UC Berkeley

Primary brain tumors are typically aggressive lesions that are difficult to treat and have a relatively poor prognosis for many patients. Magnetic resonance imaging (MRI) tools have been commonly utilized in the management of patients diagnosed with gliomas. Functional and metabolic MRI techniques have been proposed to add information regarding the tissue characteristics and biochemistry for better tumor localization, treatment planning and follow up of the disease.

Magnetic resonance spectroscopic imaging (MRSI) is a metabolic imaging technique used to analyze brain tissue chemistry. By showing metabolically active infiltrative tumor that can look similar to surrounding tissues on conventional MR images, MRSI allows for a more accurate definition of the extent of the disease. Despite those benefits, MRSI has not been widely used to care for patients with brain tumors.

Three major difficulties encountered in using MRSI in a clinical setting are limited cover- age, coarse spatial resolution and long data acquisition time. The possibility of combining compressed sensing and parallel imaging techniques to accelerate the acquisition time, in- crease coverage of the brain or resolution while ensuring high quality data without loss of information was the primary goal of this dissertation.

Several new techniques have been developed to accelerate MRSI acquisition in addition to reconstruct and optimally combine the accelerated multi-channel whole-brain 3D-MRSI data accurately and robustly. These techniques were validated on healthy volunteers and patients with brain tumors. They allowed the acquisition of MRSI data from a much larger brain volumes and finer spatial resolution than the conventional methods. The proposed AVD- GRAPPA technique reduced the 20min acquisition to < 5min with clinically interpretable spectra (high correlation to the full sampling). The improved coverage and spatial resolution will be useful for evaluating heterogeneous and infiltrative tumors, which are difficult to evaluate with current protocols. This advancement should make possible a more accurate assessment of the progression of tumors in serial studies.

The result of this dissertation suggests that magnetic resonance spectroscopic imaging is an important technique for spatially characterizing brain tumors that can be acquired in a shorter time to obtain equivalent disease related information. It is expected that shorter scan times will result in less patient discomfort, motion artifacts and will increase the scanner throughput and the usage of MRSI data in treatment management of the patients with brain tumor.