MRSI is a noninvasive tool for measuring the spatial distribution of important metabolites in human tissue. MRSI signals are quite weak, so a lengthy data gathering period is required to obtain adequate signal levels from each measurement point. To reduce the total duration of the MRSI session, clinicians usually limit the number of MRSI measurement points. When these limited data are processed, the resulting image lacks contrast and exhibits poor spatial resolution. This poor image quality is believed to be caused by the processing algorithms employed in existing equipment. the proposed effort would develop a different MRSI processing algorithm to improve image quality. Specifically, MRSI sampled data currently are processed using the FFT, which is a computationally efficient method of producing an image. However, the underlying assumptions required by the FFT are not satisfied in MRSI. In particular, MRSI data form frequency-limited records of a spatially finite tissue. For this situation, a different discrete transform based on Prolate spheroidal Wave Functions (PSWF) is required. PSWF requires more computation than the FFT, but its superior accuracy permits greatly improved resolution from a limited number of measurements. PSWF image transforms for MRSI will lead to robust, automated image processing.
