Optimizing gradient waveforms for microstructure sensitivity in diffusion-weighted MR

Abstract

Variations in gradient waveforms can provide different levels of sensitivity to microstructure parameters in diffusion-weighted MR. We present a method that identifies gradient waveforms with maximal sensitivity to parameters of a model relating microstructural features to diffusion MR signals. The method optimizes the shape of the gradient waveform, constrained by hardware limits and fixed orientation, to minimize the expected variance of parameter estimates. The waveform is defined discretely and each point optimized independently. The method is illustrated with a biomedical application in which we maximize the sensitivity to microstructural features of white matter such as axon radius, intra-cellular volume fraction and diffusion constants. Simulation experiments find that optimization of the shape of the gradient waveform improves sensitivity to model parameters for both human and animal MR systems. In particular, the optimized waveforms make axon radii smaller than 5 μm more distinguishable than standard pulsed gradient spin-echo (PGSE). The identified class of optimized gradient waveforms have dominant square-wave components with frequency that increases as the radius size decreases.