Purpose

To investigate the feasibility of compressed sensing (CS) on the morphologic measurement of lung microstructure using hyperpolarized ¹²⁹Xe diffusion MRI.

Introduction

¹²⁹Xe diffusion MRI is sensitive to acinar airway. Combined with the simplified models such as Weibel model and theoretical methods (1), multi-b diffusion MRI method can offer the information of lung structure at the alveolar level. However, this needs a long scanning time. In this work, we develop a new method combining multi-b diffusion MRI with CS (2) to accelerate the acquisition in order to measure microstructure of human lungs.

Methods

Fully sampled (FS) and undersampled ¹²⁹Xe multi-b diffusion MRI data were acquired from five healthy subjects and one COPD patient, respectively. The parameters of 2D spoiled gradient echo sequence were as follows: FOV = 384*384 mm², TR/TE = 13.9/10.9 ms, matrix = 64*64, number of slices = 4, slice thickness = 30 mm, coronal slices, receiver bandwidth = 250 Hz/pixel, flip angle = 5°, centric phase encoding direction. The bipolar diffusion weighting gradient pulse was as follows: ramp up/down time = 0.3 ms, duration time = 3.7 ms, diffusion time Δ = 5 ms, b = 0, 10, 20, 30, 40 s/cm², the diffusion gradient were located on the slice selection direction. The parameters of the prospective CS acquisition were the same as those of FS except that the phase encoding number reduced to 32 and the flip angle raised from 5°to 7°. In this work, we used the pseudo-random pattern for two-fold undersampling in phase direction. A nonlinear method was used to enforce the sparsity using the L1 norm (3). The reconstruction of the image involves solving equation min_x1/2||Fu-y||₂²+λ₁||Ψx||_*+λ₂TV(x). The maps of mean linear intercept length (L_m) were calculated using cylinder model, L_m map and global mean L_m of CS were compared with those of FS for all subjects. The scatter and Bland-Altman plots were used to verify the agreement between CS and FS measurements.

Results and Discussion

Figure 1 showed the ventilation images of b=0 and L_m maps of one healthy subject and the COPD patient by FS and CS, respectively. In the chosen example slices, some ventilation defect regions could be observed obviously in the ventilation images of the COPD patient. There were generally some differences in lung sizes and positions between FS and CS acquisition, and the differences became more obvious for COPD patient, because two acquisitions were obtained in different breath-holds. The differences in mean L_m of healthy and COPD subjects between FS and CS were 9 μm (4.1%) and 30 μm (8.1%) respectively. The mean L_m difference of the COPD patient was larger than that of the healthy subject, which was in accordance with the difference of ventilation images. The scatter plot and Bland-Altman analysis plot about the slice-by-slice L_m values between FS and CS methods were showed in Figure 2. A good correlation (R = 0.988) was observed (Fig. 2a), and all points fell within the 95% confidence intervals in the Bland-Altman plot (Fig. 2b). The mean L_m percent difference between methods was -1.9%. And 95% of the difference fell within -11.5% to +7.7%. In the quantitative comparisons of L_m, two clusters of data points were observed corresponding to the healthy and COPD groups. The results suggested that the CS method could also be used to extract the pulmonary parameters accurately with the twice shorter time in comparison to the FS method.

Conclusions

In this study, we demonstrated that the multi-b diffusion with CS method is a promising accelerated technique to obtain the lung morphometric parameters in a single breath hold. Our results showed that lung morphometric parameters using prospective CS are in agreement with that of the FS images. This provides confidence for the use of ¹²⁹Xe lung morphometry to detect changes of lung diseases.

Acknowledgements

We acknowledge the support by the National Natural Science Foundation of China (81227902, 81625011) and National Program for Support of Eminent Professionals (National Program for Support of Top-notch Young Professionals).