Single Breath-Hold MR Spectroscopy for Diffusion-Weighted Lipid Droplet Size Estimation
Abstract
Purpose
To explore the potential of diffusion-weighted spectroscopy to estimate lipid droplet diameter with short acquisition times (< 30 s).
Methods
Stimulated-echo single-direction diffusion-weighted (DW-STEAM; Gmax = 50 mT/m, δ = 25 ms) MR spectroscopy with log-spaced mixing times (TMmin = 50, TMmax = 500 ms) was modeled and used for experiments. Long (~4 min, 10 TMs) and short (< 30 s, 6 TMs) single voxel DW-STEAM with constrained T1 were compared. Cramér-Rao bound analysis estimated theoretical parameter uncertainty. Pure sunflower oil in oil-in-water emulsions were homogenized at different rates (12,000-35,000 rpm) and durations (30-120 s) to produce a range in droplet diameters (~1-45 um, validated by microscopy) in 50 mL tubes for phantoms. All DW-STEAM triglyceride peaks were fit and a Murday-Cotts model for restricted diffusion was used to estimate T1-constrained lipid droplet diameter and standard error.
Results
Cramér-Rao lower bound calculations for SNR = 100 estimate the lipid droplet diameter coefficient-of-variation <25% over ~4-20 um for both long and short sequences, consistent with signal simulations. DW-STEAM droplet diameter estimates ranged from 15 um to the maximum modeled size (100 um) spanning from the most homogenized emulsion to pure oil, respectively, and were positively biased relative to microscopy. The 95% confidence interval of the difference between estimated diameter from long versus short DW-STEAM included zero but were larger than expected from theory. Standard errors in diameter estimates increased with increasing lipid droplet size.
Conclusion
Single breath-hold DW-STEAM-based measurements of lipid droplet size are statistically consistent with longer acquisitions. However, both long and short DW-STEAM suffer from estimation uncertainties outside the narrow range of diameters for the given mixing times. This suggests that this rapid technique may be most useful to dichotomize large/unrestricted (≳ 20 um) versus small/restricted (≲ 5 um) lipid droplet sizes, rather than precise size estimation.