Optimized High-Resolution Fast Magnetic Resonance Fingerprinting with Cloud- Based Reconstruction Abstract Magnetic resonance imaging (MRI), despite its wide utility, is inherently limited due to its inability to measure tissue properties quantitatively, which is critical for objective and scanner-independent diagnosis and treatment monitoring. MR Fingerprinting (MRF) is a relatively new quantitative MRI framework for simultaneous quantification of multiple tissue properties. While MRF outperforms most conventional methods in quantitative imaging, existing MRF techniques are still handicapped by limited spatial resolution and coverage, long acquisition times, suboptimal acquisition parameters, long data reconstruction times, and complicated post- processing workflows, hindering large-scale clinical validation and translation. In this project, we will leverage the expertise of our team in MRF, machine learning, and pulse sequence optimization to develop and optimize a rapid and robust quantitative MR technique, applicable to high-resolution volumetric brain imaging. Our team has recently developed a new B1-insensitive MRF method using low flip angles and multiple magnetization preparations for improved accuracy and precision in tissue quantification compared with existing MRF methods. We will first develop and optimize this new MRF method for 3D high- resolution brain imaging, using our newly developed pulse sequence design framework. Novel fat navigator will be incorporated to improve motion robustness (Aim 1). We will leverage state-of-the-art deep learning techniques to accelerate both acquisition and post-processing (Aim 2). Finally, a complete MRF post-processing pipeline empowered by GPU cloud computing will be developed to significantly simplify the post-processing workflow and facilitate efficient clinical translation and validation of the proposed methods for patients with neurological diseases (Aim 3).