I hereby consent to the publication of this abstract under the CC-BY license Comparative Review of Algorithms and Methods for proton-density fat fraction (PDFF) quantification

Methods: An open-source toolbox available both in Python and Matlab enabled to numerically compare recent Fat-Water separation algorithms[3–10] (Table 1). Additionally, an extensible human fat spectra[11–13] library was implemented to be plugged with each algorithm. Synthetic CSE-MRI volumes were modeled with PDFF=0-100%, B0=-300:6:300Hz, 100 repetitions, gaussian noise (SNR=50:10:100) and T2*=20ms. Different numbers of echo times (NTE=3,5,7,9) and schemes (realistic minimal, IDEAL and in/out-of-phase) were considered. The Bias, precision and limits of agreement (LOA) were evaluated for PDFF, B0 and T2*. Algorithms’ discrepancies were demonstrated practically on challenging in vivo datasets (cardiac & supraclavicular). Finally, synthetic signals were processed with either the same simulated spectrum or another (NTE=9) to probe sensitivity to fat spectra. Table 1 Summary of evaluated state-of-the-art opensource Fat-Water reconstruction algorithms