão ou dano não intencional causado ao paciente pela intervenção assistencial e não pela doença base. O PSP é um documento com embasamento científico que aponta as situações de risco e descreve estratégias e ações definidas pelo serviço de saúde para a gestão do risco com objetivo de prevenção e redução dos EA em todas as fases da assistência. O implante de eletrodos para estimulação encefálica profunda (EEP) é um procedimento realizado rotineiramente no Hospital Marcelino Champagnat (HMC), localizado na cidade de Curitiba – PR, para melhorar a sintomatologia e a qualidade de vida de pacientes portadores de determinados distúrbios neurológicos crônicos. A confiabilidade e a precisão do posicionamento dos eletrodos cerebrais após a implantação de sistemas de EEP é de suma importância para a eficácia do método, sendo a Imagem por Ressonância Magnética (IRM) pós-operatória, atualmente, o padrão ouro para documentação do correto posicionamento dos eletrodos. Entretanto, a interação do sistema de EEP com o campo de radiofrequência do equipamento de IRM pode constituir uma fonte de EAs, uma vez que possibilita a indução de correntes elétricas com potencial de causar lesões térmicas ao paciente em locais de alta resistência. As recomendações de segurança dos fabricantes para a maioria dos sistemas de EEP implantados são bastante restritivas e resultam em longos tempos de exame ou em imagens de baixa qualidade, fatores que dificultam a prática em muitos centros. Estudos in vitro revelam que o aquecimento excessivo ocorre sob determinadas configurações, enquanto outras não oferecem risco fisiológico ao paciente. Estudos clínicos com base em vasta experiência sustentam a evidência de que é possível realizar exames pós-operatórios de forma satisfatória e sem a ocorrência de EAs utilizando parâmetros menos restritivos que aqueles determinados pelos fabricantes, desde que alguns padrões de segurança sejam cuidadosamente seguidos. Dessa forma, o presente trabalho propõe a elaboração de um PSP para a situação específica de pacientes com sistemas de EEP submetidos à exames de IRM no HMC, com base nas recomendações de segurança do fabricante e na revisão sistemática da literatura. De acordo com a base de dados consultada, um total de 26 artigos científicos foram considerados relevantes e permitiram identificar as possíveis fontes de risco de forma a evitá-las, colaborando com a conclusão satisfatória do PSP. Além de suprir a demanda local, o presente trabalho visou também promover a cultura de segurança do paciente e despertar a atenção para a necessidade de interposição de barreiras às diversas oportunidades de EAs que os setores de radiologia podem oferecer. A metodologia aqui proposta pode servir, ainda, de base para que outros centros de diagnóstico por imagem componham seus próprios PSPs.
In 2013, the National Program Patient Safety (PNSP) was implemented in Brazil through Resolution RDC 36, providing that the Brazilian health institutions must have a Patient Safety Plan (PSP) for situations that may lead to adverse events (AEs), which are unintentional injury or damage caused to the patient by the health care intervention and not by the primary disease. The PSP is a document with scientific basis that points to hazardous situations and describes strategies and actions defined by the health service for risk management in order to prevent and reduce AEs in all phases of patient care. Implantation of Deep Brain Stimulation (DBS) devices is considered a routine procedure at the Hospital Marcelino Champagnat (HMC), located in Curitiba – PR, and it consists in a practice widely used to improve symptoms and quality of life of patients with certain chronic neurological disorders.The reliability and accuracy of the final brain positioning of the leads, after the DBS implantation are of paramount importance to assure efficacy. Currently, post-operative Magnetic Resonance Imaging (MRI) is the gold standard for the documentation of the correct lead positioning. However, the interaction between the DBS system and the MRI radiofrequency field could represent an important source of adverse events (EAs) since it allows electric currents induction with potential to cause local thermal injuries on high resistance sites. The safety recommendations from the DBS system manufacturers for most of the already deployed systems are quite restrictive resulting in long examination times or low quality images, which compromises the practice in many centers. Thus, the present work proposes the development of a PSP based on the manufacturer's safety recommendations and a systematic review of the literature to the specific situation of patients with DBS systems undergoing MRI scans at the HMC. We found a total of 26 papers, that were considered relevant and allowed us to identify the potential sources of risk in order to avoid them, collaborating to the successful elaboration of the PSP. Besides supplying local demand, this work also aims to promote patient safety and draw attention to the need of interposing barriers in order to avoid significant AEs situations that a radiology department may be presented with. Moreover, the methodology proposed here can serve as a basis for other imaging centers to compose their own PSPs.

The number of people suffering from neurodegenerative diseases, such as Alzheimer`s disease, increased dramatically over the past centuries and is expected to increase even further within the next years. Based on predictions of the World Health Organization and Alzheimer`s Disease International, 115 million people will suffer from dementia by the year 2050. An additionally increase in other age related neurodegenerative diseases is also forecasted. Quite naturally, neurodegenerative diseases became a focus of attention of governments and health insurances, trying to control the uprising financial burden. Early detection and treatment of neurodegenerative diseases could be an important component in containing this problem. In particular, researchers focused on automatic methods to analyze patients’ imaging data. One way to detect structural changes in magnetic resonance images (MRI) is the voxel-based method approach. It was specifically implemented for various imaging modalities, e.g. T1-weighted images or diffusion tensor imaging (DTI). Voxel-based morphometry (VBM), a method specifically designed to analyze T1-weighted images, has become very popular over the last decade. Investigations using VBM revealed numerous structural brain changes related to, e.g. neurodegeneration, learning induced structural changes or aging. Although voxel-based methods are designed to be robust and reliable structural change detection methods, it is known that they can be influenced by physical and physiological factors. Dehydration, for example, can affect the volume of brain structures and possibly induce a confound in morphometric studies. Therefore, three-dimensional T1-weighted images were acquired of six young and healthy subjects during different states of hydration. Measurements during normal hydration, hyperhydration, and dehydration made it possible to assess consequential volume changes in gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). The datasets were analyzed using VBM, FreeSurfer and SIENA. A significant decrease of GM and WM volume, associated with dehydration, was found in various brain regions. The most prominent effects were found in temporal and parietal areas, in the left inferior orbito-frontal region, and in the extra-nuclear region. Moreover, we found consistent increases in CSF, an expansion around 6% of the ventricular system affecting both lateral ventricles, i.e. the third and fourth ventricle. Similar degrees of shrinkage in WM volume and increase of the ventricular system have been reported in studies of Alzheimer’s disease during disease progression and in its prestage mild cognitive impairment. Based on these findings, a potential confound in GM and WM or CSF studies due to the subjects’ hydration state cannot be excluded and should be appropriately addressed. These results underline the sensitivity of VBM and might also concern other voxel-based methods, such as Tract-Based Spatial Statistics (TBSS). TBSS was specifically designed for WM analyses and its sensitivity might be helpful for revealing the spatial relation of structural WM changes and related blood serum biomarkers. Two common brain related biomarkers are the glial protein S100B, a plasticity inducing neuro- and gliotrophin, and neuron-specific enolase (NSE), a marker for neuronal damage. However, the spatial specificity of these biomarkers for brain region has not been investigated in vivo until now. Therefore, we acquired two MRI parameters – T1- weighted and DTI - sensitive to changes in GM and WM, and obtained serum S100B and NSE levels of 41 healthy subjects. Additionally, the gene expression of S100B on the whole brain level in a male cohort of three subjects from the Allen Brain Database was analyzed. Furthermore, a female post mortal brain was investigated using double immunofluorescence labeling with oligodendrocyte markers. It could be shown that S100B is specifically related to white matter structures, namely the corpus callosum, anterior forceps and superior longitudinal fasciculus in female subjects. This effect was observed in fractional anisotropy and radial diffusivity – the latest an indicator of myelin changes. Histological data confirmed a co-localization of S100B with oligodendrocyte markers in the human corpus callosum. S100B was most abundantly expressed in the corpus callosum according to the whole genome Allen Human Brain Atlas. In addition, NSE was related to gray matter structures, namely the amygdala. This effect was detected across sexes. The data demonstrates a very high S100B expression in white matter tracts, in particular in human corpus callosum. This was the first in vivo study validating the specificity of the glial marker S100B for the human brain, and supporting the assumption that radial diffusivity represents a myelin marker. The results open a new perspective for future studies investigating major neuropsychiatric disorders. All above mentioned studies are mainly dependent on the sensitivity and accuracy of soft and hardware parameters. In particular, technical developments have improved acquisition accuracy in the field of MRI. Interestingly, very little is known about the confounding effects of variations due to hardware parameters and their possible impact on reliability and sensitivity of VBM. Recent studies have shown that different acquisition parameters may influence VBM results. Therefore age-related GM changes were investigated with VBM in 36 healthy volunteers grouped into 12 young, 12 middle-aged and 12 elderly subject. Six T1-weighted datasets were acquired per subject with a 12-channel matrix coil, as well as a 32-channel array, MP-RAGE and MP2RAGE, and with isotropic resolutions of 0.8 and 1 mm. DARTEL-VBM was applied on all images and GM, WM and CSF segments were statistically analyzed.. Paired t-tests and statistical interaction tests revealed significant effects of acquisition parameters on the estimated gray-matter-density (GMD) in various cortical and subcortical brain regions. MP2RAGE seemed slightly less prone to false positive results when comparing data acquired with different RF coils and yielded superior segmentation of deep GM structures. With the 12-channel coil, MP-RAGE was superior in detecting age-related changes, especially in cortical structures. Most differences between both sequences became insignificant with the 32-channel coil, indicating that the MP2RAGE images benefited more from the improved signal-to-noise ratio and improved parallel-imaging reconstruction). A possible explanation might be an overestimation of the GM compartment on the MP-RAGE images. In view of substantial effects obtained for all parameters, careful standardization of the acquisition protocol is advocated. While the current investigation focused on aging effects, similar results are expected for other VBM studies, like on plasticity or neurodegenerative diseases. This work has shown that voxel-based methods are sensitive to subtle structural brain changes, independent of imaging modality and scanning parameters. In particular, the studies investigated and discussed the analysis of T1- and diffusion weighted images with VBM and TBSS in the context of dehydration, blood serum sensitive biomarkers and aging were discussed. The major goal of these studies was the investigation of the sensitivity of voxel-based methods. In conclusion, sensitivity and accuracy of voxelbased methods is already high, but it can be increased significantly, using optimal hardand software parameters. It is of note, though, that these optimizations and the concomitant increase of detection sensitivity could also introduce additional confounding factors in the imaging data and interfere with the latter preprocessing and statistical computations. To avoid an interference e.g. originating from physiological parameters, a very careful selection and monitoring of biological parameters of each volunteer throughout the whole study is recommended. A potential impact of scanning parameters can be minimized by strict adherence to the imaging protocol for each study subjectwithin a study. A general increase in detection sensitivity due to optimized parameters selection in hard- and/or can not be concluded by the above mentioned studies. Although the present work addressed some of those issues, the topic of optimal selection of parameters for morphometric studies is still very complex and controversial and has to be individually decided. Further investigations are needed to define more general scanning and preprocessing standards to increase detection sensitivity without the concomitant amplification of confounding factors.

Shi, Lin.
Thesis (Ph.D.)--Chinese University of Hong Kong, 2008.
Includes bibliographical references (leaves 171-185).
Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.
Abstracts in English and Chinese.

雖然由老化引發的腦白質病變是老年人認知障礙的一個重要誘因，其機理缺並不為人所知。最新的小樣本研究表明擴散核磁造影在很大程度上是對腦白質病變最為敏感的的成像檢測手段。加深對擴散核磁造影所給出的各種指數的理解和認知對於檢測腦白質病變的病理發展以及研發試驗療法的替代標記有重要的意義。
為了獲得更具有臨床價值的擴散核磁造影指數，我們首先需要重構腦白質纖維束並沿著重構出的腦白質纖維束採集數值。然而，傳統的腦白質纖維束重構技術對於腦白質病變十分敏感。此外，不同病人所重構出的腦白質纖維束間缺乏映射關係也使我們無法有效進行大樣本統計分析。
在這個課題裡，我們提出了一個可以解決以上問題的一個全新框架。我們將專家標註過功能區的全腦白質纖維束模板配準到各個個體的腦部。此方案可自動生成個體化的全腦白質纖維束以及纖維束的功能區標註。自由形變模型被用於在全局層面對配準進行約束。所重構纖維束的曲率被用於在局部對配準進行約束。為了減輕腦白質病變對配準的影響，我們運用了一種 魯棒的主成分分析手段來檢測被病灶所干擾的纖維束。為了指導這些被干擾纖維束的配準，我們提出了一種全新的沿纖維束的區域特徵作為替代。此外，我們也探究了通過在纖維束上建立坐標系來除去離群纖維已經提供更高相關性的辦法。
我們所提出的框架被運用於一個腦小血管病變的臨床研究。在64個研究對象中約半數是腦白質病變患者。試驗結果證實此算法成功地將全腦白質纖維束模板配準到了所有研究對象上。沿著特定纖維束改採集的指數與認知測試分數的相關性顯著地超越了傳統指數所給出的結果。我們同時也發現沿著不同功能區腦白質纖維改採集的指數與相應的認知測試分數統計相關。
Although age-related white matter lesion(WML)is an important substrate for cognitive impairment in the elderly, the mechanisms whereby WML induces cognitive impairment are uncertain. Recent findings based on small studies suggested that diffusion tensor imaging (DTI) measures might be the most sensitive imaging predictors in patients with WML. Understanding the imaging predictors for such disease will be useful in monitoring disease progression and in devising surrogate marker for treatment trials.
In order to obtain DTI measurements with diagnostic significance, it is first necessary to reconstruct the white-matter fiber pathways inside the brain along which certain DTI-derived values are calculated. Nevertheless, the traditional approach of white-matter tract reconstruction, or tractography, is severely hindered by the abundant existence of lesions inside the brains of WML patients. The lack of correspondence between fiber bundles across patients also makes obtaining group statistics of individual fiber bundles dicult.
In this study, we propose a novel framework that can mitigate the aforementioned issues of traditional tractography approaches. An expert-labeled whole brain tractography template is registered onto individual patients. Fiber trajectories and anatomically meaningful fiber bundles are automatically obtained by this registration. The free-form deformations are used to regularize the transformations at the whole brain level and across fiber bundles. Fiber curvatures are penalized as the intra-fiber regularization to encourage the smoothness of transformed fibers. White matter (WM) lesion is one of the major factors affecting tractography and registration of subjects with neuro Logical disorders. The Robust Principal Component Analysis(RPCA) is used to automatically detect fiber tract segments that are affected by WM lesion and a novel along-fiber regional prior is learned from healthy subjects to facilitate the registration of these fiber tract segments. We also propose to establish bundle-wise coordinate system by utilizing low-rank constraints upon the DTI measurements. The eort elevates the summary for an anatomical bundle from a scalar statistic to a vector containing changes along the representative fiber pathway. It provides means to exclude the outlier fibers while retaining partially-complete fibers.
The proposed scheme is applied to a clinical study of cerebral small vessel diseases(SVD).Experimental results show successful registration of the whole brain tractography template onto all 64 subjects, including both healthy con¬trol subjects and SVD patients. The DTI measures measured along specific registered anatomical fiber bundles exhibit significant boost in correlation with cognitive functions as compared with traditional measures. It also shows that different anatomical WM structures correlate with multiple types of cognitive functions in different ways.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
He, Xiaotian.
Thesis (M.Phil.)--Chinese University of Hong Kong, 2012.
Includes bibliographical references (leaves 46-53).
Abstracts also in Chinese.
List of Figures --- p.ix
List of Tables --- p.xii
Chapter 1 --- Introduction --- p.1
Chapter 1.1 --- Motivation --- p.1
Chapter 1.2 --- Our Work and Contributions --- p.2
Chapter 1.3 --- Related Work --- p.4
Chapter 1.4 --- Thesis Organization --- p.5
Chapter 2 --- Background --- p.6
Chapter 2.1 --- Background of Neuroanatomy --- p.6
Chapter 2.2 --- Background on Diffusion Tensor Magnetic Resonance Imaging (DTMRI) --- p.11
Chapter 3 --- Free Form Fibers --- p.18
Chapter 3.1 --- DTI Acquisition --- p.20
Chapter 3.2 --- Fiber Model --- p.20
Chapter 3.3 --- Fiber-to-DTI Registration --- p.21
Chapter 3.3.1 --- Free-Form Fibers (FFFs) --- p.21
Chapter 3.3.2 --- Tensor-Driven Fiber-to-DTI Registration --- p.23
Chapter 3.3.3 --- Reliability Assessment by Robust Principal Component Analysis --- p.24
Chapter 3.3.4 --- Contextual Feature --- p.26
Chapter 3.3.5 --- Learning the Fiber Context Prior --- p.29
Chapter 3.3.6 --- Registration Refinement Using the Fiber Context Prior --- p.29
Chapter 4 --- Results --- p.31
Chapter 5 --- Future Work --- p.39
Chapter 5.1 --- Refinement on Large Bundles --- p.39
Chapter 5.2 --- Outlier Fiber Removal in Fiber Template --- p.40
Chapter 6 --- Conclusion --- p.44
Bibliography --- p.46

碩士
國立中興大學
通訊工程研究所
100
Support vector machine (SVM) has found great promise in magnetic resonance analysis recently. It works effectively depend on choosing proper parameter and kernel. However, there are two major challenge issues in SVM classification. One is the random select of training sample cause unstable and inconsistent result. It may lead to large variation from every experiment. The other is that classification errors via SVM usually distributed at the boundary between classes which is caused by partial volume effect. In this thesis, we have developed a new version of SVM, called iterative SVM (ISVM) that we could take only a few training samples to solve first issue. And using morphological or edge detection processing to deal with second issue, the idea is through above methods to find the boundary between classes. Then extract these areas from MR images for SVM to classify again. We combined above approaches with a preprocessing independent component analysis to improve SVM classification. And choose weighted RBF kernel instead of RBF kernel in SVM. The experimental results show the proposed method has great performance in magnetic resonance brain image classification.

Functional magnetic resonance imaging (fMRI) enables non-invasive investigation of the neural functions with excellent spatial resolution. Although fMRI has primarily been developed in young adult populations, its use is becoming widespread in paediatrics. However, there are many differences, both behavioural and physical, between adults and children requiring adjustments to imaging and analysis methodology to optimise the results in children. This thesis examines fMRI analysis methodology to improve the detection of developmental changes in the brain. The work uses an emotional and familiar face paradigm that elicits strong BOLD fMRI responses in the fusiform, a region that is still developing across childhood. This face paradigm also enables the comparison of the fusiform responses to the primary visual cortex to link to extensive results in the literature. Thirty five 4-8 year old children and fourteen adults (18-30 years old) were scanned. To address the concern of anatomical size differences between the brains of adults and children, the anatomical variability of the fusiform was measured and the validity of stereotaxic transformation into an adult template was confirmed for the children. To investigate the effect of threshold settings between the adults and children, individual subject analyses of the peak activation location, estimated signal percent change and noise values were calculated using the general linear model (GLM). Similar functional peak locations between individuals were quantitatively selected using a novel application of the activation likelihood estimation (ALE). Also, several different preprocessing steps were evaluated for their ability to correct for the increased motion frequently seen in children, in a quantitative framework (NPAIRS) using canonical variates analysis (CVA), a data driven multivariate model as well as the standard univariate GLM. Functional differences between the adults and the children were identified in the fusiform by applying these optimised procedures. The results of this thesis demonstrate that thresholding and preprocessing pipelines must be made in a group-specific fashion. These methods can also be extended to elderly populations, enabling the investigation of the complete ageing spectrum with fMRI.

Background: Functional magnetic resonance imaging (fMRI) can require extensive preprocessing to minimize noise and maximize signal. There is evidence suggesting that fixed-subject preprocessing pipelines, the current standard in fMRI preprocessing, are suboptimal compared to individual-subject pipelines. Aim: We sought to test if individual-subject preprocessing pipeline optimization, compared to fixed, resulted in stronger and more reliable brain-patterns in episodic recognition. Methodology: 27 young healthy controls were scanned via fMRI while performing forced-choice episodic recognition. Several sets of fMRI preprocessing pipelines were tested and optimized in a fixed and individual-subject manner, using methods outlined by Churchill et al. (2011). Results: Individual-subject pipeline optimization, compared to fixed, significantly increased reproducibility, significantly increased the detection of positively and negatively activated voxels, and resulted in a brain-pattern with significant correlation to a task behavioral measure. Conclusions: Individual-subject pipeline optimization, compared to fixed, led to stronger and more reliable brain-patterns that are significantly correlated with behavior.

An ever-increasing number of functional magnetic resonance imaging (fMRI) studies are now using information-based multi-voxel pattern analysis (MVPA) techniques to decode mental states. In doing so, they achieve a significantly greater sensitivity compared to when they use univariate analysis frameworks. Two most prominent MVPA methods for information mapping are searchlight decoding and classifier weight mapping. The new MVPA brain mapping methods, however, have also posed new challenges for analysis and statistical inference on the group level. In this thesis, I discuss why the usual procedure of performing t-tests on MVPA derived information maps across subjects in order to produce a group statistic is inappropriate. I propose a fully nonparametric solution to this problem, which achieves higher sensitivity than the most commonly used t-based procedure. The proposed method is based on resampling methods and preserves the spatial dependencies in the MVPA-derived information maps. This enables to incorporate a cluster size control for the multiple testing problem. Using a volumetric searchlight decoding procedure and classifier weight maps, I demonstrate the validity and sensitivity of the new approach using both simulated and real fMRI data sets. In comparison to the standard t-test procedure implemented in SPM8, the new results showed a higher sensitivity and spatial specificity. The second goal of this thesis is the comparison of the two widely used information mapping approaches -- the searchlight technique and classifier weight mapping. Both methods take into account the spatially distributed patterns of activation in order to predict stimulus conditions, however the searchlight method solely operates on the local scale. The searchlight decoding technique has furthermore been found to be prone to spatial inaccuracies. For instance, the spatial extent of informative areas is generally exaggerated, and their spatial configuration is distorted. In this thesis, I compare searchlight decoding with linear classifier weight mapping, both using the formerly proposed non-parametric statistical framework using a simulation and ultra-high-field 7T experimental data. It was found that the searchlight method led to spatial inaccuracies that are especially noticeable in high-resolution fMRI data. In contrast, the weight mapping method was more spatially precise, revealing both informative anatomical structures as well as the direction by which voxels contribute to the classification. By maximizing the spatial accuracy of ultra-high-field fMRI results, such global multivariate methods provide a substantial improvement for characterizing structure-function relationships.

Simultaneous electroencephalogram (EEG) and functional magnetic resonance imaging (fMRI) have gained increasing popularity in studying human cognition due to their potential to map the brain dynamics with high spatial and temporal fidelity. Such detailed mapping of the brain is crucial for understanding the neural mechanisms by which humans make perceptual decisions. Despite recent advances in data acquisition and analysis of simultaneous EEG-fMRI, the lack of effective computational tools for optimal fusion of the two modalities remains a major challenge. The goal of this dissertation is to provide a recipe of machine learning methods for fusion of simultaneous EEG-fMRI data. Specifically, we investigate three types of fusion approaches and apply them to study the whole-brain spatiotemporal dynamics during a rapid object recognition task where subjects discriminate face, car, and house images under ambiguity. We first use an asymmetric fusion approach capitalizing on temporal single-trial EEG variability to identify early and late neural subsystems selective to categorical choice of faces versus nonfaces. We find that the degree of interaction in these networks accounts for a substantial fraction of our bias to see faces. Based on a computational modeling of behavioral measures, we further dissociate separate neural correlates of the face decision bias modulated by varying levels of stimulus evidence. Secondly, we develop a state-space model based symmetric fusion approach to integrate EEG and fMRI in a probabilistic generative framework. We use a variational Bayesian method to infer the network connectivity among latent neural states shared by EEG and fMRI. Finally, we use a data-driven symmetric fusion approach to compare representations of the EEG and fMRI against those of a deep convolutional neural network (CNN) in a common similarity space. We show a spatiotemporal hierarchical correspondence in visual processing stages between the human brain and the CNN. Collectively, our results show that the spatiotemporal properties of neural circuits revealed by the analysis of simultaneous EEG-fMRI data can reflect the choice behavior of subjects during rapid perceptual decision making.

Early brain development is characterized by rapid growth and perpetual reconfiguration, driven by a dynamic milieu of heterogeneous processes. Moreover, potent postnatal brain plasticity engenders increased vulnerability to environmental stimuli. However, little is known regarding the ontogeny and temporal manifestations of inter- and intra-regional functional connectivity that comprise functional brain networks. Recently, resting-state functional magnetic resonance imaging (fMRI) emerged as a promising non-invasive neuroinvestigative tool, measuring spontaneous fluctuations in blood oxygen level dependent (BOLD) signal at rest that reflect baseline neuronal activity. Its application has expanded to infant populations in the past decade, providing unprecedented insight into functional organization of the developing brain, as well as early biomarkers of abnormal/ disease states. However, rapid extension of the resting-state technique to infant populations leaves many methodological issues need to be resolved prior to standardization of the technique. The purpose of this thesis is to describe a protocol for intrinsic functional connectivity analysis, and extraction of resting-state networks in infants <12 months of age using the data-driven approach independent component analysis (ICA). To begin, we review the evolution of resting-state fMRI application in infant populations, including the biological premise for neural networks. Next, we present a protocol designed such that investigators without previous knowledge in the field can implement the analysis and reliably obtain viable results consistent with previous literature. Presented protocol provides detailed, albeit basic framework for RSN analysis, with interwoven discussion of basic theory behind each technique, as well as the rationale behind selecting parameters. The overarching goal is to catalyze efforts towards development of robust, infant-specific acquisition and preprocessing pipelines, as well as promote greater transparency by researchers regarding methods used. Finally, we review the literature, current methodological challenges and potential future directions for the field of infant resting-state fMRI.

abstract: Electromagnetic fields (EMFs) generated by biologically active neural tissue are critical in the diagnosis and treatment of neurological diseases. Biological EMFs are characterized by electromagnetic properties such as electrical conductivity, permittivity and magnetic susceptibility. The electrical conductivity of active tissue has been shown to serve as a biomarker for the direct detection of neural activity, and the diagnosis, staging and prognosis of disease states such as cancer. Magnetic resonance electrical impedance tomography (MREIT) was developed to map the cross-sectional conductivity distribution of electrically conductive objects using externally applied electrical currents. Simulation and in vitro studies of invertebrate neural tissue complexes demonstrated the correlation of membrane conductivity variations with neural activation levels using the MREIT technique, therefore laying the foundation for functional MREIT (fMREIT) to detect neural activity, and future in vivo fMREIT studies. The development of fMREIT for the direct detection of neural activity using conductivity contrast in in vivo settings has been the focus of the research work presented here. An in vivo animal model was developed to detect neural activity initiated changes in neuronal membrane conductivities under external electrical current stimulation. Neural activity was induced in somatosensory areas I (SAI) and II (SAII) by applying electrical currents between the second and fourth digits of the rodent forepaw. The in vivo animal model involved the use of forepaw stimulation to evoke somatosensory neural activations along with hippocampal fMREIT imaging currents contemporaneously applied under magnetic field strengths of 7 Tesla. Three distinct types of fMREIT current waveforms were applied as imaging currents under two inhalants – air and carbogen. Active regions in the somatosensory cortex showed significant apparent conductivity changes as variations in fMREIT phase (φ_d and ∇^2 φ_d) signals represented by fMREIT activation maps (F-tests, p <0.05). Consistent changes in the standard deviation of φ_d and ∇^2 φ_d in cortical voxels contralateral to forepaw stimulation were observed across imaging sessions. These preliminary findings show that fMREIT may have the potential to detect conductivity changes correlated with neural activity.
Dissertation/Thesis
Doctoral Dissertation Biomedical Engineering 2020

Η εισαγωγή της Μαγνητικής Τομογραφίας (ΜΤ) στην κλινική πρακτική και η συμπληρωματική πληροφορία που δίνει η Φασματοσκοπία Μαγνητικού Συντονισμού (ΦΜΣ) συνιστά μια από τις πιο σημαντικές εξελίξεις στη διάγνωση ασθενών με καρκίνο εγκεφάλου [1]. Παρ’ όλα αυτά, οι εικόνες ΜΤ είναι συχνά δύσκολο να ερμηνευθούν από τους ειδικούς λόγω [2] α/ της υποκειμενικότητας και περιορισμένης εμπειρίας του παρατηρητή στην εκτίμηση εικόνων που παράγει η σχετικά νέα αυτή τεχνολογία, β/ των ποικίλων κλινικών χαρακτηριστικών των όγκων (π.χ. τύπος, διαβάθμιση κακοήθειας κλπ.) και γ/ της ιδιαιτερότητας των όγκων στην αντίθεση που παρουσιάζουν με τον περιβάλλοντα ιστό. Μόνο λιγοστές μελέτες έχουν διεξαχθεί για να χαρακτηρίσουν ιστούς εγκεφάλου μέσω της ανάλυσης ποσοτικών χαρακτηριστικών από εικόνες εγκεφάλου ΜΤ [3, 4]. Ενώ έχει ήδη τονιστεί η αναγκαιότητα συσχετισμού της διαγνωστικής και προγνωστικής πληροφορίας που προέρχεται από εικόνες ΜΤ και σήματα ΦΜΣ στη διεθνή βιβλιογραφία [5], υπάρχουν λιγοστές ανάλογες αναφορές για τον σχεδιασμό και υλοποίηση συστήματος Η/Υ αυτόματης διάγνωσης όγκων εγκεφάλου κάνοντας συνδυασμό ποσοτικής πληροφορίας προερχόμενης από εικόνες ΜΤ και σήματα ΦΜΣ [6, 7]. Οι στόχοι της παρούσας διατριβής εστιάζονται στα παρακάτω: - στη μελέτη, ανάπτυξη και η υλοποίηση υπολογιστικού συστήματος αυτόματης ταξινόμησης όγκων του εγκεφάλου μέσω της ποσοτικής ανάλυσης εικόνων ΜΤ το οποίο θα βελτιώνει την ακρίβεια ταξινόμησης σε σχέση με ήδη υπάρχοντα συστήματα [4, 8, 9], όπως αυτά περιγράφονται στην διεθνή βιβλιογραφία μεταξύ πρωτογενών και δευτερογενών όγκων εγκεφάλου καθώς και μεταξύ γλοιωμάτων και μηνιγγιωμάτων με την χρήση δέντρου ιεραρχικής απόφασης δύο επιπέδων. Επιπλέον, στην ανάδειξη πως η χρήση ενός μη-γραμμικού πολυωνυμικού μετασχηματισμού ελάχιστων τετραγώνων των χαρακτηριστικών υφής έχει ως αποτέλεσμα την βελτίωση της ακρίβειας ταξινόμησης του ταξινομητή πιθανοκρατικού νευρωνικού δικτύου. - στην επέκταση και την βελτίωση του συστήματος αυτόματης ταξινόμησης όγκων του εγκεφάλου χρησιμοποιώντας α/ ογκομετρικές ποσοτικές παραμέτρους εικόνων ΜΤ, β/ ταξινομητή μηχανών διανυσμάτων στήριξης μαζί με τη μεθοδολογία συνάθροισης αποτελεσμάτων ταξινόμησης από τυχαιοποιημένα δείγματα κατηγοριών δημιουργημένων με επαναδειγματοληψία για κάθε κόμβο δέντρου ιεραρχικής απόφασης δύο επιπέδων όπου στο πρώτο επίπεδο πραγματοποιήθηκε διαχωρισμός μεταξύ πρωτογενών και δευτερογενών όγκων εγκεφάλου και στο δεύτερο και μεταξύ γλοιωμάτων και μηνιγγιωμάτων και γ/ έναν τροποποιημένο πυρήνα ακτινικής συνάρτησης βάσης για τον ταξινομητή μηχανών διανυσμάτων στήριξης ο οποίος περιλαμβάνει την τεχνική μη-γραμμικού πολυωνυμικού μετασχηματισμού ελάχιστων τετραγώνων με στόχο την βελτίωση της ακρίβειας ταξινόμησης. - στην περαιτέρω επέκταση και την βελτίωση του συστήματος αυτόματης ταξινόμησης με την εισαγωγή χαρακτηριστικών προερχόμενων από σήματα ΦΜΣ ώστε να διερευνηθεί εάν η χρήση του μπορεί να βελτιώσει τα αποτελέσματα ταξινόμησης μεταξύ μηνιγγιωμάτων και μονήρων μεταστάσεων. Τέλος κάνοντας μια περίληψη, η παρούσα διατριβή διαπραγματεύεται τον σχεδιασμό, ανάπτυξη και υλοποίηση μεθόδων και αλγορίθμων για την επεξεργασία και ανάλυση ιατρικών εικόνων, επικεντρώνοντας ειδικότερα στην εφαρμογή των μεθόδων αυτών για την διάγνωση του τύπου των όγκων εγκεφάλου. Τα πιο βασικά συμπεράσματα που απορρέουν από την παρούσα διατριβή είναι τα ακόλουθα: α/ Το σύστημα ταξινόμησης των τύπων των όγκων εγκεφάλου που σχεδιάστηκε και υλοποιήθηκε αυξάνει τα ποσοστά ορθής ταξινόμησης σε σχέση με τα υπάρχοντα. β/ Η κωδικοποίηση των ιδιοτήτων της υφής που προέρχεται από τον σύνολο του όγκου παρέχει επιπρόσθετη πληροφορία στο σύστημα ταξινόμησης αυξάνοντας τα ποσοστά επιτυχούς διαχωρισμού. γ/ Τα χαρακτηριστικά φασματοσκοπίας μαγνητικού συντονισμού αποτελούν επιπρόσθετη αξία στο χαρακτηρισμό του τύπου των όγκων εγκεφάλου μιας και οδήγησαν στην αύξηση του ποσοστού επιτυχούς διαχωρισμού του συστήματος ταξινόμησης.
The process of brain tumor characterization requires a rather intricate assessment of the various Magnetic Resonance (MR) image and spectroscopic features and is typically performed by experienced radiologists. Despite the inherently subjective nature of many of the decisions associated with this process, an expert radiologist is able to perform this task with a significant degree of precision and accuracy. However, in the effort to deliver more effective treatment, clinicians are continuously seeking for greater accuracy in the pathological characterization of brain tissues. The aim of the present thesis was to design, implement, and evaluate a software classification system for discriminating between different brain tumor types on Magnetic Resonance Imaging (MRI), employing textural and spectroscopic features. The clinical material consisted of sixty seven T1-weighted post-contrast MR brain images (21 metastases, 19 meningiomas, and 27 gliomas), obtained from patients with verified and untreated intracranial tumors. Thirty-six 2-dimensional textural features (2D), from the image histogram and the co-occurrence and run-length matrices, were extracted from each one of 67 MR-images. Similarly, an equal number of 3-dimensional textural features (3D) were also calculated in the attempt to maximize classification performances. Finally, MR-spectroscopy features were also incorporated for improving classification accuracies. Classification methods employed included i/ a modified Probabilistic Neural Network (PNN) and Support Vector Machines (SVM) algorithms, incorporating a non-linear Least Squares Features Transformation (LSFT) into the classifiers and ii/ an ensemble classification scheme employing the LSFT-SVM classifier. The LSFT improved classifiers’ performances, increased class separability, and resulted in dimensionality reduction. For evaluating the performance of the designed classification schemes, evaluations were performed by means of the external cross validation process, which is considered indicative of the generalization performance of the designed classification system to ‘unseen’ cases. It was found that the LSFT features transformation enhanced the performance of the PNN and SVM algorithms, achieving classification accuracies of 73.48 % in distinguishing metastatic from primary tumors and 88.67% in discriminating gliomas from meningiomas. When volumetric 3-dimensional features were employed, these results improved to 88.18% for discriminating between metastatic and primary tumors and 97.33% for distinguishing gliomas from meningiomas. The textural features employed in the design of the optimum classification scheme were associated primarily with image texture homogeneity. Finally, when MR-spectroscopy features were also incorporated, classification accuracy was boosted up from 95% in discriminating meningiomas from metastasis to 100%. The MR-image features that participated in the optimum feature vector were related to the degree of homogeneity, the amount of randomness and the dispersion of the gray-tone intensity values within the texture of the tumor. These textural characteristics are related to textural parameters that physicians employ in diagnosis and they were proportional to the textural imprint of brain tumors, i.e. gliomas have heterogeneous texture while meningiomas appear to be homogeneous in MR imaging. The MR-spectroscopy feature that participated in the optimum feature vector was the Choline (Cho) / N-Acetyl Aspartate (NAA) metabolite integral ratio. It was found that both meningiomas and metastases are characterized from low concentrations of NAA while meningiomas exhibit higher concentrations of Cho than metastases, which could be attributed to increased synthesis of tumor cell membranes. Finally, the proposed system might be of value as an assisting tool for brain tumor characterization on volumetric MRI series.

Tese de mestrado integrado em Engenharia Biomédica e Biofísica (Radiações em Diagnóstico e Terapia), apresentada à Universidade de Lisboa, através da Faculdade de Ciências, 2012
A neuroanatomia humana tem sido objecto de estudo científico desde que surgiu o interesse na organização do corpo humano e nas suas funções, quer como um todo quer através das partes que o constituem. Para atingir este fim, as autópsias foram a primeira forma de revelar algum conhecimento, o qual tem vindo a ser catalogado e sistematizado à medida que a medicina evolui. Passando por novas técnicas de conservação e tratamento de tecido humano, de que são exemplo as dissecções de Klinger, nas quais se fazem secções de material conservado criogenicamente, bem como por estudos histológicos através da utilização de corantes, conseguiu-se uma forma complementar de realizar estes estudos. Permanecia, no entanto, a impossibilidade de analisar in vivo a estrutura e função dos diferentes sistemas que constitutem o Homem. Com o surgimento das técnicas imagiológicas o diagnóstico e monitorização do corpo humano, bem como das patologias a ele associadas, melhoraram consideravelmente. Mais recentemente, com o aparecimento da ressonância magnética (MRI: do Inglês "Magnetic Resonance Imaging"), tornou-se possível estudar as propriedades magnéticas do tecido, reflectindo as suas características intrínsecas com base na aplicação de impulsos de radiofrequência. Através de ressonância magnética é possível estudar essas propriedades em vários núcleos atómicos, sendo mais comum o estudo do hidrogénio, pois somos maioritariamente consistituídos por água e gordura. Uma vez que só é possível medir variações do campo magnético, aplicam-se impulsos de radiofrequência para perturbar o equilíbrio dos spins e medir os seus mecanismos de relaxação, os quais, indirectamente, reflectem a estrutura do tecido. Contudo, o sinal medido é desprovido de qualquer informação espacial. De facto, para podermos proceder a essa quantificação, é necessária a utilização de gradientes de campo magnético, que permitem modificar localmente a frequência de precessão dos protões, através da alteração local do campo magnético, permitindo assim, adquirir o sinal de forma sequencial. A informação obtida constitui uma função variável no espaço e através da transformação de Fourier pode ser quantificada em frequências espaciais, sendo estes dados armazenados no espaço k. O preencimento deste espaço, caracterizado por frequências espaciais, bem como os gradientes de campo magnético que são aplicados, permitem determinar a resolução da imagem que podemos obter, aplicando uma transformação de Fourier inversa. O estudo da ressonância magnética não se restringe à análise da estrutura mas também ao estudo da função e difusão das moléculas de água. A difusão é um processo aleatório, que se traduz pelo movimento térmico das moléculas de água, e o seu estudo permite inferir sobre o estado do tecido e microestrutura associada, de uma forma não invasiva e in vivo. A técnica de imagiologia de ressonância magnética ponderada por difusão (DWI: do Inglês "Diffusion Weighted Imaging") permite o estudo da direccionalidade das moléculas de água e extracção de índices que reflectem directamente a integridade dos tecidos biológicos. De modo a sensibilizar as moléculas de água à difusão, é necessário aplicar sequências de ressonância magnética modificadas, nas quais se aplicam gradientes de campo magnético de difusão para quantificar o deslocamento das moléculas e a sua relação com o coeficiente de difusão das mesmas. Num ambiente livre e sem barreiras a difusão das moléculas de água é isotrópica, uma vez que se apresenta igual em todas as direcções. Todavia, tal não se verifica no corpo humano. A presença destas barreiras leva a que, na verdade, apenas possa ser medido um coeficiente de difusão aparente. Este, por sua vez, traduz a interacção entre as moléculas de água com a microestrutura e, como tal, uma anisotropia na sua difusão. Como caso particular de difusão anisotrópica a nível cerebral, tem-se a difusão das moléculas de água na matéria branca, uma vez que esta apresenta uma direccionalidade preferencial de acordo com a orientação dos axónios, visto estarem presentes menos restrições à sua propagação, ao contrário do que acontece com a direcção perpendicular (devido à membrana celular e às bainhas de mielina). Por oposição, a matéria cinzenta, constituída pelo aglomerado dos corpos celulares dos neurónios, e o líquido cefalorraquidiano apresentam uma difusão sem direcção preferencial (i.e. aproximadamente isotrópica). A informação obtida através da difusão das moléculas de água encontra-se limitada pelo número de direcções segundo o qual aplicamos os gradientes de difusão. Deste modo, surgiu a imagiologia por tensor de difusão (DTI: do Inglês "Diffusion Tensor Imaging"). Esta técnica permite extrair informação acerca da tridimensionalidade da distribuição da difusão de moléculas de água através da aplicação de seis gradientes de difusão não colineares entre si. A distribuição destas moléculas pode, então, ser vista como um elipsóide, no qual o principal vector próprio do tensor representa a contribuição da difusão das moléculas segundo a direcção do axónio (ou paralela), sendo os dois restantes componentes responsáveis pela contribuição transversal. Além da difusividade média (MD: do Inglês "Mean Diffusivity") e das contribuições da difusão paralela (MD//) e perpendicular (MD ) às fibras, é também possível extrair outros índices, como a anisotropia fraccional (FA: do Inglês "Fractional Anisotropy"), que fornece informação acerca da percentagem de difusão anisotrópica num determinado voxel. Para a matéria branca, tal como já foi referido, existe difusão preferencial e, portanto, a anisotropia fraccional será elevada. Por outro lado, para a matéria cinzenta e para o líquido cefalorraquidiano, verificar-se-á uma FA reduzida, devido à ausência de anisotropia. Todavia, regiões com reduzida anisotropia fraccional podem camuflar regiões de conformação de cruzamento de fibras, ou fibras muito anguladas, que a imagiologia por tensor de difusão não consegue resolver. A razão para esta limitação reside no número reduzido de diferentes direcções de difusão que são exploradas, assim como o pressuposto de que a distribuição das moléculas de água é Gaussiana em todo o cérebro, o que não é necessariamente verdade. A fim de se ultrapassar estas limitações, novas técnicas surgiram, nomeadamente as de elevada resolução angular (HARDI: do Inglês "High Angular Resolution Diffusion Imaging"). Estas fazem uso de uma aquisição em função de múltiplas direcções de gradiente e de uma diferente modelação dos dados obtidos, dividindo-se em dois tipos. As técnicas livres de modelos permitem extrair uma função de distribuição da orientação das fibras num determinado voxel directamente do sinal e/ou transformações da função densidade de probabilidade do deslocamento das moléculas de água. Contrariamente, as técnicas baseadas em modelos admitem existir determinados constrangimentos anatómicos e que o sinal proveniente de um determinado voxel é originado por um conjunto de sinais individuais de fibras, caracterizados por uma distribuição preferencial das direcções das fibras. Todos estes métodos têm como objectivo principal recuperar a direcção preferencial da difusão das moléculas de água e reconstruir um trajecto tridimensional que represente a organização das fibras neuronais, pelo que se designam métodos de tractografia. Esta representa a única ferramenta não invasiva de visualização in vivo da matéria branca cerebral e o seu estudo tem revelado uma grande expansão associada ao estabelecimento de marcador biológico para diversas patologias. Adicionalmente, esta técnica tem vindo a tornar-se uma modalidade clínica de rotina e de diversos protocolos de investigação, sendo inclusivamente utilizada para complementar o planeamento em cirurgia, devido à natureza dos dados que gera. Particularmente no caso de dissecções manuais, nas quais os dados de tractografia são manuseados por pessoal especializado, com vista a realizar a parcelização de diferentes tractos de interesse, o processo é moroso e dependente do utilizador, revelando-se necessária a automatização do mesmo. Na realidade, já existem técnicas automáticas que fazem uso de algoritmos de agregação1, nos quais fibras são analisadas e agrupadas segundo características semelhantes, assim como técnicas baseadas em regiões de interesse, em que se extraem apenas os tractos seleccionados entre as regiões escolhidas. O objectivo principal desta dissertação prende-se com a análise automática de dados de tractografia, bem como a parcelização personalizada de tractos de interesse, também esta automática. Em primeiro lugar, foi desenvolvido um algoritmo capaz de lidar automaticamente com funções básicas de carregamento dos ficheiros de tractografia, o seu armazenamento em variáveis fáceis de manusear e a sua filtragem básica de acordo com regiões de interesse de teste. Neste processo de filtragem é feita a avaliação das fibras que atravessam a região de interesse considerada. Assim, após a localização das fibras entre as regiões de interesse os tractos resultantes podem ser guardados de duas formas, as quais têm, necessariamente, que ser especificadas antes de utilizar o software: um ficheiro que contém todas as fibras resultantes da parcelização e outro que contém o mapa de densidade associado, isto é, o número de fibras que se encontra em cada voxel. Após esta fase inicial, a flexibilidade e complexidade do software foi aumentando, uma vez que foram implementados novos filtros e a possibilidade de utilizar regiões de interesse de diferentes espaços anatómicos padrão. Fazendo uma análise a esta última melhoria, pode referir-se que, através de um procedimento de registo não linear da imagem anatómica do espaço padrão ao espaço individual de cada sujeito, foi possível, de forma automática, guardar o campo de deformações que caracteriza a transformação e, assim, gerar regiões de interesse personalizadas ao espaço do sujeito. Estas regiões de interesse serviram depois para a parcelização básica e para seleccionar tractos, mas também para filtragens adicionais, como a exclusão de fibras artefactuosas2 e um filtro especial, no qual apenas os pontos que ligam directamente as diferentes regiões são mantidos. Além do que já foi referido, recorreu-se também à aplicação de planos de interesse que actuam como constrangimentos neuroanatómicos, o que não permite, por exemplo, no caso da radiação óptica, que as fibras se propaguem para o lobo frontal. Esta ferramenta foi utilizada com sucesso para a parcelização automática do Fascículo Arcuado, Corpo Caloso e Radiação Óptica, tendo sido feita a comparação com a dissecção manual, em todos os casos. O estudo do Fasciculo Arcuado demonstrou ser o teste ideal para a ferramenta desenvolvida na medida que permitiu identificar o segmento longo, assim como descrito na literatura. O método automático de duas regiões de interesse deu a origem aos mesmos resultados obtidos manualmente e permitiu confirmar a necessidade de estudos mais aprofundados. Aumentando a complexidade do estudo, realizou-se a parcelização do Corpo Caloso de acordo com conectividade estrutural, isto é, com diferentes regiões envolvidas em funções distintas. Procedeu-se deste modo, e não com base em informação acerca de divisões geométricas, uma vez que estas já demonstraram incongruências quando correlacionadas com subdivisões funcionais. O uso adicional de regiões de interesse para a exclusão de fibras demonstrou-se benéfico na obtenção dos mapas finais. Finalmente, incluiu-se a utilização de um novo filtro para realizar a parcelização da Radiação Óptica, comparando os resultados para DTI e SD(do Inglês "Spherical Deconvolution"). Foi possível determinar limitações na primeira técnica que foram, no entanto, ultrapassadas pela utilização de SD. O atlas final gerado apresenta-se como uma mais-valia para o planeamento cirúrgico num ambiente clínico. O desenvolvimento desta ferramenta resultou em duas apresentações orais em conferências internacionais e encontra-se, de momento, a ser melhorada, a fim de se submeter um artigo de investigação original. Embora se tenha chegado a um resultado final positivo, tendo em conta a meta previamente estabelecida, está aberto o caminho para o seu aperfeiçoamento. Como exemplo disso, poder-se-á recorrer ao uso combinado das duas abordagens de parcelização automática e à utilização de índices específicos dos tractos, o que poderá trazer uma nova força à delineação dos tractos de interesse. Adicionalmente, é também possível melhorar os algoritmos de registo de imagem, tendo em conta a elevada variabilidade anatómica que alguns sujeitos apresentam. Como nota final, gostaria apenas de salientar que a imagiologia por difusão e, em particular, a tractografia, têm ainda muito espaço para progredir. A veracidade desta afirmação traduz-se pela existência de uma grande variedade de modelos e algoritmos implementados, sem que, no entanto, exista consenso na comunidade científica acerca da melhor abordagem a seguir.
Diffusion weighted imaging (DWI) has provided us a non-invasive technique to determine physiological information and infer about tissue microstructure. The human body is filled with barriers affecting the mobility of molecules and preventing it from being constant in different directions (anisotropic diffusion). In the brain, the sources for this anisotropy arise from dense packing axons and from the myelin sheath that surrounds them. Only with Diffusion Tensor Imaging (DTI) it was possible to fully characterize anisotropy by offering estimations for average diffusivities in each voxel. However, these methods were limited, not being able to reflect the index of anisotropic diffusion in regions with complex fibre conformations. It was possible to reduce those problems through the acquisition of many gradient directions with High Angular Resolution Diffusion Imaging (HARDI). There are model-free approaches such as Diffusion Spectrum Imaging (DSI) and Q-ball Imaging (QBI) which retrieve an orientation distribution function (ODF) directly from the water molecular displacement. Another method is Spherical Deconvolution, which is a model-based approach based on the computation of a fibre orientation distribution (FOD) from the deconvolution of the diffusion signal and a chosen fibre response function. Reconstructing the fibre orientations from the diffusion profile, generates a three-dimensional reconstruction of neuronal fibres (Tractography) whether in a deterministic, probabilistic or global way. Tractography has two main purposes: non-invasive and in vivo mapping of human white matter and neurosurgical planning. In order to achieve those purposes it is common to apply parcellation techniques which can be subdivided into ROI-based or Clustering base. The aim of this project is to develop an automated method of tract-based parcellation of different brain regions. This tool is essential to retrieve information about the architecture and connectivity of the brain, overcoming time consuming and expertise related issues derived from manual dissections. Firstly we investigated basic functions to handle diffusion and tractography data. In particular, we focused on how to load track files, filter them according to regions of interest and save the output in different formats. Results were always compared with manual dissection. The developed tool increased complexity by introduction a new filtering and the use of regions of interest from different standard spaces, created trough non-linear registrations. Three major tracts of interest were analysed: Arcuate Fasciculus, Corpus Callosum and Optic Radiation.

T1rho弛豫是旋轉坐標系中的自旋晶格弛豫，它決定橫向磁化向量在存有自旋鎖定射頻脈衝情況下的衰減，自旋鎖定脈衝與橫向磁化向量同向。T1rho磁共振成像對於低頻運動過程敏感，故可研究水與其周大分子物質環境間的交互作用，有鑒別組織內早期生化改變的潛力。
衰老與慢性高血壓是常見腦退行性疾病的兩個主要危險因素。但是正常腦衰老過程及慢性高血壓兩個因素與腦組織T1rho是否有相關性，尚缺乏研究。序貫性測量SD老鼠自5至15月齡、WKY（血壓正常）和SHR（患有自發性高血壓）老鼠自6至12月齡的雙側丘腦、海馬、和皮質的腦組織T1rho值。發現三組老鼠的丘腦、海馬及皮質的T1rho均隨年齡增長而增高；且SHR的顯著高於WKY老鼠。
T1rho值與椎間盤退變等級的相關性已有報導。但相比T2值，T1rho在評價椎間盤退變方面是否優於或如何優於T2值尚缺乏研究。將椎間盤髓核及纖維環的T1rho和T2值與5級和8級椎間盤退變等級系統做比較；發現髓核的T1rho及T2與椎間盤退變等級的相關性均呈二次函數降低，且無顯著差別(P=0.40)。纖維環的T1rho及T2與椎間盤退變等級的相關性呈線性函數降低，T2降低的斜率明顯比T1rho降低的斜率要平坦(P<0.001)。故T1rho值比T2值更加適合評價纖維環退變，而兩者在評價髓核時相似。
肝纖維化是幾乎所有慢性肝病的常見特徵，包括大分子物質在細胞外基質的沉積。選用四氯化碳CCl4腹腔注射6周來製造肝纖維化模型。肝臟T1rho在注射後的第二天輕度上升，然後持續上升，直到注射六周後T1rho達最高值，此後T1rho隨CCl4注射停止而降低。顯示T1rho磁共振成像對於監測慢性注射CCl4誘導的肝纖維化及肝損傷有價值。當沒有明顯肝纖維化時，肝T1rho輕微受水腫及急性炎症的影響。
為將肝臟T1rho磁共振成像轉化到臨床使用，我們研究了其可行性，以及正常志願者肝臟T1rho值分佈範圍。發現採用六個自旋鎖定時間來測量健康志願者肝T1rho，結果有較高的可重複性和一致性，肝T1rho平均值為42.5ms，分佈範圍為38.8到46.5ms。採用三個自鎖鎖定時間點掃描，可以減少一半掃描時間，且可以得到可信的肝T1rho值，但採用兩個自旋鎖定時間點則不行。
T1rho relaxation is spin-lattice relaxation in the rotating frame. It determines the decay of the transverse magnetization in the presence of a spin-lock radiofrequency pulse, which applied along the transverse magnetization. T1rho MRI is sensitive to low frequency motional processes, so it can be used to investigate the interaction between water molecules and their macromolecular environment. T1rho imaging is suggested to have the potential to identify early biochemical changes in tissues.
Aging and chronic hypertension are two major risk factors for common neurodegenerative disease. However, whether normal brain aging and chronic spontaneous hypertensive are associated with brain T1rho values changes were not reported. We longitudinally measured the T1rho value in rat brain of Sprague-Dawley (SD) rats from 5-month to 15-month, and spontaneous hypertensive rats (SHR) with Wistar Kyoto (WKY) rats from 6-month to 12-month. The T1rho values in three brain regions of thalamus, hippocampus, and cortices increased with aging process, and were significantly higher in SHR than WKY rats.
For intervertebral disc, the correlation between T1rho and degenerative grade has been reported. However, whether and how T1rho specifically offer better evaluation of disc degeneration compared with T2 was not studied previously. T1rho and T2 value of nucleus pulposus (NP) and annulus fibrosus (AF) was compared with reference to the five-level and eight-level semi-quantitative disc degeneration grading systems. For NP, T1rho and T2 decreased quadratically with disc degeneration grades and had no significant trend difference (P=0.40). In NP, T1rho and T2 decrease in a similar pattern following disc degeneration. For AF, T1rho and T2 decreased linearly and the slopes of T2 were significantly flatter than those of T1rho (P<0.001). Therefore, the T1rho is better suited for evaluating AF in degenerated disc than T2.
Liver fibrosis, a common feature of almost all causes of chronic liver disease, involves macromolecules accumulated within the extracellular matrix. Male Sprague-Dawley rats received intraperitoneal injection of 2 ml/kg CCl4 twice weekly for up to 6 weeks. Then CCl4 was withdrawn for recovery. The liver T1rho values increased slightly on day 2, then increased further and were highest at week 6 post CCl4 insults, and decreased upon the withdrawal of the CCl4 insult. This study demonstrated that T1rho MRI is a valuable imaging biomarker for liver injury and fibrosis induced by CCl4. Liver T1rho value was only mildly affected by edema and acute inflammation when there was no apparent fibrosis.
To translate liver T1rho MRI to clinical application, the technical feasibility of T1rho MRI in human liver was explored and the normal range of T1rho values in healthy volunteers was determined. We found it is feasible to obtain consistent liver T1rho measurement for healthy human liver with six spin-lock time (SLT) points of 1, 10, 20, 30, 40, and 50ms; the mean liver T1rho value of the healthy subjects was 42.5ms, with a range of 38.8-46.5ms. Adopting 3-SLT points of 1, 20, and 50ms for T1rho measurement could provide reliable measurement and reduce the scanning time, while 2-SLT points of 1 and 50ms do not provide reliable measurement.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Detailed summary in vernacular field only.
Zhao, Feng.
Thesis (Ph.D.)--Chinese University of Hong Kong, 2013.
Includes bibliographical references (leaves 119-143).
Abstracts also in Chinese.
ABSTRACT --- p.i
ACKNOWLEDGEMENTS --- p.vi
LIST OF FIGURES --- p.viii
LIST OF TABLES --- p.xvi
LIST OF ABBREVIATIONS --- p.xvii
CONTENTS --- p.xxi
Chapter Chapter 1 --- Introduction --- p.1
Chapter 1.1 --- Conventional Magnetic Resonance Imaging --- p.1
Chapter 1.1.1 --- Basic Principle of Conventional Magnetic Resonance Imaging --- p.1
Chapter 1.1.2 --- T1 Relaxation --- p.2
Chapter 1.1.3 --- T2 Relaxation --- p.3
Chapter 1.2 --- T1rho Magnetic Resonance Imaging --- p.3
Chapter 1.2.1 --- T1rho Relaxation --- p.3
Chapter 1.2.2 --- Principle of T1rho Magnetic Resonance Imaging --- p.4
Chapter 1.2.3 --- Radiofrequency Pulse for T1rho Magnetic Resonance Imaging --- p.5
Chapter 1.2.4 --- T1rho-weighted Contrast Imaging and Application --- p.10
Chapter 1.2.5 --- Quantitative T1rho Mapping and Application --- p.11
Chapter 1.2.6 --- T1rho Dispersion and Application --- p.13
Chapter 1.3 --- Thesis Overview --- p.14
Chapter Chapter 2 --- T1rho MRI in brain aging of animal model --- p.19
Chapter 2.1 --- Introduction --- p.19
Chapter 2.2 --- Materials and Methods --- p.20
Chapter 2.2.1 --- Animal Model of Brain Aging --- p.20
Chapter 2.2.2 --- T1rho Data Acquisition --- p.21
Chapter 2.2.3 --- T1rho Data Processing --- p.23
Chapter 2.2.4 --- T1rho Measurement and Statistical Analysis --- p.24
Chapter 2.3 --- Results --- p.27
Chapter 2.4 --- Discussion --- p.38
Chapter 2.5 --- Summary --- p.42
Chapter Chapter 3 --- T1rho MRI in lumbar disc degeneration of human subjects --- p.43
Chapter 3.1 --- Introduction --- p.43
Chapter 3.2 --- Methods --- p.45
Chapter 3.2.1 --- Subjects --- p.45
Chapter 3.2.2 --- MR Image Acquisition --- p.46
Chapter 3.2.2.1 --- T2-weighted MRI --- p.46
Chapter 3.2.2.2 --- T2 Mapping Imaging --- p.47
Chapter 3.2.2.3 --- T1rho MRI --- p.47
Chapter 3.2.3 --- Data Processing --- p.49
Chapter 3.2.4 --- Data Measurement and Statistical Analysis --- p.49
Chapter 3.3 --- Results --- p.52
Chapter 3.3.1 --- Range of T1rho/T2 Values for Discs --- p.52
Chapter 3.3.2 --- The Relationship between NP T1rho/T2 Values and 8-level Degeneration Grading of Discs --- p.52
Chapter 3.3.3 --- The Relationship between NP T1rho/T2 Values and 5-level Degeneration Grading of Discs --- p.55
Chapter 3.3.4 --- The Relationship between AF T1rho/T2 Values and 8-level Degeneration Grading of Discs --- p.58
Chapter 3.3.5 --- The Relationship between AF T1rho/T2 Values and 8-level Degeneration Grading of Discs --- p.61
Chapter 3.4 --- Discussion --- p.64
Chapter 3.5 --- Summary --- p.69
Chapter Chapter 4 --- T1rho MRI in rat liver fibrosis model induced by CCl4 insult --- p.71
Chapter 4.1 --- Introduction --- p.71
Chapter 4.2 --- Materials and Methods --- p.73
Chapter 4.2.1 --- Animal Preparation --- p.73
Chapter 4.2.2 --- MR Image Acquisition --- p.74
Chapter 4.2.2.1 --- T2-weighted MRI --- p.75
Chapter 4.2.2.2 --- T1rho MRI --- p.75
Chapter 4.2.3 --- Data Processing --- p.76
Chapter 4.2.4 --- Data Measurement and Statistical Analysis --- p.78
Chapter 4.2.5 --- Histology Analysis --- p.79
Chapter 4.3 --- Results --- p.80
Chapter 4.3.1 --- T1rho Measurement Reproducibility --- p.80
Chapter 4.3.2 --- Rat Liver T1rho Values at Different Time Phase --- p.81
Chapter 4.3.3 --- Relative Rat Liver Signal Intensity on T2WI at Different Time Phase --- p.83
Chapter 4.3.4 --- Histology Results --- p.84
Chapter 4.4 --- Discussion --- p.86
Chapter 4.5 --- Summary --- p.91
Chapter Chapter 5 --- T1rho MRI in liver of healthy human subjects --- p.93
Chapter 5.1 --- Introduction --- p.93
Chapter 5.2 --- Methods --- p.95
Chapter 5.2.1 --- Subjects --- p.95
Chapter 5.2.2 --- MR Image Acquisition --- p.96
Chapter 5.2.2.1 --- T2-weighted MRI --- p.96
Chapter 5.2.2.2 --- T1rho MRI --- p.97
Chapter 5.2.3 --- T1rho Data Processing --- p.99
Chapter 5.2.4 --- T1rho Measurement --- p.100
Chapter 5.3 --- Results --- p.102
Chapter 5.3.1 --- T1rho Measurement Reproducibility --- p.105
Chapter 5.3.2 --- T1rho Value Agreement of the Fasting Status with Post Meal Status --- p.105
Chapter 5.3.3 --- T1rho Value Agreement for T1rho Maps Constructed by Different Spin-lock Time Points --- p.106
Chapter 5.3.4 --- T1rho Value Range of Healthy Human Subjects --- p.108
Chapter 5.4 --- Discussion --- p.108
Chapter 5.5 --- Summary --- p.113
Chapter Chapter 6 --- General discussion and further work --- p.115
References: --- p.119
LIST OF PUBLICATIONS --- p.138

Healthy white matter in the brain and spinal cord is composed primarily of myelinated axons and glial cells. Myelinated axons transfer information between the peripheral nervous system and the central nervous system (CNS) as well as between centres within the CNS. Demyelination, a hallmark of neurodegenerative autoimmune diseases such as multiple sclerosis (MS), can cause nerve damage and degrade signal propagation. Magnetic resonance imaging (MRI) methods thought to assess myelin integrity and the structural integrity of axons are improving both the diagnosis and understanding of white matter diseases such as MS. Current methods, however, are sensitive to many different pathologies, making the interpretation of individual MRI results difficult. For this dissertation, several quantitative MRI methods were developed and compared, including single component T1 and T2 relaxometry, multicomponent T2 relaxometry, diffusion tensor imaging (DTI), and quantitative magnetization transfer imaging (qMTI). These methods were tested on agarose gels, fixed rat spinal cords, healthy control mice, and the cuprizone mouse model of demyelination. Quantitative MRI measurements were correlated to ultrastructural measurements of white matter to determine the influence myelin content and axonal structure have on different MRI methods. Cellular distributions measured in electron micrographs of the corpus callosum correlated strongly to several different quantitative MRI metrics. The largest Spearman correlation coefficient varied depending on cellular type: longitudinal relaxation rates (RA/T1) vs. the myelinated axon fraction ( r = 0.90/-0.90), the qMTI-derived bound pool fraction (f) vs. the myelin sheath fraction ( r = 0.93), and the DTI-derived axial diffusivity vs. the non-myelinated cell fraction (r = 0.92). Using Pearson’s correlation coefficient, f was strongly correlated to the myelin sheath fraction (r = 0.98) with a linear equation predicting myelin content (5.37f −0.25). Of the calculated MRI metrics, f was the strongest indicator of myelin content while longitudinal relaxation rates and diffusivity measurements were the strongest indicators of changes in tissue structure. Multiparametric MRI measurements of relaxation, diffusion, and magnetization transfer give a more complete picture of white matter integrity.

Infinite-Dimensional Systems (IDSs) which have been made possible by recent advances in mathematical and computational tools can be used to model complex real phenomena. However, due to physical, economic, or stringent non-invasive constraints on real systems, the underlying characteristics for mathematical models in general (and IDSs in particular) are often missing or subject to uncertainty. Therefore, developing efficient estimation techniques to extract missing pieces of information from available measurements is essential. The human brain is an example of IDSs with severe constraints on information collection from controlled experiments and invasive sensors. Investigating the intriguing modeling potential of the brain is, in fact, the main motivation for this work. Here, we will characterize the hemodynamic behavior of the brain using functional magnetic resonance imaging data. In this regard, we propose efficient estimation methods for two classes of IDSs, namely Partial Differential Equations (PDEs) and Fractional Differential Equations (FDEs). This work is divided into two parts. The first part addresses the joint estimation problem of the state, parameters, and input for a coupled second-order hyperbolic PDE and an infinite-dimensional ordinary differential equation using sampled-in-space measurements. Two estimation techniques are proposed: a Kalman-based algorithm that relies on a reduced finite-dimensional model of the IDS, and an infinite-dimensional adaptive estimator whose convergence proof is based on the Lyapunov approach. We study and discuss the identifiability of the unknown variables for both cases. The second part contributes to the development of estimation methods for FDEs where major challenges arise in estimating fractional differentiation orders and non-smooth pointwise inputs. First, we propose a fractional high-order sliding mode observer to jointly estimate the pseudo-state and input of commensurate FDEs. Second, we propose a modulating function-based algorithm for the joint estimation of the parameters and fractional differentiation orders of non-commensurate FDEs. Sufficient conditions ensuring the local convergence of the proposed algorithm are provided. Subsequently, we extend the latter technique to estimate smooth and non-smooth pointwise inputs. The performance of the proposed estimation techniques is illustrated on a neurovascular-hemodynamic response model. However, the formulations are efficiently generic to be applied to a wide set of additional applications.

Trabalho Final do Mestrado Integrado em Medicina apresentado à Faculdade de Medicina
O síndrome de Usher (USH) é uma doença autossómica recessiva rara que cursa com alterações da visão e audição apresentando heterogeneidade clínica e genética. Vários estudos psicofísicos e de imagiologia estrutural evidenciaram também a existência de défices olfativos em doentes com USH. No entanto, o efeito desta condição no circuito central de processamento olfativo ainda não foi avaliado através de imagiologia funcional. Deste modo, procurámos comparar a atividade cerebral relacionada com uma tarefa olfativa nos córtices orbito-frontal (COF) e piriforme (CP) entre doentes com USH e indivíduos saudáveis. Além disso, foi realizada uma análise de classificação entre grupos de modo a avaliar o potencial da imagiologia funcional para discriminar doentes com USH de indivíduos saudáveis.Vinte e seis indivíduos saudáveis sem história de disfunção olfativa e 27 doentes com USH (4 USH1, 21 USH2, 2 USH3) foram incluídos neste estudo. Todos os sujeitos realizaram a mesma tarefa de deteção olfativa durante as sessões de ressonância magnética funcional para avaliar as respostas evocadas no COF e CP. Quatro níveis de concentração de butanol foram apresentados a cada participante. As regiões cerebrais foram definidas funcionalmente através do Neurosynth, uma ferramenta de meta-análise automatizada. Na análise univariada foi ajustado um modelo linear geral multi-sujeito com efeitos aleatórios e os parâmetros beta estimados de cada região foram usados para a comparação entre grupos. Na análise de classificação foi ajustado um modelo linear geral com sujeitos separados e foram criados mapas de estatística t para cada sujeito. Estes mapas foram classificados através de um modelo de regressão logística.Verificou-se a existência de um efeito de interação entre o grupo e o nível de butanol no COF direito (F(2,365;118,247)=3,032, p=0,043) e no CP direito (F(3,150)=4,537, p=0,004). Não se verificou nenhum efeito significativo na ativação cerebral evocada pelo estímulo olfativo no COF e CP esquerdos. O contraste planeado da ativação cerebral da maior concentração de odor menos a da menor concentração de odor entre grupos revelou uma diferença significativa no COF direito (t(51)=2,339, p=0,023). O mesmo contraste mostrou uma diferença significativa entre doentes e controlos no CP direito (t(51)=-3.380, p=0.001).Quanto à análise de classificação de doentes versus controlos, apresentamos um modelo preditivo com precisão de 71,7% (p=0,0072), sensibilidade de 67,7% (p=0,0328), especificidade de 77,3% (p=0,0041) e AUC de 0,785 (p=0,0087).Estes resultados evidenciam uma diminuição da ativação no CP direito e um aumento compensatório da ativação no COF direito em doentes com USH reforçando a noção de olfação disfuncional neste síndrome. Além disso, sugerem que os padrões de ativação cerebral em regiões olfativas medidos por ressonância magnética funcional permitem discriminar doentes com USH de indivíduos saudáveis sendo uma técnica promissora em termos de diagnóstico deste síndrome.
Usher syndrome (USH) is a rare autosomal recessive disease, affecting vision and audition, and showing clinical and genetic heterogeneity. Evidence of olfactory impairment in USH patients has emerged through psychophysical and structural imaging studies. However, the effect of this condition in the central olfactory processing network has not yet been evaluated through functional imaging studies. We sought to compare olfactory task-related activity in the orbitofrontal (OFC) and piriform (PC) cortices between USH patients and healthy subjects. Also, a classification analysis between these groups was carried out to assess functional imaging potential of discriminating USH patients.Twenty-six age- and gender-matched controls with no history of olfactory dysfunction and 27 USH patients (4 USH1, 21 USH2, 2 USH3) were studied. Functional magnetic resonance imaging (fMRI) was used with an olfactory detection task to evaluate responses in the OFC and PC. Four butanol concentration levels were presented to each participant. These regions were functionally defined using an automated meta-analysis toolbox, Neurosynth. In the univariate analyses a multi-subject general linear model (GLM) with random effects was performed and the beta estimates from each region were used to compare between groups. In the classification analysis a separate-subject GLM was performed and t-statistic maps were created for each subject which then were used as input to a logistic regression classifier.An interaction effect between group and butanol level was found in the right OFC (F(2.365;118.247)=3.032, p=0.043). Also, an interaction effect between group and butanol level emerged in the right PC (F(3,150)=4.537, p=0.004). Stimulus-evoked activation in both the left OFC and left PC did not show any significant effect. Planned contrast of the highest odor concentration minus the lowest odor concentration activation between groups revealed a significant difference in the right OFC (t(51)=2.339, p=0.023). The same contrast showed a significant difference between USH patients and controls in the right PC (t(51)=-3.380, p=0.001).As for the USH patients vs controls classification analysis we report a predictor model with accuracy of 71.7% (p=0.0072), sensitivity of 67.7% (p=0.0328), specificity of 77.3% (p=0.0041) and an AUC of 0.785 (p=0.0087).These data provide evidence of decreased activation in the right PC and increased compensatory activation in the right OFC in USH patients reinforcing the notion of dysfunctional olfactory sensory function. Also, it shows that olfactory fMRI patterns can discriminate USH patients from controls which holds promise in USH diagnosis improvement.

We identified human visual field maps, LO1 and LO2, in object-selective lateral occipital cortex. Using transcranial magnetic stimulation (TMS), we assessed the functions of these maps in the perception of orientation and shape. TMS of LO1 disrupted orientation, but not shape, discrimination, whereas TMS of LO2 disrupted shape, but not orientation, discrimination. This double dissociation suggests that specialized and independent processing of different visual attributes occurs in LO1 and LO2.