Optimized Molecular Imaging through Magnetic Resonance for Improved Target Definition in Radiation Oncology

Abstract

Magnetic resonance spectroscopy (MRS) and spectroscopic imaging (MRSI) are a key modality in radiation oncology for brain and prostate tumors. Improved target definition for radiation therapy (RT) and distinction of changes due to RT from tumor recurrence have been greatly aided by MRSI. However, current applications of MRS/MRSI have limitations due to mainly the fast Fourier transform (FFT) and noise. Optimization of MRS/MRSI is possible by more advanced signal processing via the fast Padé transform (FPT). As a quotient of two polynomials, the FPT markedly improves the resolution of in vivo MR time signals encoded from the brain and reliably reconstructs all spectral parameters of metabolites. Due to high spectral density with numerous multiplet resonances, MRS/MRSI of the prostate is exceedingly difficult. The FPT applied to MRS data as encoded from normal and malignant prostate resolves all the genuine resonances, including multiplets and closely overlapping peaks. With synthesized time signals, the FPT fully retrieves all the input spectral parameters with machine accuracy. Such super-resolution is achieved without fitting or numerical integration of peak areas, thereby yielding the most accurate metabolite concentrations. This needs only short signal lengths that imply improved signal-to-noise ratios. These ratios are further enhanced by eliminating “noisy” Froissart doublets as confluent pole-zero pairs. Hence, only the true information is reconstructed by the FPT, as the prerequisite for clinically meaningful interpretations of in vivo time signals. With these long sought capabilities of advanced Padé-based signal processing, MRS and MRSI are poised to reach their full potential in radiation oncology.