Phase‐cycled balanced SSFP imaging for non‐contrast‐enhanced functional lung imaging

Abstract

Purpose To introduce phase‐cycled balanced SSFP (bSSFP) acquisition as an alternative in Fourier decomposition MRI for improved robustness against field inhomogeneities. Methods Series 2D dynamic lung images were acquired in 5 healthy volunteers at 1.5 T and 3 T using bSSFP sequence with multiple RF phase increments and compared with conventional single RF phase increment acquisitions. The approach was evaluated based on functional map homogeneity analysis, while ensuring image and functional map quality by means of SNR and contrast‐to‐noise ratio analyses. Results At both field strengths, functional maps obtained with phase‐cycled acquisitions displayed improved robustness against local signal losses compared with single‐phase acquisitions. The coefficient of variation (mean ± SD, across volunteers) measured in the ventilation maps resulted in 29.7 ± 2.6 at 1.5 T and 37.5 ± 3.1 at 3 T for phase‐cycled acquisitions, compared with 39.9 ± 5.2 at 1.5 T and 49.5 ± 3.7 at 3 T for single‐phase acquisitions, indicating a significant improvement (p<0.05$$ p<0.05 $$) in ventilation map homogeneity. Conclusions Phase‐cycled bSSFP acquisitions improve robustness against field inhomogeneity artifacts and significantly improve ventilation map homogeneity at both field strengths. As such, phase‐cycled bSSFP may serve as a robust alternative in lung function assessments.

Dynamic mode decomposition of dynamic MRI for assessment of pulmonary ventilation and perfusion

“…Regarding the pulse sequence selection, the proposed method was demonstrated on dynamic images obtained using the conventional 2D balanced SSFP (bSSFP) pulse sequence due to its wide availability. Nevertheless, the DMD algorithm is independent of the underlying pulse sequence and can be utilized with other pulse sequences capable of dynamic acquisition, such as standard or transient spoiled gradient echo (GRE), 3,31 ultra‐fast SSFP, 33 tiny golden angle UTE, 34 or multi‐acquisition bSSFP 12 …”

“…Exemplary DMD modes obtained from a volunteer measurement are displayed in Figure 1 with their respective frequencies and mode amplitudes. Last, fractional ventilation map (hereafter referred to as ventilation map) and normalized perfusion map (hereafter referred to as perfusion map) were calculated as previously explained 12,25 . In this work, perfusion maps were normalized using the 99th percentile of voxels inside an area encompassing lungs including large vessels.…”

“…Nevertheless, the DMD algorithm is independent of the underlying pulse sequence and can be utilized with other pulse sequences capable of dynamic acquisition, such as standard or transient spoiled gradient echo (GRE), 3,31 ultra-fast SSFP, 33 tiny golden angle UTE, 34 or multi-acquisition bSSFP. 12 Last, the proposed method can be advanced by incorporating additional priors to DMD. Recently, extensions to DMD have been proposed to improve the eigenvalue prediction, such as forward-backward algorithm 35 to improve noise resilience, multiresolution DMD algorithm to sift out information at different time-scales 36 akin to low-rank plus sparse methods, 37,38 or compressed sensing DMD algorithms 39,40 to relax sampling requirements.…”

“…Recently, extensions to DMD have been proposed to improve the eigenvalue prediction, such as forward-backward algorithm 35 to improve noise resilience, multiresolution DMD algorithm to sift out information at different time-scales 36 akin to low-rank plus sparse methods, 37,38 or compressed sensing DMD algorithms 39,40 to relax sampling requirements. These extensions can be incorporated to exploit the low-rank and sparsity properties present in the dynamic lung MR images 41,42 to further improve the performance of DMD algorithm.…”

“…As such, these methods can suffer from the amplitude and frequency estimation issues. 9 To improve the robustness of amplitude and frequency estimation, modifications to the FD method such as truncation of acquisitions into complete time-series 12 ; and more advanced techniques such as matrix pencil (MP) decomposition 13 which rely on Hankel matrix structure have been previously proposed. Yet, the truncation method decreases the scan efficiency by excluding collected measurements during post-processing; whereas the MP implementation described previously 13 requires a multi-step approach with low-pass and high-pass filtering operations before the spectral analysis.…”