$k$ -Space-Based Enhancement of Pulmonary Hyperpolarized 129Xe Ventilation Images

Abstract

Hyperpolarized (HP) noble gas (e.g., 3He/129Xe) MR images are susceptible to noise and artifacts due to the rapid attenuation of nonrenewable HP magnetization along with the scan time. However, a little attention is paid to this issue through postprocessing techniques. Here, a $k$ -space-based analysis method is proposed to improve the overall signal-to-noise ratio (SNR) and lessen the blurring of pulmonary HP 129Xe ventilation images. This method performs $k$ -space weighting using low-frequency boost and a high-frequency modulation which offsets the magnetization loss per excitation, and then integrates two outputs in image domain through a convex combination model. Furthermore, a nonreference quality metric, called second-derivative measure of improvement by entropy (SMIE), is introduced to assess the image quality of HP gas MRI. The simulation results demonstrate that the enhanced images are statistically significantly different to the original ones regarding the SNR, peak SNR, structural similarity, and SMIE (each $P$ -value is less than 0.0001). In vivo results indicate that the proposed method significantly upsets the SNR and SMIE in human pulmonary HP 129Xe ventilation images (all $P$ -values are less than 0.05), while maintaining the appearance of ventilation defects or fine structures. In this case, the proposed scheme has potential for improving the understandability and/or differentiation of regions of interest in the lung.