Magnetic resonance spectroscopic imaging (MRSI) has been recognized as a powerful tool for noninvasive imaging of metabolites of the body. However its research and clinical applications  have been limited due to the long data acquisition time, poor spatial resolution, and low signal-to-noise ratio (SNR) .  We have recently developed a subspace-based approach, called SPectroscopic Imaging by exploiting spatiospectral CorrElation (SPICE), to high-resolution MRSI. SPICE exploits the spatiospectral correlations of high-dimensional MRSI data and adopts special data acquisition and reconstruction schemes to recover high-resolution, high-SNR, MRSI images from sparsely sampled data. In this talk, I will give a quick review of MRS followed by the basic principle of SPICE. I will show MRSI results from phantoms and healthy subjects to demonstrate the feasibility and capability of SPICE. I will also show preliminary results on using SPICE and PET to guide radiation therapy in gliomas. At last, I will show how the underlying mathematical model and imaging principle of SPICE can be applied to dynamic MR.