@phdthesis{Neumann2014,
  author    = {Neumann, Daniel},
  title     = {Advances in Fast MRI Experiments},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-108165},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2014},
  abstract  = {Magnetic Resonance Imaging (MRI) is a non-invasive medical imaging technique, that is rou- tinely used in clinical practice for detection and diagnosis of a wide range of different diseases. In MRI, no ionizing radiation is used, making even repeated application unproblematic. This is an important advantage over other common imaging methods such as X-rays and Computer To- mography. One major drawback of MRI, however, are long acquisition times and associated high costs of experiments. Since the introduction of MRI, several important technical developments have been made to successfully reduce acquisition times. In this work, novel approaches were developed to increase the efficiency of MRI acquisitions. In Chapter 4, an improved radial turbo spin-echo (TSE) combined acquisition and reconstruction strategy was introduced. Cartesian turbo spin-echo sequences [3] are widely used especially for the detection and diagnosis of neurological pathologies, as they provide high SNR images with both clinically important proton density and T2 contrasts. TSE acquisitions combined with radial sampling are very efficient, since it is possible to obtain a number of ETL images with different contrasts from a single radial TSE measurement [56-58]. Conventionally, images with a particular contrast are obtained from both radial and Cartesian TSE acquisitions by combining data from different echo times into a single image. In the radial case, this can be achieved by employing k-space weighted image contrast (KWIC) reconstruction. In KWIC, the center region of k-space is filled exclusively with data belonging to the desired contrast while outer regions also are assembled with data acquired at other echo times. However, this data sharing leads to mixed contrast contributions to both Cartesian and radial TSE images. This is true especially for proton density weighted images and therefore may reduce their diagnostic value. In the proposed method, an adapted golden angle reordering scheme is introduced for radial TSE acquisitions, that allows a free choice of the echo train length and provides high flexibility in image reconstruction. Unwanted contrast contaminations are greatly reduced by employing a narrow-band KWIC filter, that restricts data sharing to a small temporal window around the de- sired echo time. This corresponds to using fewer data than required for fully sampled images and consequently leads to images exhibiting aliasing artifacts. In a second step, aliasing-free images are obtained using parallel imaging. In the neurological examples presented, the CG-SENSE algorithm [42] was chosen due to its stable convergence properties and its ability to reconstruct arbitrarily sampled data. In simulations as well as in different in vivo neurological applications, no unwanted contrast contributions could be observed in radial TSE images reconstructed with the proposed method. Since this novel approach is easy to implement on today's scanners and requires low computational power, it might be valuable for the clinical breakthrough of radial TSE acquisitions. In Chapter 5, an auto-calibrating method was introduced to correct for stimulated echo contribu- tions to T2 estimates from a mono-exponential fit of multi spin-echo (MSE) data. Quantification of T2 is a useful tool in clinical routine for the detection and diagnosis of diseases as well as for tis- sue characterization. Due to technical imperfections, refocusing flip angles in a MSE acquisition deviate from the ideal value of 180○. This gives rise to significant stimulated echo contributions to the overall signal evolution. Therefore, T2 estimates obtained from MSE acquisitions typically are notably higher than the reference. To obtain accurate T2 estimates from MSE acquisitions, MSE signal amplitudes can be predicted using the extended phase graph (EPG, [23, 24]) algo- rithm. Subsequently, a correction factor can be obtained from the simulated EPG T2 value and applied to the MSE T2 estimates. However, EPG calculations require knowledge about refocus- ing pulse amplitudes, T2 and T1 values and the temporal spacing of subsequent echoes. While the echo spacing is known and, as shown in simulations, an approximate T1 value can be assumed for high ratios of T1/T2 without compromising accuracy of the results, the remaining two parameters are estimated from the data themselves. An estimate for the refocusing flip angle can be obtained from the signal intensity ratio of the second to the first echo using EPG. A conventional mono- exponential fit of the MSE data yields a first estimate for T2. The T2 correction is then obtained iteratively by updating the T2 value used for EPG calculations in each step. For all examples pre- sented, two iterations proved to be sufficient for convergence. In the proposed method, a mean flip angle is extracted across the slice. As shown in simulations, this assumption leads to greatly reduced deviations even for more inhomogeneous slice profiles. The accuracy of corrected T2 values was shown in experiments using a phantom consisting of bottles filled with liquids with a wide range of different T2 values. While T2 MSE estimates were shown to deviate significantly from the spin-echo reference values, this is not the case for corrected T2 values. Furthermore, applicability was demonstrated for in vivo neurological experiments. In Chapter 6, a new auto-calibrating parallel imaging method called iterative GROG was pre- sented for the reconstruction of non-Cartesian data. A wide range of different non-Cartesian schemes have been proposed for data acquisition in MRI, that present various advantages over conventional Cartesian sampling such as faster acquisitions, improved dynamic imaging and in- trinsic motion correction. However, one drawback of non-Cartesian data is the more complicated reconstruction, which is ever more problematic for non-Cartesian parallel imaging techniques. Iterative GROG uses Calibrationless Parallel Imaging by Structured Low-Rank Matrix Completion (CPI) for data reconstruction. Since CPI requires points on a Cartesian grid, it cannot be used to directly reconstruct non-Cartesian data. Instead, Grappa Operator Gridding (GROG) is employed in a first step to move the non-Cartesian points to the nearest Cartesian grid locations. However, GROG requires a fully sampled center region of k-space for calibration. Combining both methods in an iterative scheme, accurate GROG weights can be obtained even from highly undersampled non-Cartesian data. Subsequently, CPI can be used to reconstruct either full k- space or a calibration area of arbitrary size, which can then be employed for data reconstruction with conventional parallel imaging methods. In Chapter 7, a new 2D sampling scheme was introduced consisting of multiple oscillating effi- cient trajectories (MOET), that is optimized for Compressed Sensing (CS) reconstructions. For successful CS reconstruction of a particular data set, some requirements have to be met. First, ev- ery data sample has to carry information about the whole object, which is automatically fulfilled for the Fourier sampling employed in MRI. Additionally, the image to be reconstructed has to be sparse in an arbitrary domain, which is true for a number of different applications. Last, data sam- pling has to be performed in an incoherent fashion. For 2D imaging, this important requirement of CS is difficult to achieve with conventional Cartesian and non-Cartesian sampling schemes. Ra- dial sampling is often used for CS reconstructions of dynamic data despite the streaking present in undersampled images. To obtain incoherent aliasing artifacts in undersampled images while at the same time preserving the advantages of radial sampling for dynamic imaging, MOET com- bines radial spokes with oscillating gradients of varying amplitude and alternating orientation orthogonal to the readout direction. The advantage of MOET over radial sampling in CS re- constructions was demonstrated in simulations and in in vivo cardiac imaging. MOET provides superior results especially when used in CS reconstructions with a sparsity constraint directly in image space. Here, accurate results could be obtained even from few MOET projections, while the coherent streaking artifacts present in the case of radial sampling prevent image recovery even for smaller acceleration factors. For CS reconstructions of dynamic data with sparsity constraint in xf-space, the advantage of MOET is smaller since the temporal reordering is responsible for an important part of incoherency. However, as was shown in simulations of a moving phantom and in the reconstruction of ungated cardiac data, the additional spatial incoherency provided by MOET still leads to improved results with higher accuracy and may allow reconstructions with higher acceleration factors.},
  subject      = {Kernspintomografie},
  language  = {en}
}
@phdthesis{Eirich2022,
  author    = {Eirich, Philipp},
  title     = {Accelerated non-Cartesian cardiovascular MR Imaging at 3T and 7T},
  doi       = {10.25972/OPUS-25397},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-253974},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2022},
  abstract  = {In this work, accelerated non-Cartesian Magnetic Resonance Imaging (MRI) methods were established and applied to cardiovascular imaging (CMR) at different magnetic field strengths (3T and 7T). To enable rapid data acquisition, highly efficient spiral k-space trajectories were created. In addition, hybrid sampling patterns such as the twisting radial lines (TWIRL) k-space trajectory were studied. Imperfections of the dynamic gradient system of a MR scanner result in k-space sampling errors. Ultimately, these errors can lead to image artifacts in non-Cartesian acquisitions. Among other reasons such as an increased reconstruction complexity, they cause the lack of spiral sequences in clinical routine compared to standard Cartesian imaging. Therefore, the Gradient System Transfer Functions (GSTFs) of both scanners were determined and used for k-space trajectory correction in post-correction as well as in terms of a pre-emphasis. The GSTF pre-emphasis was implemented as a fully automatic procedure, which enabled a precise correction of arbitrary gradient waveforms for double-oblique slice orientations. Consequently, artifacts due to trajectory errors could be mitigated, which resulted in high image quality in non-Cartesian MRI. Additionally, the GSTF correction was validated by measuring pre-emphasized spiral gradient outputs, which showed high agreement with the theoretical gradient waveforms. Furthermore, it could be demonstrated that the performance of the GSTF correction is superior to a simple delay compensation approach. The developed pulse sequences were applied to gated as well as real-time CMR. Special focus lied on the implementation of a spiral imaging protocol to resolve the beating heart of animals and humans in real time and free breathing. In order to achieve real-time CMR with high spatiotemporal resolution, k-space undersampling was performed. For this reason, efficient sampling strategies were developed with the aim to facilitate compressed sensing (CS) during image reconstruction. The applied CS approach successfully removed aliasing artifacts and yielded high-resolution cardiac image series. Image reconstruction was performed offline in all cases such that the images were not available immediately after acquisition at the scanner. Spiral real-time CMR could be performed in free breathing, which led to an acquisition time of less than 1 minute for a whole short-axis stack. At 3T, the results were compared to the gold standard of electrocardiogram-gated Cartesian CMR in breath hold, which revealed similar values for important cardiovascular functional and volumetric parameters. This paves the way to an application of the developed framework in clinical routine of CMR. In addition, the spiral real-time protocol was transferred to swallowing and speech imaging at 3T, and first images were presented. The results were of high quality and confirm the straightforward utilization of the spiral sequence in other fields of MRI. In general, the GSTF correction yielded high-quality images at both field strengths, 3T and 7T. Off-resonance related blurring was mitigated by applying non-Cartesian readout gradients of short duration. At 7T, however, B1-inhomogeneity led to image artifacts in some cases. All in all, this work demonstrated great advances in accelerating the MRI process by combining efficient, undersampled non-Cartesian k-space coverage with CS reconstruction. Trajectory correction using the GSTF can be implemented at any scanner model and enables non-Cartesian imaging with high image quality. Especially MRI of dynamic processes greatly benefits from the presented rapid imaging approaches.},
  subject      = {Kernspintomografie},
  language  = {en}
}
@phdthesis{Portmann2023,
  author    = {Portmann, Johannes},
  title     = {Accelerated inversion recovery MRI of the myocardium using spiral acquisition},
  doi       = {10.25972/OPUS-30282},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-302822},
  school      = {Universit{\"a}t W{\"u}rzburg},
  year      = {2023},
  abstract  = {This work deals with the acceleration of cardiovascular MRI for the assessment of functional information in steady-state contrast and for viability assessment during the inversion recovery of the magnetization. Two approaches are introduced and discussed in detail. MOCO-MAP uses an exponential model to recover dynamic image data, IR-CRISPI, with its low-rank plus sparse reconstruction, is related to compressed sensing. MOCO-MAP is a successor to model-based acceleration of parametermapping (MAP) for the application in the myocardial region. To this end, it was augmented with a motion correction (MOCO) step to allow exponential fitting the signal of a still object in temporal direction. Iteratively, this introduction of prior physical knowledge together with the enforcement of consistency with the measured data can be used to reconstruct an image series from distinctly shorter sampling time than the standard exam (< 3 s opposed to about 10 s). Results show feasibility of the method as well as detectability of delayed enhancement in the myocardium, but also significant discrepancies when imaging cardiac function and artifacts caused already by minor inaccuracy of the motion correction. IR-CRISPI was developed from CRISPI, which is a real-time protocol specifically designed for functional evaluation of image data in steady-state contrast. With a reconstruction based on the separate calculation of low-rank and sparse part, it employs a softer constraint than the strict exponential model, which was possible due to sufficient temporal sampling density via spiral acquisition. The low-rank plus sparse reconstruction is fit for the use on dynamic and on inversion recovery data. Thus, motion correction is rendered unnecessary with it. IR-CRISPI was equipped with noise suppression via spatial wavelet filtering. A study comprising 10 patients with cardiac disease show medical applicability. A comparison with performed traditional reference exams offer insight into diagnostic benefits. Especially regarding patients with difficulty to hold their breath, the real-time manner of the IR-CRISPI acquisition provides a valuable alternative and an increase in robustness. In conclusion, especially with IR-CRISPI in free breathing, a major acceleration of the cardiovascular MR exam could be realized. In an acquisition of less than 100 s, it not only includes the information of two traditional protocols (cine and LGE), which take up more than 9.6 min, but also allows adjustment of TI in retrospect and yields lower artifact level with similar image quality.},
  subject      = {Kernspintomografie},
  language  = {en}
}
@phdthesis{Sturm2015,
  author    = {Sturm, Volker J{\"o}rg Friedrich},
  title     = {\(^{19}F\) Magnetresonanztomographie zur Bildgebung von Infektionen im Zeitverlauf},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-122851},
  school      = {Universit{\"a}t W{\"u}rzburg},
  pages     = {114},
  year      = {2015},
  abstract  = {Im Rahmen dieser Arbeit sollten die M{\"o}glichkeiten der MR Tomographie erkundet werden bakterielle Infektionen im Zeitverlauf darzustellen. Genauer gesagt sollte das Potential der MR Tomographie anhand eines durch eine Infektion induzierten lokalisierten Abszesses unter Verwendung dreier unterschiedlicher MRT Methoden untersucht werden: Mittels nativem \(T_2\) Kontrast; der Verwendung von superparamagnetischen Eisenoxid Partieln (USPIO) als \(T_2^*\) Kontrastmittel; und dem Einsatz von Perfluorkarbonen (PFC) als \(^{19}F\) MRT Marker (siehe Kapitel 3). Wie erwartet f{\"u}hrte die durch die Infektion hervorgerufene Entz{\"u}ndung zu ver{\"a}nderten \(T_2\)-Zeiten, welche auf \(T_2\)-gewichteten MR Bildern eine Lokalisierung des Abszessbereiches erlauben. Jedoch eigneten sich diese Daten aufgrund der graduellen {\"A}nderung der \(T_2\)-Zeiten nicht, um eine klare Grenze zwischen Abszess und umliegendem Gewebe zu ziehen. Superparamagnetische Eisenoxidpartikel andererseit haben als MRT Kontrastmittel bereits in den letzten Jahren ihre F{\"a}higkeit unter Beweis gestellt Entz{\"u}ndungen [53, 58, 64] darzustellen. Die Anreicherung dieser Partikel am Rande des Abszesses [53], wie sie auch in unseren MR Daten zu beobachten war, erlaubte eine relativ scharfe Abgrenzung gegen{\"u}ber dem umgebenden Gewebe in der chronischen Phase der Infektion (Tag 9 p.i.). Hingegen gen{\"u}gte die nur sehr sp{\"a}rlichen Anreicherung von USPIO Partikeln in der akuten Phase der Infektion (Tag 3 p.i.) nicht f{\"u}r eine entsprechende Abgrenzung [58]. Aufgrund der sehr geringen biologischen H{\"a}ufigkeit und den sehr kurzen Relaxationszeiten von endogenem Fluor eignen sich Perfluorkarbone als Markersubstanz in der MR Tomographie von biologischen Systemen. Insbesondere da PFC Emulsionen durch phagozytierende Zellen aufgenommen werden und im Bereich von Entz{\"u}ndungen akkumulieren [30, 59]. In dieser Arbeit konnte anhand der erhaltenen MRT Daten eine Akkumulation von Perfluorkarbonen nicht nur in der chronischen Phase, sondern auch in der akuten Phase nachgewiesen werden. Diese Daten erlauben somit zu allen untersuchten Zeitpunkten eine Abgrenzung zwischen Infektion und umliegenden Gewebe. Aufgrund der besagten Vorteile wurden die Perfluorkarbone gew{\"a}hlt, um die M{\"o}glichkeiten der MR Tomographie zu testen, quantitative Informationen {\"u}ber die schwere der Infektion zu liefern. Als Referenz f{\"u}r die Bakterienbelastung wurden die Biolumineszenzbildgebung (BLI) [49, 50] und die Standardmethode zur Bestimmung der Bakterienbelastung cfu (koloniebildenden Einheiten) herangezogen. Eine Gegen{\"u}berstellung der zeitlichen Verl{\"a}ufe der durch die Biolumineszenzbildgebung und durch die cfu erhaltenen Daten liefert eine qualitative {\"U}bereinstimmung mit den durch die 19F MR Tomographie erhaltenen Daten. Dies trifft hierbei sowohl auf die {\"u}ber den gesamten Infektionsbereich hinweg summierten Signalamplituden, als auch auf das Volumen zu, in dem Fluor am Ort der Infektion akkumuliert wurde. Im Gegensatz zur Methode der cfu Bestimmung sind die MR Tomographie und die Biolumineszenzbildgebung nicht invasiv und erlauben die Verfolgung des Infektionsverlaufes an einem einzelnen Individuum. Hierzu ben{\"o}tigt, im Gegensatz zur MR Tomographie, die Methode der Biolumineszenzbildgebung jedoch einen speziellen Pathogenstamm. Dar{\"u}ber hinaus ist hervorzuheben, dass die MR Tomographie zudem die M{\"o}glichkeit bietet auch morphologische Informationen {\"u}ber den Infektionsbereich und seine Umgebung zu akquirieren. Gerade weil jede dieser Methoden die mit der Infektion einhergehenden Prozesse aus einer leicht anderen Blickrichtung betrachtet, erscheint es sinnvoll diese etablierte Untersuchungsplattform bestehend aus MRT, BLI und cfu {\"u}ber die in dieser Arbeit bearbeitete Fragestellung hinaus n{\"a}her zu untersuchen. Insbesondere der Aspekt inwieweit die drei Methoden sich gegenseitig erg{\"a}nzen, k{\"o}nnte einen tieferen Einblick in die Wechselwirkung zwischen Pathogen und Wirt erlauben. Auch wenn f{\"u}r die betrachtete Fragestellung bereits der hierdurchgef{\"u}hrte semiquanitative Ansatz zur Bestimmung der relativen Fluormengen am Ort der Infektion ausreichte, so ist doch im Allgemeinen w{\"u}nschenswert probenbezogen die Sensitivit{\"a}t der Spule und damit die G{\"u}te der Spulenabstimmung zu bestimmen. Hierzu ist jedoch die Aufnahme von \(B_1\)-Karten unabdingbar und wird entsprechend im Kapitel 4 \(Bloch-Siegert B_1^+-Mapping\) n{\"a}her addressiert. Der Schwerpunkt liegt hierbei, wie der Kapitelname bereits andeutet, auf der Bloch-Siegert Methode, die insbesondere in der pr{\"a}sentierten Implementierung in einer Turbo/ Multi Spin Echo Sequenz eine effiziente Nutzung der relativ langen \(T_\)2-Zeiten der Perfluorkarbone erlaubt. Da zudem die Bloch-Siegert-Methode eine rein phasenbasierte Methode ist, kann neben der aus den Daten erzeugten \(B_1\)-Karte zugleich ein unverf{\"a}lschtes Magnitudenbild generiert werden, wodurch eine sehr effiziente Nutzung der vorhandenen Messzeit erm{\"o}glicht wird. Diese Eigenschaft ist insbesondere f{\"u}r \(^{19}F\) Bildgebung von besonderem Interesse, da hier f{\"u}r jede Messung, aufgrund der {\"u}blicherweise relativ geringen Konzentration an Fluoratomen, lange Messzeiten ben{\"o}tigt werden. Zusammenfassend konnte anhand des untersuchten Tiermodells sowohl die F{\"a}higkeit der MR Tomographie nachgewiesen werden Infektionen im Zeitverlauf darzustellen, als auch die F{\"a}higkeit der MR Tomographie quantitative Informationen {\"u}ber den Verlauf der Infektion zu liefern. Desweiteren konnte eine M{\"o}glichkeit aufgezeigt werden, welche das Potential hat in vertretbarem Zeitrahmen auch in vivo B1+-Karten auf dem Fluorkanal zu erstellen und so einen zentralen Unsicherheitsfaktor, f{\"u}r Relaxometry und absolute Quantifizierung von \(^{19}F\) Daten in vivo, zu beseitigen.},
  subject      = {Kernspintomografie},
  language  = {de}
}