LAUSR

Quantitative MR parametric maps of the heart have become routine elements of the cardiac magnetic resonance imaging (MRI) toolkit at many institutions, adding diagnostic and prognostic benefit above standard assessment of myocardial volumes, function, and fibrosis. Beyond wellestablished T1, T2, and T2* parameters, methods to assess a variety of other MRI physiological parameters of interest, such as perfusion, diffusion, and proton density fat fraction (PDFF), are at various stages of development. While these methods are unquestionably expanding the capabilities of modern cardiac imaging, significant drawbacks remain. Most often, the various parametric mapping sequences are acquired separately, slice-by-slice under breathholds, increasing patient fatigue, and resulting in images that are not necessarily coregistered. Maps may be derived from any one of several different pulse sequences, with unique sensitivity profiles affected by field strength, system imperfections, and patient-specific physiology. Finally, the models used to fit signal evolutions generally remain simplistic and often do not account for known confounders to parameter estimation. Significant effort is being made to address these issues, but for now, the Society for Cardiovascular Magnetic Resonance consensus statement on cardiac MRI parametric mapping recommends that extreme care be taken when evaluating these measures, as direct comparisons of quantitative results across scanner vendor, sequence type, and software version remains challenging. Cardiac magnetic resonance fingerprinting (cMRF) provides an intriguing, alternate solution to many of these problems. Based on the paradigm-shifting seminal MRF work introduced by Ma et al, cMRF can acquire multiple parameter maps simultaneously, with the number of parameters limited in principle only by dictionary size and the ability of the sequence to generate distinguishable signal contrast behavior. Additionally, confounding system imperfections and parameter interactions can be modeled into the matching dictionaries, with the caveat that additional parameters in the dictionaries rapidly expands their size. For the heart, where dictionaries must be generated uniquely based on the patientspecific ECG signal, the incorporation of additional parameters must be balanced against increased computation time. In recent work, Jaubert et al described a cMRF technique to compute T1, T2, and PDFF maps using a Dixon framework to decompose signals from fat and water prior to dictionary matching. This approach, which they term Dixon-cMRF, enables the use of reasonable dictionary sizes for patient-specific dictionary generation (additional fatrelated parameters are not required as part of the dictionary matching step, nor is a separate inhomogeneity map scan required). In this issue of JMRI, the next natural step in the development of Dixon-cMRF is reported, with an assessment of the bias and precision of the method. First, the authors conducted a thorough repeatability analysis of each estimated parameter in a phantom and in healthy volunteers, with comparisons to established methods (MOLLI, T2-GRASE, and a gradient echo PDFF sequence for T1, T2, and PDFF, respectively). Second, Dixon-cMRF’s performance against these established techniques was evaluated in a small clinical cohort with known or suspected heart disease. Overall, Dixon-cMRF achieved good precision in the phantom and in healthy subjects, with repeatability measures for each parameter generally similar to or better than those measured by the corresponding established techniques. In the small patient cohort, precontrast T1, T2, and PDFF measurements from the Dixon-cMRF maps were comparable to healthy volunteers, with similar or better image quality compared to conventional mapping methods. Small, statistically significant differences in measured mean values for native T1 and T2 were found between the Dixon-cMRF parameter estimates and conventional methods in both the healthy and clinical cohorts, but no differences were found between PDFF, postcontrast T1, or derived synthetic extracellular volume fraction. Several important limitations to the study must be noted. All experiments were performed on one scanner, and the sample size used for the repeatability analysis was small