Optimizing 3D noncartesian gridding reconstruction for hyperpolarized ¹²⁹Xe MRI-focus on preclinical applications

The goal of this work is to characterize and optimize gridding reconstruction of 3D radial hyperpolarized (HP) 129Xe MRI. In support of this objective, we developed a flexible, open source reconstruction software package in MATLAB to optimally reconstruct radially acquired, undersampled HP 129Xe MRI. Using this framework, we demonstrate the effects of 5 key reconstruction parameters: overgridding, gridding kernel function, kernel sharpness, kernel extent, and the density compensation algorithm. We further demonstrate how each parameter can be tuned to optimize a high-resolution 3D radially acquired HP 129Xe image of a ventilated mouse. Specifically, wrap-around artifact, caused by non-selective RF excitation of signal in the trachea, was eliminated by overgridding onto a finely spaced k-space grid; high-frequency aliasing was reduced using iterative density compensation; image SNR and sharpness were optimized by tuning kernel sharpness; and computational burden was minimized by defining an appropriate kernel extent. Compared to our previous reconstruction methods, this optimized method extended visualization from the 5th to 6th generation of mouse airway, while maintaining comparable SNR. Although optimized here for preclinical mouse MRI, this work suggests that 3D radial acquisition offers many broader advantages to undersampled HP gas MRI. Using the methods presented here, we maintained image quality across datasets acquired with various degrees of undersampling and differing SNR by adjusting only a single parameter. These methods are now available to optimize radially acquired hyperpolarized gas images in both the clinical and preclinical arena.

Duke Scholars

Author Scott Robertson Radiology

Author Bastiaan Driehuys Radiology