Performance of chemical shift-based water-fat separation with self-calibrated fat spectrum is sensitive to echo times.

Chemical shift-based water-fat separation method utilises water-fat resonance frequency difference to decompose signals into water and fat partitions in magnetic resonance imaging (MRI) on a pixel-wise basis. It provides an effective way to measure fat fraction, or to suppress fat signal which might obscure underlying pathology. IDEAL (Iterative decomposition of water and fat with echo asymmetry and least-squares estimation) algorithm with multi-peak fat spectral modelling has been developed. Recent studies have discussed the performance of this algorithm assuming that the frequencies and relative amplitudes of fat peaks are constant among all subjects. However, the fat spectra vary in different tissues, thus a self-calibration method which estimates the fat spectrum directly from the data provides more accurate results. In this work, we analyse the performance of multi-peak IDEAL algorithm with self-calibrated fat spectrum by theoretical calculation, simulation, and experiments, and find optimal echo time increments which provide reliable water-fat separation.