Abstract
Knowledge of the extent and location of viable tissue is important to clinical diagnosis. In principle, sodium (^23^Na) and potassium (^39^K) MRI could noninvasively provide information about tissue viability. In practice, imaging of these nuclei is difficult because, compared with water protons (^1^H), ^23^Na and ^39^K have lower MR sensitivities (9.2 and 0.051%, respectively), and lower in vivo concentrations (ca. 1000‐fold). On the other hand, the relatively short T~1~ relaxation times of ^23^Na and ^39^K (ca. 30 and 10 ms, respectively) suggest that optimized imaging pulse sequences may in part alleviate the weak signal of these nuclei. In this study, numerical simulations of high‐speed imaging sequences were developed and used to maximize ^23^Na and ^39^K image signal‐to‐noise ratio (SNR) per unit time within the constraints of existing gradient hardware. The simulation demonstrated that decreasing receiver bandwidth at the expense of echo time (TE) results in a substantial increase in ^23^Na and ^39^K image SNR/time despite the short T~2~ and T~2~* of these nuclei. Referenced to the available ^1^H signal on existing 1.5 T scanners, the simulation suggested that it should be possible to acquire three‐dimensional ^23^Na images of the human heart with 7 × 7 × 7 mm resolution and ^39^K images with 26 × 26 × 26 mm resolution in 30 min. Experimentally in humans at 1.5 T, three‐dimensional ^23^Na images of the heart were acquired in 15 min with 6 × 6 × 12 mm resolution and signal‐to‐noise ratios of 11 and 7 in the left ventricular cavity and myocardium, respectively, which is very similar to the predicted result. The results demonstrate that by choosing imaging pulse sequence parameters that fully exploit the short relaxation times of ^23^Na and ^39^K, potassium MRI is improved but remains impractical, whereas sodium MRI improves to the point where ^23^Na imaging of the human heart may be clinically feasible on existing 1.5 T scanners.