@phdthesis{Klauer2018,
  author    = {Klauer, Peter},
  title     = {Vollst{\"a}ndig integrierter Traveling-Wave-MPI-MRI-Hybridscanner},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-161314},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2018},
  abstract  = {Magnetic Particle Imaging (MPI) ist ein neuartiges tomographisches Bildgebungsverfahren, welches in der Lage ist, dreidimensional die Verteilung von superparamagnetischen Nanopartikeln zu detektieren. Aufgrund des direkten Nachweises des Tracers ist MPI ein sehr schnelles und sensitives Verfahren [12] und ben{\"o}tigt f{\"u}r eine Einordnung des Tracers (z.B. im Gewebe) eine weitere bildgebende Modalit{\"a}t wie die Magnetresonanztomographie (MRI) oder die Computertomographie. Die strukturelle Einordnung wird h{\"a}ufig mit dem Fusion-Imaging-Verfahren durchgef{\"u}hrt, bei dem die Proben separat in den Ger{\"a}ten vermessen und die Datens{\"a}tze retrospektiv korreliert werden [75][76]. In einem ersten Experiment wurde bereits ein Traveling-Wave-MPI-Scanner (TWMPI) [17] mit einem Niederfeld-MRI-Scanner kombiniert und die ersten Hybridmessung durchgef{\"u}hrt [15]. Der technische Aufwand, zwei separate Ger{\"a}te aufzubauen sowie die Tatsache, dass ein MRI-Ger{\"a}t bei 30mT sehr lange ben{\"o}tigt, diente als Motivation f{\"u}r ein integriertes TWMPIMRI- Hybridsystem, bei dem das dynamische lineare Gradientenarray (dLGA) eines TWMPI-Scanners intrinsisch das B0-Feld f{\"u}r ein MRI-Ger{\"a}t erzeugen sollte. Das Ziel dieser Arbeit war es, die Grundlagen f{\"u}r einen integrierten TWMPI-MRIHybridscanner zu schaffen. Die Geometrie des dLGAs sollte dabei nicht ver{\"a}ndert werden, damit TWMPI-Messungen weiterhin ohne Einschr{\"a}nkungen m{\"o}glich sind. Zusammenfassend werden hier noch mal die wichtigsten Schritte und Ergebnisse dieser Arbeit aufgezeigt. Zu Beginn dieser Arbeit wurde mittels Magnetfeldsimulationen nach einer geeigneten Stromverteilung gesucht, um allein mit dem dLGA ein ausreichend homogenes Magnetfeld erzeugen zu k{\"o}nnen. Die Ergebnisse der Simulationen zeigten, dass bereits zwei unterschiedliche Str{\"o}me in 14 der 20 Einzelspulen des dLGAs gen{\"u}gten, um ein Field of View (FOV) mit der Gr{\"o}ße 36mm x 12mm mit ausreichender Homogenit{\"a}t zu erreichen. Die Homogenit{\"a}t innerhalb des FOVs betrug dabei 3000 ppm. F{\"u}r die angestrebte Feldst{\"a}rke von 235mT waren Stromst{\"a}rken von 129A und 124A n{\"o}tig. Die hohen Str{\"o}me des dLGAs erforderten die Entwicklung eines daf{\"u}r angepassten Verst{\"a}rkers. Das urspr{\"u}ngliche Konzept, welches auf einem linear angesteuerten Leistungstransistors aufbaute, wurde in zahlreichen Schritten so weit verbessert, dass die n{\"o}tigen Stromst{\"a}rken stabil an- und ausgeschaltet werden konnten. Mithilfe eines Ganzk{\"o}rper-MRIs konnte erstmals das B0-Feld des dLGAs, welches durch den selbstgebauten Verst{\"a}rker erzeugt wurde, gemessen und mit der Simulation verglichen werden. Zwischen den beiden Verl{\"a}ufen zeigte sich eine qualitativ gute {\"U}bereinstimmung. Das Finden des NMR-Signals stellte wegen des selbstgebauten Verst{\"a}rkers eine Herausforderung dar, da zu diesem Zeitpunkt die n{\"o}tige Pr{\"a}zision noch nicht erreicht wurde und der wichtigste Parameter, die Magnetfeldst{\"a}rke im dLGA, nicht gemessen werden konnte. Dagegen konnte die L{\"a}nge der Pulse f{\"u}r die Spin-Echo- Sequenz sehr gut gemessen werden, jedoch war der optimale Wert noch nicht bekannt. Durch iterative Messungen wurden die richtigen Einstellungen gefunden, die nach {\"A}nderungen an der Hardware jeweils angepasst wurden. Die Performanz des Verst{\"a}rkers konnte anhand wiederholter Messungen des NMRSignals genauer untersucht werden. Es zeigte sich, dass die Pr{\"a}zision weiter verbessert werden musste, um reproduzierbare Ergebnisse zu erhalten. Mithilfe des NMR-Signals konnten auch das B0-Feld ausgemessen werden. Es zeigte eine gute {\"U}bereinstimmung zur Simulation. Mithilfe von vier Segmentspulen des dLGAs war es m{\"o}glich einen linearen Gradienten entlang der z-Achse zu erzeugen. Ein Gradient wurde zus{\"a}tzlich zum B0-Feld geschaltet und ebenfalls ausgemessen. Auch dieser Verlauf zeigte eine gute {\"U}bereinstimmung zur Simulation. Mithilfe des Gradienten wurde erfolgreich die Frequenzkodierung und die Phasenkodierung implementiert, durch die bei beiden Messungen zwei Proben anhand des Ortes unterschieden werden konnten. Damit war die Entwicklung des MRIScanners abgeschlossen. Der Aufbau des TWMPI-Scanners ben{\"o}tigte neben dem Bau des dLGAs die Anfertigung von Sattelspulen. F{\"u}r die MPI-Messungen konnte der fehlende Teil der Sendekette sowie die gesamte Empfangskette von einer fr{\"u}heren Version benutzt werden. Auch f{\"u}r das MPI wurde die Funktionalit{\"a}t mithilfe einer Punktprobe und eines Phantoms {\"u}berpr{\"u}ft, allerdings hier in zwei Dimensionen. Die Erweiterung zu einem Hybridscanner erforderte weitere Modifikationen gegen{\"u}ber einem reinen TWMPI- bzw. MRI-Scanner. Es musste ein Weg gefunden werden, die Beschaltung des dLGAs f{\"u}r die jeweilige Modalit{\"a}t z{\"u}gig anzupassen. Daf{\"u}r wurde ein Steckbrett gebaut, das es erlaubt, die Verkabelung des dLGAs in kurzer Zeit zu {\"a}ndern. Außerdem mussten innerhalb des dLGAs die Sattelspulen und die Empfangsspule des TWMPIs sowie die Empfangsspule des MRIs untergebracht werden. Ein modulares System erlaubte die gleichzeitige Anordnung aller Komponenten innerhalb des dLGAs. Das messbare FOV des MRIs ist der Homogenit{\"a}t des B0-Feldes angepasst, das FOV des TWMPI ist ausgedehnter. Zum Ende dieser Arbeit wurde erfolgreich eine Hybridmessung durchgef{\"u}hrt. Das Phantom bestand aus je zwei Kugeln gef{\"u}llt mit {\"O}l und mit einem MPI-Tracer (Resovist). Mit TWMPI war die r{\"a}umliche Abbildung der Resovistkugeln m{\"o}glich, w{\"a}hrend mit MRI die der {\"O}lkugeln m{\"o}glich war. Diese in situ Messung zeigte die erfolgreiche Umsetzung des Konzeptes f{\"u}r den TWMPI-MRI-Hybridscanner. Zusammenfassend wurden in dieser Arbeit die Grundlagen f{\"u}r einen TWMPIMRI- Hybridscanner gelegt. Die gr{\"o}ßte Schwierigkeit bestand darin, ein ausreichend homogenes B0-Feld f{\"u}r das MRI zu erzeugen, mit dem man ein gutes NMRSignal aufnehmen konnte. Mit einer einfachen Stromverteilung, bestehend aus zwei unterschiedlichen Str{\"o}men, konnte ein ausreichend homogenes B0-Feld erzeugt werden. Durch komplexere Stromverteilungen l{\"a}sst sich die Homogenit{\"a}t noch verbessern und somit das FOV vergr{\"o}ßern. Die MRI-Bildgebung wurde in dieser Arbeit f{\"u}r eine Dimension implementiert und soll in fortf{\"u}hrenden Arbeiten auf 2D und 3D ausgedehnt werden. Letztendlich soll anhand eines MRI-Bildes die Partikelverteilung des MPI-Tracers in Lebewesen deren Anatomie zugeordnet werden. In [76][77][78] sind die ersten pr{\"a}klinischen Anwendungen mit dem TWMPI-Scanner durchgef{\"u}hrt worden. Diese Anwendungen erlangen eine h{\"o}here Aussagekraft durch die zus{\"a}tzlichen Informationen eines TWMPI-MRI-Hybridscanners. In weiteren Arbeiten sollte zus{\"a}tzlich die Gr{\"o}ße des FOVs f{\"u}r das MRI erweitert werden. Außerdem macht es Sinn, einen elektronischen Schalter zum Umschalten des dLGAs zwischen MRI und MPI zu realisieren. Die n{\"a}chste Version des Hybridscanners k{\"o}nnte beispielsweise ein komplett neu gestaltetes dLGA enthalten, in dem jede Segmentspule in radialer Richtung einmal geteilt wird und dadurch in eine innere und eine {\"a}ußere Spule zerlegt wird. F{\"u}r das MRI werden die beiden Spulenteile gegen geschaltet, um ein homogenes Feld in radialer Richtung zu erhalten. F{\"u}r das TWMPI werden die Spulenteile gleichgeschaltet, um einen m{\"o}glichst starken Feldgradienten zu erreichen. In dieser Arbeit wurde f{\"u}r die n{\"a}chste Version eines TWMPI-MRI-Hybridscanners viel Wissen generiert, das {\"a}ußerst hilfreich f{\"u}r das neue Design sein wird. Anhand der Vermessung des B0-Feldes hat sich gezeigt, dass die simulierten Magnetfelder gut mit den gemessenen Magnetfeldern {\"u}bereinstimmen. Außerdem wurde viel gelernt {\"u}ber die Kombination von TWMPI mit MRI.},
  subject      = {Magnetpartikelbildgebung},
  language  = {de}
}
@phdthesis{MendesPereira2019,
  author    = {Mendes Pereira, Lenon},
  title     = {Morphological and Functional Ultrashort Echo Time (UTE) Magnetic Resonance Imaging of the Human Lung},
  doi       = {10.25972/OPUS-18317},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-183176},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2019},
  abstract  = {In this thesis, a 3D Ultrashort echo time (3D-UTE) sequence was introduced in the Self-gated Non-Contrast-Enhanced Functional Lung Imaging (SENCEFUL) framework. The sequence was developed and implemented on a 3 Tesla MR scanner. The 3D-UTE technique consisted of a nonselective RF pulse followed by a koosh ball quasi-random sampling order of the k-space. Measurements in free-breathing and without contrast agent were performed in healthy subjects and a patient with lung cancer. A gating technique, using a combination of different coils with high signal correlation, was evaluated in-vivo and compared with a manual approach of coil selection. The gating signal offered an estimation of the breathing motion during measurement and was used as a reference to segment the acquired data into different breathing phases. Gradient delays and trajectory errors were corrected during post-processing using the Gradient Impulse Response Function. Iterative SENSE was then applied to determine the fully sampled data. In order to eliminate signal changes caused by motion, a 3D image registration was employed, and the results were compared to a 2D image registration method. Ventilation was assessed in 3D and regionally quantified by monitoring the signal changes in the lung parenchyma. Finally, image quality and quantitative ventilation values were compared to the standard 2D-SENCEFUL technique. 3D-UTE, combined with an automatic gating technique and SENCEFUL MRI, offered ventilation maps with high spatial resolution and SNR. Compared to the 2D method, UTE-SENCEFUL greatly improved the clinical quality of the structural images and the visualization of the lung parenchyma. Through-plane motion, partial volume effects and ventilation artifacts were also reduced with a three-dimensional method for image registration. UTE-SENCEFUL was also able to quantify regional ventilation and presented similar results to previous studies.},
  subject      = {Kernspintomografie},
  language  = {en}
}
@phdthesis{Raghuraman2020,
  author    = {Raghuraman, Sairamesh},
  title     = {New RF coil arrays for Static and Dynamic Musculoskeletal Magnetic Resonance Imaging},
  doi       = {10.25972/OPUS-20416},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-204165},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2020},
  abstract  = {Magnetic Resonance Imaging at field strengths up to 3 T, has become a default diagnostic modality for a variety of disorders and injuries, due to multiple reasons ranging from its non-invasive nature to the possibility of obtaining high resolution images of internal organs and soft tissues. Despite tremendous advances, MR imaging of certain anatomical regions and applications present specific challenges to be overcome. One such application is MR Musculo-Skeletal Imaging. This work addresses a few difficult areas within MSK imaging from the hardware perspective, with coil solutions for dynamic imaging of knee and high field imaging of hand. Starting with a brief introduction to MR physics, different types of RF coils are introduced in chapter 1, followed by sections on design of birdcage coils, phased arrays and their characterization in chapter 2. Measurements, calculations and simulations, done during the course of this work, have been added to this chapter to give a quantitative feel of the concepts explained. Chapter 3 deals with the construction of a phased array receiver for dynamic imaging of knee of a large animal model, i.e. minipig, at 1.5 T. Starting with details on the various aspects of an application that need to be considered when an MR RF array is designed, the chapter details the complex geometry of the region of interest in a minipig and reasons that necessitate a high density array. The sizes of the individual elements that constitute the array have been arrived at by studying the ratio of unloaded to loaded Q factors and choosing a size that provides the best ratio but still maintains a uniform SNR throughout the movement of the knee. To have a minimum weight and to allow mechanical movement of the knee, the Preamplifiers were located in a separate box. A movement device was constructed to achieve adjustable periodic movement of the knee of the anesthetized animal. The constructed array has been characterized for its SNR and compared with an existing product coil to show the improvement. The movement device was also characterized for its reproducibility. High resolution static images with anatomical details marked have been presented. The 1/g maps show the accelerations possible with the array. Snapshots of obtained dynamic images trace the cruciate ligaments through a cycle of movement of the animal's knee. The hardware combination of a high density phased array and a movement device designed for a minipig's knee was used as a 'reference' and extended in chapter 4 for a human knee. In principle the challenges are similar for dynamic imaging of a human knee with regards to optimization of the elements, the associated electronics and the construction of the movement device. The size of the elements were optimized considering the field penetration / sensitivity required for the internal tissues. They were distributed around the curvature of the knee keeping in mind the acceleration required for dynamic imaging and the direction of the movement. The constructed movement device allows a periodic motion of the lower half of the leg, with the knee placed within the coil, enabling visualization of the tissues inside, while the leg is in motion. Imaging has been performed using dynamic interleaved acquisition sequence where higher effective TR and flip angles are achieved due to a combination of interleaving and segmentation of the sequence. The movement device has been characterized for its reproducibility while the SNR distribution of the constructed RF array has been compared with that of a commercially available standard 8 channel array. The results show the improvement in SNR and acceleration with the constructed geometry. High resolution static images, dynamic snapshots and the 3D segmentation of the obtained images prove the usefulness of the complete package provided in the design, for performing dynamic imaging at a clinically relevant field strength. A simple study is performed in chapter 5 to understand the effects of changes in overlap for coil configurations with different loads and at different frequencies. The noise levels of individual channels and the correlation between them are plotted against subtle changes in overlap, at 64 and 123 MHz. SNR for every overlap setup is also measured and plotted. Results show that achieving critical overlap is crucial to obtain the best possible SNR in those coil setups where the load offered by the sample is low. Chapter 6 of the thesis work deals with coil design for high field imaging of hand and wrists at 7 T, with an aim to achieve ultra high resolution imaging. At this field strength due to the increase in dielectric effects and the resulting decrease in homogeneity, whole body transmit coils are impractical and this has led engineers to design local transmit coils, for specific anatomies. While transmit or transceive arrays are usually preferred, to mitigate SAR effects, the spatial resolution obtained is limited. It is shown that a solution to this, with regards to hand imaging, can be a single volume transmit coil, along with high density receive arrays optimized for different regions of the hand. The use of a phased array for reception provides an increased SNR / penetration under high resolution. A volume transmit coil could pose issues in homogeneity at 7 T, but the specific anatomy of hand and wrist, with comparatively less water content, limits dielectric effects to have homogeneous B_1+ profile over the hand. To this effect, a bandpass birdcage and a 12 channel receive array are designed and characterized. Images of very high spatial resolution (0.16 x 0.16 x 0.16 mm3) with internal tissues marked are presented. In vivo 1/g maps show that an acceleration of up to 3 is possible and the EM simulation results presented show the uniform field along with SAR hotspots in the hand. To reduce the stress created due to the 'superman' position of imaging, provisions in the form of a holder and a hand rest have been designed and presented. Factors that contributed to the stability of the presented design are also listed, which would help future designs of receive arrays at high field strengths. In conclusion, the coils and related hardware presented in this thesis address the following two aspects of MSK imaging: Dynamic imaging of knee and High resolution imaging of hand / wrist. The presented hardware addresses specific challenges and provides solutions. It is hoped that these designs are steps in the direction of improving the existing coils to get a better knowledge and understanding of MSK diseases such as Rheumatoid Arthritis and Osteoarthritis. The hardware can aid our study of ligament reconstruction and development. The high density array and transmit coil design for hand / wrist also demonstrates the benefits of the obtained SNR at 7 T while maintaining SAR within limits. This design is a contribution towards optimizing hardware at high field strength, to make it clinically acceptable and approved by regulatory bodies.},
  subject      = {MRI},
  language  = {en}
}
@phdthesis{Schindele2016,
  author    = {Schindele, Andreas},
  title     = {Proximal methods in medical image reconstruction and in nonsmooth optimal control of partial differential equations},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-136569},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2016},
  abstract  = {Proximal methods are iterative optimization techniques for functionals, J = J1 + J2, consisting of a differentiable part J2 and a possibly nondifferentiable part J1. In this thesis proximal methods for finite- and infinite-dimensional optimization problems are discussed. In finite dimensions, they solve l1- and TV-minimization problems that are effectively applied to image reconstruction in magnetic resonance imaging (MRI). Convergence of these methods in this setting is proved. The proposed proximal scheme is compared to a split proximal scheme and it achieves a better signal-to-noise ratio. In addition, an application that uses parallel imaging is presented. In infinite dimensions, these methods are discussed to solve nonsmooth linear and bilinear elliptic and parabolic optimal control problems. In particular, fast convergence of these methods is proved. Furthermore, for benchmarking purposes, truncated proximal schemes are compared to an inexact semismooth Newton method. Results of numerical experiments are presented to demonstrate the computational effectiveness of our proximal schemes that need less computation time than the semismooth Newton method in most cases. Results of numerical experiments are presented that successfully validate the theoretical estimates.},
  subject      = {Optimale Kontrolle},
  language  = {en}
}
@phdthesis{Seiberlich2008,
  author    = {Seiberlich, Nicole},
  title     = {Advances in Non-Cartesian Parallel Magnetic Resonance Imaging using the GRAPPA Operator},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-28321},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2008},
  abstract  = {Magnetic Resonance Imaging (MRI) is an imaging modality which provides anatomical or functional images of the human body with variable contrasts in an arbitrarily positioned slice without the need for ionizing radiation. In MRI, data are not acquired directly, but in the reciprocal image space (otherwise known as k-space) through the application of spatially variable magnetic field gradients. The k-space is made up of a grid of data points which are generally acquired in a line-by-line fashion (Cartesian imaging). After the acquisition, the k-space data are transformed into the image domain using the Fast Fourier Transformation (FFT). However, the acquisition of data is not limited to the rectilinear Cartesian sampling scheme described above. Non-Cartesian acquisitions, where the data are collected along exotic trajectories, such as radial and spiral, have been shown to be beneficial in a number of applications. However, despite their additional properties and potential advantages, working with non-Cartesian data can be complicated. The primary difficulty is that non-Cartesian trajectories are made up of points which do not fall on a Cartesian grid, and a simple and fast FFT algorithm cannot be employed to reconstruct images from non-Cartesian data. In order to create an image, the non-Cartesian data are generally resampled on a Cartesian grid, an operation known as gridding, before the FFT is performed. Another challenge for non-Cartesian imaging is the combination of unusual trajectories with parallel imaging. This thesis has presented several new non-Cartesian parallel imaging methods which simplify both gridding and the reconstruction of images from undersampled data. In Chapter 4, a novel approach which uses the concepts of parallel imaging to grid data sampled along a non-Cartesian trajectory called GRAPPA Operator Gridding (GROG) is described. GROG shifts any acquired k-space data point to its nearest Cartesian location, thereby converting non-Cartesian to Cartesian data. The only requirements for GROG are a multi-channel acquisition and a calibration dataset for the determination of the GROG weights. Chapter 5 discusses an extension of GRAPPA Operator Gridding, namely Self-Calibrating GRAPPA Operator Gridding (SC-GROG). SC-GROG is a method by which non-Cartesian data can be gridded using spatial information from a multi-channel coil array without the need for an additional calibration dataset, as required in standard GROG. Although GROG can be used to grid undersampled datasets, it is important to note that this method uses parallel imaging only for gridding, and not to reconstruct artifact-free images from undersampled data. Chapter 6 introduces a simple, novel method for performing modified Cartesian GRAPPA reconstructions on undersampled non-Cartesian k-space data gridded using GROG to arrive at a non-aliased image. Because the undersampled non-Cartesian data cannot be reconstructed using a single GRAPPA kernel, several Cartesian patterns are selected for the reconstruction. Finally, Chapter 7 discusses a novel method of using GROG to mimic the bunched phase encoding acquisition (BPE) scheme. In MRI, it is generally assumed that an artifact-free image can be reconstructed only from sampled points which fulfill the Nyquist criterion. However, the BPE reconstruction is based on the Generalized Sampling Theorem of Papoulis, which states that a continuous signal can be reconstructed from sampled points as long as the points are on average sampled at the Nyquist frequency. A novel method of generating the "bunched" data using GRAPPA Operator Gridding (GROG), which shifts datapoints by small distances in k-space using the GRAPPA Operator instead of employing zig-zag shaped gradients, is presented in this chapter. With the conjugate gradient reconstruction method, these additional "bunched" points can then be used to reconstruct an artifact-free image from undersampled data. This method is referred to as GROG-facilitated Bunched Phase Encoding, or GROG-BPE.},
  subject      = {NMR-Tomographie},
  language  = {en}
}
@phdthesis{Ziener2011,
  author    = {Ziener, Christian H.},
  title     = {Spindephasierung im Kroghschen Kapillarmodell des Myokards},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-73762},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2011},
  abstract  = {Der Zusammenhang zwischen den Parametern der Mikrostruktur des Myokards und der Spindephasierung wird hergestellt. Zur Beschreibung der Mikrostruktur des Myokards wurde das Kroghsche Kapillarmodell genutzt. In diesem Modell wird das Myokard auf eine einzige Kapillare reduziert, die von einem konzentrischen Gewebszylinder umgeben ist. In dem Gewebszylinder findet die Dephasierung und Diffusion statt. Mathematisch wird die Dephasierung durch die Bloch-Torrey-Gleichung beschrieben. Experimentell wurde der Signal-Zeit-Verlauf mittels einer PRESS-Sequenz und einer Gradienten-Echo-Sequenz gemessen. Mit den in dieser Arbeit vorgestellten Methoden ist der Zusammenhang zwischen Kapillarradius und Freien Induktionszerfall bekannt.},
  subject      = {Herzmuskel},
  language  = {de}
}