TY - THES A1 - Breuer, Felix T1 - Development and Applications of Efficient Strategies for Parallel Magnetic Resonance Imaging T1 - Entwicklung und Anwendungen von effizienten Strategien in der Parallelen Magnetresonanztompgraphie N2 - Virtually all existing MRI applications require both a high spatial and high temporal resolution for optimum detection and classification of the state of disease. The main strategy to meet the increasing demands of advanced diagnostic imaging applications has been the steady improvement of gradient systems, which provide increased gradient strengths and faster switching times. Rapid imaging techniques and the advances in gradient performance have significantly reduced acquisition times from about an hour to several minutes or seconds. In order to further increase imaging speed, much higher gradient strengths and much faster switching times are required which are technically challenging to provide. In addition to significant hardware costs, peripheral neuro-stimulations and the surpassing of admissable acoustic noise levels may occur. Today’s whole body gradient systems already operate just below the allowed safety levels. For these reasons, alternative strategies are needed to bypass these limitations. The greatest progress in further increasing imaging speed has been the development of multi-coil arrays and the advent of partially parallel acquisition (PPA) techniques in the late 1990’s. Within the last years, parallel imaging methods have become commercially available,and are therefore ready for broad clinical use. The basic feature of parallel imaging is a scan time reduction, applicable to nearly any available MRI method, while maintaining the contrast behavior without requiring higher gradient system performance. PPA operates by allowing an array of receiver surface coils, positioned around the object under investigation, to partially replace time-consuming spatial encoding which normally is performed by switching magnetic field gradients. Using this strategy, spatial resolution can be improved given a specific imaging time, or scan times can be reduced at a given spatial resolution. Furthermore, in some cases, PPA can even be used to reduce image artifacts. Unfortunately, parallel imaging is associated with a loss in signal-to-noise ratio (SNR) and therefore is limited to applications which do not already operate at the SNR limit. An additional limitation is the fact that the coil array must provide sufficient sensitivity variations throughout the object under investigation in order to offer enough spatial encoding capacity. This doctoral thesis exhibits an overview of my research on the topic of efficient parallel imaging strategies. Based on existing parallel acquisition and reconstruction strategies, such as SENSE and GRAPPA, new concepts have been developed and transferred to potential clinical applications. N2 - In den späten 80er Jahren entwickelte sich die Magnetresonanz-Tomographie (MRT), die bis dato lediglich in Forschungseinrichtungen etabliert war, zu einem der wichtigsten Verfahren in der klinischen Diagnostik. Allerdings erfordern nahezu alle bestehenden klinischen Anwendungsgebiete sowohl eine hohe räumliche als auch eine hohe zeitliche Auflösung für eine optimale Detektion und Klassifizierung von Krankheitsbildern. Der bisherige Ansatz, diesen zunehmenden Anforderungen an die klinische MRT gerecht zu werden, bestand vor allem in der stetigen Verbesserung von Gradientensystemen die mit immer höheren Gradientenstärken und schnelleren Schaltzeiten aufwarteten. Die technischen Fortschritte, sowie schnelle Bildgebungsmethoden erlaubten es, Messzeiten von etwa einer Stunde auf nur wenige Minuten oder sogar Sekunden zu reduzieren. Eine weitere Verkürzung der Experimentdauer mittels noch leistungsfähigeren Gradientensystemen ist jedoch technisch schwierig zu realisieren. Ausserdem gehen enorm hohe Entwicklungs und Materialkosten mit den erhöhten Anforderungen einher. Es kommt hinzu, dass noch stärkere Gradienten und noch schnellere Schaltzeiten zu peripheren Neurostimulationen und zur Überschreitung von zulässigen akustischen Grenzwerten führen können. Heutige Gradientensysteme arbeiten schon sehr nahe an den Grenzen der zulässigen Sicherheitsbestimmungen. Deshalb werden alternative Strategien benötigt, um weitere Messzeitverkürzungen realisieren zu können. Der bisher erfolgreichste Ansatz bestand in der Entwicklung von Mehr-Kanal-Spulen-Anordnungen und damit verknüpft der darauffolgenden Einführung der parallellen Bildgebung in den späten 90er Jahren. In den letzten 5 Jahren haben sich parallele Bildgebungsmethoden an den klinischen Tomographen etabliert und nahezu alle Herstellerfirmen stellen diese Technik kommerziell zur Verfügung. Die parallele Bildgebung ermöglicht eine Messzeitverkürzung, die prinzipiell auf jede bestehende Bildgebungsmethode angewendet werden kann, ohne dabei das Kontrastverhalten zu verändern und ohne höhere Gradientenleistung zu beanspruchen. In der parallellen Bildgebung übernimmt die Mehr-Kanal-Spulen-Anordnung teilwiese die Ortskodierung, die normalerweise durch zeitaufwendiges Schalten von Magnetfelgradienten erzeugt wird. Mit dieser Strategie kann bei gleicher Messzeit die örtliche Auflösung verbessert, oder bei gleicher Auflösung die Messzeit verkürzt werden. Ausserdem können mit hilfe der parallelen MRT in manchen Fällen Bildartefakte signifikant reduziert werden. Allerdings ist mit der parallelen Bildgebung immer ein Signal zu Rausch (SNR) Verlust verbunden, der diese Methode auf klinische Anwendungen begrenzt, die nicht bereits am SNR-Limit betrieben werden. Ausserdem muß die Spulenanordnung genug Sensitivitätsvariationen über das zu untersuchende Objekt bereitstellen, um ausreichende Kodierfunktion zu gewährleisten. Diese Dissertationsarbeit liefert einen Überblick über meine Forschungsarbeit zum Thema “Entwicklung und Anwendung von effizienten Strategien in der parallelen MRT”. Basierend auf bestehenden parallelen Akquisitions und Rekonstruktionstechniken, wie beispielsweise SENSE und GRAPPA, wurden neue Konzepte entwickelt und auf mögliche klinische Fragestellungen angewandt. KW - NMR-Bildgebung KW - Paralleler Prozess KW - Parallele Bildgebung KW - SENSE KW - GRAPPA KW - TGRAPPA KW - CAIPIRINHA KW - dynamische Bildgebung KW - Parallel imaging KW - SENSE KW - GRAPPA KW - TGRAPPA KW - CAIPIRINHA KW - dynamic imaging Y1 - 2006 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-20683 ER - TY - JOUR A1 - Gerdes, Antje B. M. A1 - Wieser, Matthias J. A1 - Mühlberger, Andreas A1 - Weyers, Peter A1 - Alpers, Georg W. A1 - Plichta, Michael M. A1 - Breuer, Felix A1 - Pauli, Paul T1 - Brain activations to emotional pictures are differentially associated with valence and arousal ratings N2 - Several studies have investigated the neural responses triggered by emotional pictures, but the specificity of the involved structures such as the amygdala or the ventral striatum is still under debate. Furthermore, only few studies examined the association of stimuli’s valence and arousal and the underlying brain responses. Therefore, we investigated brain responses with functional magnetic resonance imaging of 17 healthy participants to pleasant and unpleasant affective pictures and afterwards assessed ratings of valence and arousal. As expected, unpleasant pictures strongly activated the right and left amygdala, the right hippocampus, and the medial occipital lobe, whereas pleasant pictures elicited significant activations in left occipital regions, and in parts of the medial temporal lobe. The direct comparison of unpleasant and pleasant pictures, which were comparable in arousal clearly indicated stronger amygdala activation in response to the unpleasant pictures. Most important, correlational analyses revealed on the one hand that the arousal of unpleasant pictures was significantly associated with activations in the right amygdala and the left caudate body. On the other hand, valence of pleasant pictures was significantly correlated with activations in the right caudate head, extending to the nucleus accumbens (NAcc) and the left dorsolateral prefrontal cortex. These findings support the notion that the amygdala is primarily involved in processing of unpleasant stimuli, particularly to more arousing unpleasant stimuli. Reward-related structures like the caudate and NAcc primarily respond to pleasant stimuli, the stronger the more positive the valence of these stimuli is. KW - Psychologie KW - emotional pictures KW - amygdala KW - caudate KW - valence KW - arousal Y1 - 2010 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-68153 ER - TY - JOUR A1 - Dawood, Peter A1 - Breuer, Felix A1 - Stebani, Jannik A1 - Burd, Paul A1 - Homolya, István A1 - Oberberger, Johannes A1 - Jakob, Peter M. A1 - Blaimer, Martin T1 - Iterative training of robust k‐space interpolation networks for improved image reconstruction with limited scan specific training samples JF - Magnetic Resonance in Medicine N2 - To evaluate an iterative learning approach for enhanced performance of robust artificial‐neural‐networks for k‐space interpolation (RAKI), when only a limited amount of training data (auto‐calibration signals [ACS]) are available for accelerated standard 2D imaging. Methods In a first step, the RAKI model was tailored for the case of limited training data amount. In the iterative learning approach (termed iterative RAKI [iRAKI]), the tailored RAKI model is initially trained using original and augmented ACS obtained from a linear parallel imaging reconstruction. Subsequently, the RAKI convolution filters are refined iteratively using original and augmented ACS extracted from the previous RAKI reconstruction. Evaluation was carried out on 200 retrospectively undersampled in vivo datasets from the fastMRI neuro database with different contrast settings. Results For limited training data (18 and 22 ACS lines for R = 4 and R = 5, respectively), iRAKI outperforms standard RAKI by reducing residual artifacts and yields better noise suppression when compared to standard parallel imaging, underlined by quantitative reconstruction quality metrics. Additionally, iRAKI shows better performance than both GRAPPA and standard RAKI in case of pre‐scan calibration with varying contrast between training‐ and undersampled data. Conclusion RAKI benefits from the iterative learning approach, which preserves the noise suppression feature, but requires less original training data for the accurate reconstruction of standard 2D images thereby improving net acceleration. KW - complex‐valued machine learning KW - data augmentation KW - deep learning KW - GRAPPA KW - parallel imaging KW - RAKI Y1 - 2023 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:20-opus-312306 VL - 89 IS - 2 SP - 812 EP - 827 ER -