Noise Performance of IDEAL considering Relaxation and Multiple Spectral Peaks of Fat

INTRODUCTION Water/fat resolved MR imaging is gaining considerable interest in a number of clinical applications. Chemical shift encoding-based imaging approaches, like three-point Dixon [1] or IDEAL [2], provide high quality water/fat separation. Both approaches treat the spectral composition of water and fat as idealized single peak spectra with infinite signal lifetime. However, in reality, relaxation effects are always present. Therefore, Yu et al. [3] proposed to incorporate a common T2 for water and fat into the signal model (T2-IDEAL). Moreover, the fat spectrum is much more complicated. Human fatty tissue contains methyl, methylene and olefinic methine protons in different concentrations and compositions [4,5]. Recently, Brodsky et al. [6] proposed a separation algorithm that takes the multi-peak nature of fat into account. The fat spectrum in this multi-peak (MP)-IDEAL approach is considered to be spatially invariant and normalized. In this paper, these refinements to the original IDEAL approach are studied with respect to model limitations and noise performance. METHODS AND MATERIALS