Single breath-held acquisition of coregistered 3D 129Xe lung ventilation and anatomical proton images of the human lung with compressed sensing

Purpose: To develop and assess a method for acquiring coregistered proton anatomical and hyperpolarized 129Xe ventilation MR images of the lungs with compressed sensing (CS) in a single breath hold. Methods: Retrospective CS simulations were performed on fully sampled ventilation images acquired from one healthy smoker to optimize reconstruction parameters. Prospective same-breath anatomical and ventilation images were also acquired in five ex-smokers with an acceleration factor of 3 for hyperpolarized 129Xe images, and were compared to fully sampled images acquired during the same session. The following metrics were used to assess data fidelity: mean absolute error (MAE), root mean square error, and linear regression of the signal intensity between fully sampled and undersampled images. The effect of CS reconstruction on two quantitative imaging metrics routinely reported [ percentage ventilated volume (% VV) and heterogeneity score] was also investigated. Results: Retrospective simulations showed good agreement between fully sampled and CS-reconstructed (acceleration factor of 3) images with MAE (root mean square error) of 3.9% (4.5%). The prospective same-breath images showed a good match in ventilation distribution with an average R2 of 0.76 from signal intensity linear regression and a negligible systematic bias of + 0.1% in % VV calculation. A bias of -1.8% in the heterogeneity score was obtained. Conclusion: With CS, high-quality 3D images of hyperpolarized 129Xe ventilation (resolution 4.2 x 4.2 x 7.5 mm3) can be acquired with coregistered 1H anatomical MRI in a 15-s breath hold. The accelerated acquisition time dispenses with the need for registration between separate breath-hold 129Xe and 1H MRI, enabling more accurate % VV calculation.