Introduction

A recent study¹ evaluated repeatability of obstructive patterns in obstructed vs. healthy lungs via static ventilation imaging using 3D radial ultrashort TE oxygen-enhanced MRI (UTE OE-MRI). Here, we extended this technique from steady-state to dynamic imaging and evaluated the feasibility of 3D dynamic UTE OE-MRI in the assessment of regional O₂ delivery, uptake and wash-out in healthy lungs.

METHODS

Four healthy subjects were enrolled in a HIPAA-compliant study with IRB approval. Each subject underwent DESPOT1² and dynamic OE-MRI sequentially using 3D radial UTE sequence on a 1.5T scanner (HDxt, GE Healthcare, Waukesha, WI). Subjects were positioned supine and breathed freely via facemask with 21% (normoxic) or 100% O₂ (hyperoxic) flowing at 15 L/min throughout the scan. For DESPOT1, five flip angles (2°, 4°, 6°, 10°, and 14°) were chosen based on the expected T₁ estimates from the literature. While breathing medical air, all subjects were scanned using respiratory-gated free-breathing 3D radial UTE (TE=80μs) sequence with variable readout gradients³ at each flip angle with 32cm FOV, 30,000 projections, 1.25mm isotropic resolution, total scan time ~14min. With the same FOV and an 8° flip, we then performed dynamic OE-MRI using the same UTE sequence. The dynamic scan consisted of 14 or 34 dynamic time series to cover two wash-in and wash-out cycles: subjects breathed 21% O₂ for the first 2 timeframes, and then the gas supply alternated between 100% and 21% O₂ with 3 (70/52s temporal) or 8 (24s temporal) timeframes at each oxygen concentration, total scan time 13-16 min. The image resolution was down-sampled to a uniform lower 1cm³ resolution for T₁ and PSE assessment. Image Analysis: Lungs were segmented using a 3D region-growing method with automatically selected seed points. We used linear least squares fitting to calculate voxelwise T₁ values from magnitude UTE volumes after 3D rigid registration. The proposed post-processing workflow¹ was integrated for the analysis of dynamic time series. First, all the later frames were registered to the first normoxic volume (timeframe#1). Second, the Jacobian determinant derived from registration was applied voxelwise to account for density changes due to physiological volume differences between normoxic vs. hyperoxic breathing. Third, whole lung and aortic median percent signal enhancement (MPSE) was calculated using registered, density corrected volumes with respect to timeframe#1. Fourth, voxelwise T₁ estimates from DESPOT1 were used as the baseline T₁ at timeframe#1 to estimate $$$T_{1}(t)$$$ at subsequent timepoints to derive changes in partial pressure of oxygen, $$$\triangle PO_{2}(t)=\frac{\frac{1}{T_{1}(t)}-\frac{1}{T_{1,baseline}}}{r_{O2}}$$$, where r_O2=2.49×10^-4/s mmHg⁴. The O₂ wash-in and wash-out portions of $$$\triangle PO_{2}(t)$$$ curves were fitted using exponential functions⁵, $$$\triangle PO_{2}(t)=\triangle PO_{2max}\times(1-e^{-\frac{t}{\tau_{up}}})$$$ and $$$\triangle PO_{2}(t)=\triangle PO_{2max}\times e^{-\frac{t}{\tau_{down}}}$$$, to compute the time constants ($$$\tau_{up}$$$ and $$$\tau_{down}$$$) individually, where $$$\triangle PO_{2max}$$$ is the maximum ΔPO₂ after switching air to 100% O₂. The association of MPSE time curves in the lungs vs. aorta was evaluated using Spearman correlation.

RESULTS

Table 1 summarizes the averaged measurements of all subjects derived from whole lung parametric maps. Figure 1 shows typical 3D T₁ maps. Figure 2 shows time series examples of 3D PSE and ΔPO₂ maps in the lungs and PSE in the aorta from a subject at 52s temporal resolution. The MPSE time series in the lungs significantly correlated with that in the aorta (ρ=0.996, p<0.0001). Figure 3 shows examples of 3D maximum PSE, ΔPO_2max, wash-in and wash-out time constant maps. Figure 4 shows the averaged global MPSE and median ΔPO₂ over time curves for all subjects at two temporal resolutions.

DISCUSSION

This study used 3D dynamic UTE OE-MRI to obtain quantitative measures of oxygen wash-in and wash-out for a 24-second temporal resolution corresponding to a 13.2-min scan time. An additional ~14-min DESPOT1 scan enables ΔPO₂ measures. The whole lung T₁ estimates are comparable to the literature value 1200ms using inversion recovery sequence⁶. Compared to the measurements of asthmatics assessed from a coronal slice using dynamic OE-MRI⁵, we observed relatively higher median ΔPO_2max and shorter time constants ($$$\tau_{up}$$$ and $$$\tau_{down}$$$) in this normal cohort. Experiments with varying temporal resolutions suggest that global MPSE time curves may be more robust than ΔPO₂ time curves at a higher temporal resolution where signal-to-noise ratio degrades. Further work improvements include: 1) investigating the optimal temporal resolution for reproducibility and 2) evaluating this technique in asthmatics.

CONCLUSION

The feasibility of using 3D dynamic UTE OE-MRI with full chest coverage and isotropic spatial resolution to achieve adequate temporal resolution for characterizing quantitative O₂ wash-in and wash-out dynamics, together with the use of conventional MRI equipment and low-cost O₂ gas makes this technique promising for clinical application in evaluating obstructive lung disease.

Acknowledgements

This project was supported in part by grant UL1TR000427 to UW ICTR from NIH/NCATS, Research and Development Fund from Departments of Radiology and Medical Physics, University of Wisconsin-Madison, and grant support from GE Healthcare. We also acknowledge the research technologists who supported the MRI scanning in this work, including Kelli Hellenbrand, RT, Jenelle Fuller, RT, and Sara John, RT.