Introduction

Perfluoropropane (C₃F₈), due to its high inertness and low solubility in water and blood¹, enables detailed measurement of ¹⁹F gas wash-in dynamics over consecutive breath-holds^2,3. Since parameters deduced from dynamic gas imaging strongly depend on consistent volumes of inspired gas, variations in inhaled gas volumes can impair the accuracy and inter-scan repeatability of the evaluated parameters. A previous study showed an increase in repeatability and correlation with spirometric measurements when limiting the inhaled gas volumes experimentally⁴. However, the results strongly rely on the patient’s compliance and the volume control may reduce the patient's comfort and requires an advanced experimental setup. This study investigates a retrospective correction of the inhaled volumes (CIV), without restricting the patients breathing and modifications in the experimental procedure.

Methods

24 patients with chronic obstructive pulmonary disease (COPD) with different severity (Global Initiative for Chronic Obstructive Disease (GOLD) stages⁵: I (4 patients), II (12 patients) and III (8 patients)) underwent lung function testing and ¹⁹F MRI (1.5T, MAGNETOM Avanto, Siemens Healthcare, Erlangen, Germany) on the same day, followed by a second 19F MR scan after a maximum of 7 days. ¹⁹F imaging was performed with a transmit birdcage coil and a 16-channel phased-array receive coil (Rapid Biomedical, Rimpar, Germany). A 3D spoiled gradient echo sequence with golden-angle stack of stars sampling with the following parameters was used: TE 2.1ms, TR 4.9ms, flip angle 35°, FOV 400x400x400mm³, matrix size 64x64x16, 5/8 partial Fourier along the partition dimension and pixel bandwidth 345Hz/pixel. For the gas wash-in, the patients inhaled a mixture of 79% C₃F₈ and 21% O₂ for eight consecutive breaths, with a ¹⁹F scan in each breath-hold (duration: 6.3s) where all inspired volumes were measured using pneumotachometers (TSD117A-MRI, BIOPAC Systems, Goleta, CA). The images were reconstructed using the parallel imaging and compressed sensing algorithm (PICS) of the BART toolbox⁶. The eight scans were reconstructed regularizing the total variation in the time domain (λ = 0.01) and applying local low rank regularization in image and time domain (λ = 0.01). Wash-in (WI) and fractional ventilation (FV) maps were generated by fitting the signal of each voxel to the respective mono-exponential model:
$$
I(t) \propto (1-exp(- \frac{t}{WI})) \qquad I(t) \propto (1-(1-FV)^t),
$$
where t represents the number of the respective breath-hold. CIV was implemented by replacing each timepoint with the ratio of the actually inhaled volume to the mean of all inhaled volumes. This adjustment shifts the timepoints along the signal curve, simulating a consistent volume for each breath-hold. A graphical representation of the method can be found in Figure 1. The assessed parameters are the median of the derived wash-in time constants (WI median), the interquartile range of the wash-in time constants (WI IQR) and the ventilated volume percentage (WI VVP), defined as the percentage of voxels with wash-in time constants <= WI median+3 breaths, the median of the fractional ventilation maps (FV median) and the interquartile range of the fractional ventilation (FV IQR).

Results

The Pearson correlation coefficient (r) of the ventilation parameters with the predicted forced expiratory volume in one second (FEV1%pred) increased with CIV, which is shown in Figure 2. For FV median, r increased from 0.63 without CIV to 0.70 with CIV; for FV IQR, from 0.17 to 0.24; for WI median, from -0.52 to -0.72; for WI IQR, from -0.28 to -0.59; and for WI VVP, from 0.59 to 0.64. As displayed in Figure 3, the coefficient of variation (COV) of FV median, WI median and WI IQR was significantly reduced with CIV (all p<0.045). The intraclass correlation coefficient (ICC) with CIV exceeded the ICC without CIV for all assessed parameters, as shown in Table 1. Figure 4 shows representative fractional ventilation and wash-in maps of the first and second scan with and without CIV.

Discussion

The presented results show an increased correlation of ventilation parameters with FEV1%pred and an improved repeatability. This suggests that the proposed method may provide a more accurate assessment of dynamic ventilation parameters. However, CIV is not able to correct a varying diaphragm position due to variations in inhaled gas volumes, which may lead to inaccurate measurements in the lower parts of the lung.

Conclusion

The CIV is a retrospective correction which is simple in implementation and can achieve an increased correlation as well as increased inter-scan repeatability of dynamic ventilation parameters in ¹⁹F gas wash-in MRI. Since CIV is implemented in the post-processing, it can be applied in the evaluation of existing data and without altering the experimental setup, as long as the inhaled gas volumes were measured.

Acknowledgements

This work was funded by the German Center for Lung Research (DZL). The authors would like to express their gratitude to the radiographers Frank Schröder and Melanie Pfeifer from the Department of Radiology for their support with the MR measurements and patient care.