3D Whole-Brain MRSI at 7T

This grant proposal for the Erwin Schrödinger Fellowship of the Austrian Science Fund deals with high resolution whole-brain MR spectroscopic imaging (MRSI) at high magnetic fields like 7T. MRSI has great potential for detecting and differentiating brain diseases and psychiatric disorders. High quality data were achieved when combining the advantages of 7T with a high resolution, an acceleration method called parallel imaging, and other methodological improvements. Yet, the acceleration of parallel imaging is not sufficient for whole-brain MRSI at a high resolution. Furthermore, the main magnetic field at 7T is inhomogeneous in many brain regions relevant for brain diseases and psychiatric disorders, resulting in degraded data quality. Both problems prevent MRSI to be used extensively for clinical and neuroscientific studies. Therefore, we propose to use special coils for simultaneously acquiring signal and improving the homogeneity of the main magnetic field, and to use advanced acceleration methods such as "SPICE" and "compressed sensing". With that, high resolution whole-brain MRSI at 7T is within reach. This would be the first time that whole-brain MRSI measurements are possible at 7T. Together with the high resolution, an unprecedented high quality and impact on clinical studies is expected. With this grant proposal we want to answer if the coil proposed to be used increases the homogeneity of the main magnetic field enough to make MRSI possible in the whole brain. Further, which combination of the advanced acceleration methods provides the best results, and whether the best method provides enough acceleration for high resolution whole-brain MRSI at 7T. Finally, we want to test if whole-brain MRSI at 7T is feasible with this methodology. To answer these questions, the homogeneity of the main magnetic field will be assessed and compared between the proposed coil being enabled, and being disabled. The different acceleration methods will be compared using a multi-compartment phantom with known metabolic concentrations as gold standard, while a standard MRSI sequence with a long measurement time will be used as gold standard for in-vivo measurements. The project is proposed to be performed within a two-year stay at the MGH/HST Athinoula A. Martinos Center and a one-year return phase at the Medical University of Vienna in collaboration with Prof. Ovidiu Andronesi. By collaborations with Dr. Jason Stockmann, Dr. Chao Ma, and Prof. Elfar Adalsteinsson, a coil for simultaneously acquiring signal and homogenizing the main magnetic field at 7T, the SPICE reconstruction, and know-how and source code for compressed sensing is available at the Martinos Center. Thereby, the Martinos Center is the optimal institution to conduct the proposed studies.

Magnetic Resonance Spectroscopic Imaging (MRSI) is a method to investigate the concentrations of specific metabolites such as glutamate or creatine using an MR scanner. These metabolites often change in diseases such as tumours, or multiple sclerosis. MRSI is non-invasive, but high-resolution measurements of MRSI data in the brain are restricted by long measurement times (~30 minutes), and by inhomogeneities of the main magnetic field of the scanner. Therefore, the goal of this project was to develop an acceleration method for 3D MRSI and to use advanced hardware ("AC/DC coils") for homogenizing the main magnetic field of the MR scanner to enable high-resolution 3D MRSI of the whole brain at a field strength of 7 T. The acceleration was proposed in this project by two methods on top of the acceleration by using spiral trajectories for signal detection: One called "SPICE" which uses a mathematical model to partially split the spatial and the spectroscopic dimensions of the data and therefore improves the data quality while accelerating at the same time. The other method randomly omits data samples during acquisition in the spatial dimension, and uses specific prior knowledge (e.g. that MR signal does not change rapidly within the brain) to reconstruct the missing data. The AC/DC hardware uses the same MR coils for signal acquisition ("AC", this signal is a high-frequency AC current), and homogenizing the main magnetic field of the MR scanner. This is achieved by adding a direct current ("DC") to the coils, which causes a magnetic field, which in turn can be used to homogenize the main magnetic field. During the project it turned out that spiral trajectories for detecting the MRSI signal provided data of low quality, and therefore concentric ring trajectories were used instead. Also, the SPICE method did not provide high quality results, as the spatial maps of different metabolites were too similar, which was not the case for the fully sampled "gold standard" reconstruction. In contrast, the compressed sensing method showed good results for a 2D dataset up to accelerations of about 2.0-3.0. The homogenizing of the main magnetic field of the scanner using AC/DC coils worked very well, which was shown in volunteers to improve the data quality and the volume of quantifiable brain volume. In brain tumour patients this methodology was shown to improve the detection of the metabolite "D-2-hydroxyglutarate" so that its detection correlates better with the tumour tissue. The detection of this metabolite was shown to correlate with the outcome of the disease. In summary, this project may help to improve brain MRSI by making the acquisition faster, and the methodology more reliable. It may therefore become more useful, available and reliable in clinical routine or neuroscience.