This thesis describes the development, implementation and application of various quantitative functional magnetic resonance imaging (fMRI) approaches at ultra-high magnetic field including the assessment with regards to applicability and reproducibility. Functional MRI (fMRI) commonly uses the blood oxygenation level dependent (BOLD) contrast to detect functionally induced changes in the oxy-deoxyhaemoglobin composition of blood which reflect cerebral neural activity. As these blood oxygenation changes do not only occur at the activation site but also downstream in the draining veins, the spatial specificity of the BOLD signal is limited. Therefore, the focus has moved towards more quantitative fMRI approaches such as arterial spin labelling (ASL), vascular space occupancy (VASO) or calibrated fMRI which measure quantifiable physiologically and physically relevant parameters such as cerebral blood flow (CBF), cerebral blood volume (CBV) or cerebral metabolic rate of oxygen (CMRO2), respectively. In this thesis a novel MRI technique was introduced which allowed the simultaneous acquisition of multiple physiological parameters in order to beneficially utilise their spatial and temporal characteristics. The advantages of ultra-high magnetic field were utilised to achieve higher signal-to-noise and contrast-to-noise ratios compared to lower field strengths. This technique was successfully used to study the spatial and temporal characteristics of CBV, CBF and BOLD in the visual cortex. This technique is the first one that allows simultaneous acquisition of CBV, CBF and BOLD weighted fMRI signals in the human brain at 7 Tesla. Additionally, this thesis presented a calibrated fMRI technique which allowed the quantitative estimation of changes in cerebral oxygen metabolism at ultra-high field. CMRO2 reflects the amount of thermodynamic work due to neural activity and is therefore a significant physical measure in neuroscience. The calibrated fMRI approach presented in this thesis was optimised for the use at ultra-high field by adjusting the MRI parameters as well as implementing a specifically designed radio-frequency (RF) pulse. A biophysical model was used to calibrate the fMRI data based on the simultaneous acquisition of BOLD and CBF weighted MRI signals during a gas-breathing challenge. The reproducibility was assessed across multiple brain regions and compared to that of various physiologically relevant parameters. The results indicate that the degree of intra-subject variation for calibrated fMRI is lower than for the classic BOLD contrast or ASL. Consequently, calibrated fMRI is a viable alternative to classic fMRI contrasts with regards to spatial specificity as well as functional reproducibility. This calibrated fMRI approach was also compared to a novel direct calibration technique which relies on complete venous oxygenation saturation during the calibration scan via a gas-breathing challenge. This thesis introduced several reliable quantitative fMRI approaches at 7 Tesla and the results presented are a step forward to the wider application of quantitative fMRI.

In the field of dynamic capillaroscopy, measurements of the capillary blood cell velocity (CBV) give significant insight into the human body. For instance, diabetes, hypertension and peripheral arterial occlusive disease all affect CBV. However, the videos used to measure CBV – captured with a microscope – are often displaced in relation to the microscope by small motions of the finger or toe. Stabilisation algorithms are commonly used to reduce this problem, in order to carry out measurements such as CBV using the stabilised video. Artificial capillaroscopy videos were used to compare the stabilisation algorithms Mutual information, Single-step DFT, Block matching and Phase correlation in terms of computational time; RMSE, PSNR and MSE; and resistance to blurring effects. Single-step DFT was indicated to be the best suited algorithm in all aforementioned metrics. / Inom dynamisk kapilläroskopi ger mätningar av kapillär blod-cell hastighet (CBV) signifikanta insikter inom den mänskliga kroppen. Till exempel, diabetes, hypertoni och perifer arteriell ocklusiv sjukdom påverkar CBV. Däremot är videorna som används för att mäta CBV – tagna med ett mikroskop – ofta förskjutna i relation till mikroskopet på grund av små rörelser av ett finger eller en tå. Stabiliseringsalgoritmer används vanligen för att reducera detta problemet med avsikt att därefter använda den stabiliserade videon för att mäta viktiga egenskaper, som till exempel CBV. Artificiella kapilläroskopivideor användes för att jämföra stabiliseringsalgorithmerna Mutual information, Single-step DFT, Block matching and Phase correlation inom beräkningstid; RMSE, PSNR och MSE; och resistens mot suddighet. Single-step DFT indikerades som den bäst lämpade algoritmen inom de ovannämnda måtten.

This thesis describes the development, implementation and application of various quantitative functional magnetic resonance imaging (fMRI) approaches at ultra-high magnetic field including the assessment with regards to applicability and reproducibility. Functional MRI (fMRI) commonly uses the blood oxygenation level dependent (BOLD) contrast to detect functionally induced changes in the oxy-deoxyhaemoglobin composition of blood which reflect cerebral neural activity. As these blood oxygenation changes do not only occur at the activation site but also downstream in the draining veins, the spatial specificity of the BOLD signal is limited. Therefore, the focus has moved towards more quantitative fMRI approaches such as arterial spin labelling (ASL), vascular space occupancy (VASO) or calibrated fMRI which measure quantifiable physiologically and physically relevant parameters such as cerebral blood flow (CBF), cerebral blood volume (CBV) or cerebral metabolic rate of oxygen (CMRO2), respectively. In this thesis a novel MRI technique was introduced which allowed the simultaneous acquisition of multiple physiological parameters in order to beneficially utilise their spatial and temporal characteristics. The advantages of ultra-high magnetic field were utilised to achieve higher signal-to-noise and contrast-to-noise ratios compared to lower field strengths. This technique was successfully used to study the spatial and temporal characteristics of CBV, CBF and BOLD in the visual cortex. This technique is the first one that allows simultaneous acquisition of CBV, CBF and BOLD weighted fMRI signals in the human brain at 7 Tesla. Additionally, this thesis presented a calibrated fMRI technique which allowed the quantitative estimation of changes in cerebral oxygen metabolism at ultra-high field. CMRO2 reflects the amount of thermodynamic work due to neural activity and is therefore a significant physical measure in neuroscience. The calibrated fMRI approach presented in this thesis was optimised for the use at ultra-high field by adjusting the MRI parameters as well as implementing a specifically designed radio-frequency (RF) pulse. A biophysical model was used to calibrate the fMRI data based on the simultaneous acquisition of BOLD and CBF weighted MRI signals during a gas-breathing challenge. The reproducibility was assessed across multiple brain regions and compared to that of various physiologically relevant parameters. The results indicate that the degree of intra-subject variation for calibrated fMRI is lower than for the classic BOLD contrast or ASL. Consequently, calibrated fMRI is a viable alternative to classic fMRI contrasts with regards to spatial specificity as well as functional reproducibility. This calibrated fMRI approach was also compared to a novel direct calibration technique which relies on complete venous oxygenation saturation during the calibration scan via a gas-breathing challenge. This thesis introduced several reliable quantitative fMRI approaches at 7 Tesla and the results presented are a step forward to the wider application of quantitative fMRI.:1 Introduction 3 2 Background to Functional Magnetic Resonance Imaging 7 2.1 Magnetic Resonance 7 2.1.1 Quantum Mechanics 7 2.1.2 The Classical Point of View 10 2.1.3 Radio Frequency Pulses 12 2.1.4 Relaxation Effects 13 2.1.5 The Bloch Equations 15 2.2 Magnetic Resonance Imaging 16 2.2.1 Data Acquisition 16 2.2.2 Image Formation 17 2.2.2.1 Slice Selection 17 2.2.2.2 Frequency Encoding 18 2.2.2.3 Phase Encoding 19 2.2.2.4 Mathematics of Image Formation 20 2.2.2.5 Signal Formation 22 2.3 Advanced Imaging Methods 24 2.3.1 Echo-Planar Imaging (EPI) 24 2.3.2 Partial Fourier Acquisition 25 2.3.3 Generalised Autocalibrating Partially Parallel Acquisition (GRAPPA) 25 2.3.4 Inversion Recovery (IR) 26 2.3.5 Adiabatic Inversion 26 2.3.5.1 Hyperbolic Secant (HS) RF pulses 28 2.3.5.2 Time Resampled Frequency Offset Corrected Inversion (tr-FOCI) RF Pulses 28 2.4 Physiological Background 29 2.4.1 Neuronal Activity 30 2.4.2 Energy Metabolism 31 2.4.3 Physiological Changes During Brain Activation 32 2.4.4 The BOLD Contrast 34 2.4.5 Disadvantages of the BOLD Contrast 35 2.5 Arterial Spin Labelling (ASL) 35 2.5.1 Pulsed Arterial Spin Labelling 37 2.5.2 Arterial Spin Labelling at Ultra-High Field 41 2.6 Vascular Space Occupancy (VASO) 42 2.6.1 VASO at Ultra-High Field 44 2.6.2 Slice-Saturation Slab-Inversion (SS-SI) VASO 45 2.7 Calibrated Functional Magnetic Resonance Imaging 47 2.7.1 The Davis Model 47 2.7.2 The Chiarelli Model 50 2.7.3 The Generalised Calibration Model (GCM) 52 3 Materials and Methods 53 3.1 Scanner Setup 53 3.2 Gas Delivery and Physiological Monitoring System 53 3.3 MRI Sequence Developments 55 3.3.1 Tr-FOCI Adiabatic Inversion 55 3.3.2 Optimisation of the PASL FAIR QUIPSSII Sequence Parameters 60 3.3.3 Multi-TE Multi-TI EPI 64 4 Experiment I: Comparison of Direct and Modelled fMRI Calibration 68 4.1 Background Information 68 4.2 Methods 69 4.2.1 Experimental Design 69 4.2.2 Visuo-Motor Task 70 4.2.3 Gas Manipulations 71 4.2.4 Scanning Parameters 71 4.2.5 Data Analysis 72 4.2.6 M-value Modelling 72 4.2.7 Direct M-Value Estimation 73 4.3 Results 74 4.4 Discussion 79 4.4.1 M-value Estimation 79 4.4.2 BOLD Time Courses 82 4.4.3 M-Maps and Single Subject Analysis 82 4.4.4 Effects on CMRO2 Estimation 83 4.4.5 Technical Limitations and Implications for Calibrated fMRI 84 4.5 Conclusion 89 5 Experiment II: Reproducibility of BOLD, ASL and Calibrated fMRI 90 5.1 Background Information 90 5.2 Methods 91 5.2.1 Experimental Design 91 5.2.2 Data Analysis 91 5.2.3 Reproducibility 93 5.2.4 Learning and Habituation Effects 95 5.3 Results 95 5.4 Discussion 101 5.4.1 Breathing Manipulations 102 5.4.2 Functional Reproducibility 107 5.4.3 Habituation Effects on Reproducibility 109 5.4.4 Technical Considerations for Calibrated fMRI 110 5.5 Conclusion 112 6 Experiment III: Simultaneous Acquisition of BOLD, ASL and VASO Signals 113 6.1 Background Information 113 6.2 Methods 114 6.2.1 SS-SI VASO Signal Acquisition 114 6.2.2 ASL and BOLD Signal Acquisition 114 6.2.3 Experimental Design 114 6.2.4 Data Analysis 115 6.3 Results 115 6.4 Discussion 116 6.5 Conclusion 120 7 Conclusion and Outlook 121

L’IRM apporte des informations morphologiques concernant la tumeur, mais également des informations concernant sa micro-vascularisation. En TEP/TDM, l’accumulation de la 18F-FDopa dans les cellules tumorales résulte de l’activité métabolique plus importante que celle des tissus sains. Nous avons étudié 28 gliomes pour lesquels nous avons analysé les données provenant d’IRM et de TEP/TDM. Une méthode de recalage a été développée afin de combiner les informations issues des deux modalités TEP et IRM et d’extraire des volumes d’intérêt sur la base des données conjointes TEP et IRM. L’analyse du contenu de ces volumes d’intérêt par un modèle de mélange gaussien a permis de différencier, dans ces volumes, les tissus tumoraux et les tissus sains, et d’obtenir ainsi des volumes tumoraux et de référence communs pour les modalités TEP et IRM. Des paramètres issus de la TEP ou de l’IRM ont ensuite été calculés dans ces volumes communs aux deux modalités, pour caractériser les tumeurs et les tissus sains. L’analyse discriminante linéaire (ADL) des données TEP/TDM et d’IRM combinées permet de discriminer les différentes classes tissulaires. Les courbes Receiver Operating Characteristic ROC combinées à l’ADL permettent d’évaluer les critères multiples [SUVmax , rCBV] et [rk1 , rCBV] et conduisent à des AUC respectives de 0,88 et 0,92. En considérant les informations combinées [SUVmax , rCBV], nous avons obtenu une sensibilité de détection des tumeurs de haut grade de 95% pour une spécificité correspondante de 60% ainsi qu’une valeur prédictive négative de 52% pour une valeur prédictive positive de 95%. De même, avec le critère [rk1 , rCBV], nous avons obtenu une spécificité de 60% pour 95% de sensibilité de détection des tumeurs de haut grade ainsi qu’une valeur prédictive négative de 60% pour une valeur prédictive positive de 95%. Nos travaux montrent que la fusion des informations microvasculaires et métaboliques est possible. Dans le cas du diagnostic différentiel des gliomes, l’information microvasculaire n’apporte cependant pas d’information plus discriminante que l’information métabolique seule. / MRI provides morphological information about tumour, but also provides information regarding the micro-vascularization of the tumour. In PET/CT, the accumulation of 18F-FDopa in tumour cells results from the metabolic activity greater than that of healthy tissues. We studied 28 gliomas for which we analysed data from MRI and PET/CT. A registration method has been developed to combine information from both PET and MRI and to extract volumes of interest consistent with the information included in the two modalities. In these volumes, the tumour compartment and normal tissue compartment were identified using a Gaussian mixture model. Parameters from PET or MRI data were then calculated in these compartments. ROC analyses combined with linear discriminant analyses were used to assess whether joint observation of standardized uptake value (SUVmax ) and relative Cerebral Blood Volume (rCBV) or of relative rk1 and rCBV could distinguish between low grade and high grade tumours. We found that using this joint analysis, 82.4% of high-grade tumors and 70.0% of low-grade tumors were correctly classified (AUC of 0.88 for [SUVmax , rCBV] and of 0.92 for [rk1 , rCBV]). Considering the [SUVmax , rCBV] combined information, the sensitivity for detecting high-grade tumors was 95% with a specificity of 60%. The negative predictive value was 52% for a positive predictive value of 95%. Similarly, considering the [rk1 , rCBV] combined information, we also a specificity of 60% associated with a 95% sensitivity for detecting high-grade tumors, with a negative predictive value of 60% and positive predictive value of 95%. Our work shows that joint analysis of microvascular and metabolic information is possible by combining PET and MR imaging data. However, we found that, in our patient population, the microvascular information given by MR did not bring information more discriminating than the metabolic information derived from PET only.

Der ischämische Schlaganfall ist ein ernstzunehmendes Ereignis, welches rascher Rekanalisationstherapie bedarf. Hierfür stehen mehrere Therapieansätze zur Verfügung. Bildgebungsgestützte Patientenselektion zur individuell geeigneten Therapie kann das abschließende klinische Behandlungsergebnis des einzelnen Patienten maßgeblich verbessern. Der Alberta Stroke Program Early CT Score (ASPECTS), eine einfach und schnell anwendbare 10-Punkte-Skala zur Auswertung von Schädel-CT-Untersuchungen, wurde bereits als hilfreicher Prädiktor für das klinische Behandlungsergebnis nach erfolgreicher thrombolytischer Therapie identifiziert. Ein Nachteil der nativen Schädel-CT ist, dass der Infarktkern erst mit mehreren Stunden Verzögerung erkennbar wird. Das aktuelle Ausmaß des Infarktkerns kann durch Bestimmung des zerebralen Blutvolumens (CBV) anhand von Perfusions-CT-Untersuchungen schneller ermittelt werden.  Diese Studie analysiert retrospektiv multimodale CT-Bildgebung einer Patientenkohorte von 51 Patienten mit akutem ischämischen Schlaganfall aufgrund eines Verschlusses im M1-Segment der Arteria cerebri media bezüglich des Behandlungsergebnisses nach endovaskulärer Therapie. Die CT-Daten wurden mit kommerzieller Computersoftware nachverarbeitet. Zwei erfahrene Neuroradiologen werteten getrennt voneinander Nativ- und Perfusions-CT-Daten mithilfe des ASPECTS aus. Die Befunde der Patientengruppe mit schlechtem klinischen Behandlungsergebnis wurden mit denen der Gruppe mit gutem Ergebnis verglichen. Variablen, welche statistisch signifikante Unterschiede aufwiesen, wurden daraufhin untersucht. Zwischen den Ergebnissen der Rekanalisationstherapie, den Zeitintervallen oder den nativen CT-ASPECTS-Ergebnissen beider Gruppen bestand kein signifikanter Unterschied. Signifikant unterschieden sich die Gruppen im Hinblick auf das Patientenalter. Die übrigen Basischarakteristika der beiden Patientengruppen unterschieden sich nicht signifikant. Es ergaben sich weiterhin signifikante Unterschiede im Bezug auf den ASPECTS für zerebralen Blutfluss (CBF-ASPECTS) und die Differenz zwischen ASPECTS für zerebrales Blutvolumen (CBV-ASPECTS) und CBF-ASPECTS [Δ(CBV - CBF)-ASPECTS]. Für CBV-ASPECTS > 7 konnte die höchste Sensitivität (84 %) und Spezifität (79 %) zur Voraussage eines guten klinischen Behandlungsergebnisses ermittelt werden.  Diese Studie zeigt, dass durch ASPECTS ausgewertete CT-Perfusionsparameter eine optimale Voraussagekraft für das klinische Behandlungsergebnis nach erfolgreicher Rekanalisationstherapie besitzen und diese sensitiver und spezifischer als der native CT-ASPECTS sind. Der ASPECTS erlaubt einen einfachen und schnellen quantitativen Überblick über die tatsächliche aktuelle Situation des einzelnen Patienten. Die Berücksichtigung dieser Parameter bei Therapieentscheidungen könnte helfen, Patienten der geeigneten Therapie zuzuführen und die Anzahl vergeblicher Rekanalisationsbehandlungen zu reduzieren.