Severity Evaluation in Cystic Fibrosis Using Oxygen-enhanced MRI: Comparison to Hyperpolarized Helium-3 MRI
Wei Zha1, Stanley J Kruger1, Robert V Cadman1, Kevin M Johnson1,2, Andrew D Hahn1, Scott K Nagle1,2,3, and Sean B Fain1,2,4

1Department of Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, 2Department of Radiology, University of Wisconsin-Madison, Madison, WI, United States, 3Department of Pediatrics, University of Wisconsin-Madison, Madison, WI, United States, 4Department of Biomedical Engineering, University of Wisconsin-Madison, Madison, WI, United States

Synopsis

Oxygen-enhanced MRI using 3D radial ultrashort echo time sequence (OE-MRI) is a promising alternative to evaluate ventilation and defects with wider accessibility and better affordability. Eleven cystic fibrosis (CF) subjects with different severities of disease underwent OE-MRI and HP-MRI. The disease severity ranks on the percent signal enhancement map (PSE) derived from OE-MRI was compared to the whole lung ventilation defect percent (VDP) measured from HP-MRI as a reference standard using Spearman rank correlation. The moderate association between VDP and PSE suggest OE-MRI shows promise for differentiating disease severity in CF.

Purpose

To assess the capability of oxygen-enhanced 3D radial ultrashort echo time (UTE) MRI (OE-MRI) in detecting disease severity of cystic fibrosis (CF) using ventilation defect percent (VDP) from hyperpolarized helium-3 MRI (HP-MRI) as a reference standard.

Methods

Eleven (11) CF subjects with various severities were enrolled in HIPPA-compliant studies with IRB approval. Each underwent OE-MRI and HP-MRI scans at 1.5 T (Signa HDx, GE Healthcare, Waukesha, WI).

OE-MRI: The air-breathing (V021) and 100% oxygen-breathing (V100) UTE images were acquired sequentially with the following scan parameters: 3.2 mm isotropic resolution, FOV = 32x32 cm, TR/TE=2.9/0.08 ms, flip angle = 8°, with an 8-channel cardiac surface coil (GE Healthcare, Waukesha, WI). The total scan time is 9 min consisting of the following intervals: 3.5 min. air-breathing at V021, 2 min. wash-in to avoid transient effects, followed by 3.5 min. of 100% O2-breathing. Real-time gating to end-expiration with a 50% acceptance window was used to minimize respiratory motion and obtain consistent lung inflation. V021 and V100 images were reconstructed offline at high (0.7 mm3) resolution for anatomy and at 1cm3 using a Fermi filter to improve signal-to-noise ratio for OE MRI.1

HP-MRI: The 3He scan used a fast 2D multislice gradient-echo sequence with the following parameters: TR/TE = 6.5/2.9 ms; flip angle = 7°; FOV = 40 x 40 cm, slice thickness = 10 mm. Proton MRI was performed prior to the 3He scan using a 2D single-shot multislice fast spin echo sequence under equivalent breath-hold conditions with acquisition matrix, field of view, slice thickness, and position matching the 3He scan.

Analysis: V021 and V100 images were intensity corrected using the open source N4ITK2 and co-registered using ANTs B-spline deformable registration3. Density changes in normoxic vs. hyperoxic breathing were corrected using the adapted slope model.4 The corrected V021 was given by $$$V_{021C} =T(V_{021})/det(J_{T})^{S}$$$ where $$$S =\frac{log({T(V_{021})})-log(V_{100})}{log(det(J_{T}))}$$$, T is the deformable transformation and det(JT) represents the Jacobian determinant of the deformation field. The density corrected percent signal enhancement (PSE) was calculated as $$$PSE =(V_{100}-V_{021C})/V_{021C}$$$. A radiologist with seven years of experience blindly and independently ranked the subjects for disease severity from 1 to 11 with 1 being least severe. Proton images were registered to 3He using 3D affine registration by ANTs.5 The whole lung VDP was measured by the adaptive K-means method.6

Relative severity rankings from OE-MRI and VDP from HP-MRI were compared using Spearman rank correlation in SAS v9.4 (SAS Institute Inc., Cary NC).

Results

Lung volumes measured at V021 were consistently larger than at V100 (Table 1). An example of unregistered vs. registered and density corrected PSE map demonstrates improved alignment (orange arrow in Figure 1), while the density correction accounted for local parenchymal density changes that influence in perceived O2 concentration. Regional patterns of ventilation defects were not consistently well matched between HP-MRI and OE-MRI (Figure 2). However, disease severity, determined from OE-MRI alone (Table 2), were positively related to the whole lung VDP values (ρ = 0.62. p = 0.046, 95% confidence interval = [0.015, 0.89]).

Discussion

The consistent bias in lung volumes observed in V021 relative to V100 reflects the lower density in V021. The applied deformable registration and density correction improved the spatial consistency of PSE maps with anatomy in the presence of systematic physiological volume differences. OE-MRI may ultimately have advantages over other functional MRI techniques in its potential for dissemination. Future work will investigate the waveforms from the bellow signal to assess the effect of uneven breathing, and develop a histogram based approach for quantifying regional disease in PSE maps rather than depending strictly on ordinal ranking.

Conclusion

OE-MRI shows promise in characterizing disease severity of CF using HP-MRI as a reference standard. Further advancements in data acquisition, post-processing, and analysis will improve the capability of OE-MRI in regional defect assessment in obstructive lung disease.

Acknowledgements

We acknowledge support from GE Healthcare, NIH UL1TR000427, and NIH KL2TR000428.

References

1.Kruger S, Fain S, Johnson K, et al., Oxygen-enhanced 3D radial ultrashort echo time magnetic resonance imaging in the healthy human lung. NMR in Biomed. 2014; 27:1535-1541.

2.Tustison N, Avants B, Cook P, et al., N4ITK: Improved N3 bias correction, IEEE Trans Med Imaging. 2010; 29(6):1310-1320.

3.Tustison N and Avants B. Explicit B-spline regularization in diffeomorphic image registration. Frontiers in Neuroinformatics. 2013; 7:39.

4.Staring M, Bakker M, Stolk J, et al., Towards local progression estimation of pulmonary emphysema using CT. Med. Phys. 2014; 41(2):021905.

5.https://github.com/stnava/ANTs

6.Zha W, Kruger S, Cadman R, et al. An adaptive K-means approach for assessment of ventilation defects in asthma and cystic fibrosis using hyperpolarized helium-3 MRI. ISMRM 2015.

Figures

Table 1 Lung volume differences measured from air- (V021) and 100% O2-breathing (V100) scans

Figure 1 An example of unregistered vs. registered PSE with density correction and corresponding color-coded Jacobian determinant det(JT) image and det(JT)^S used in the adapted slope model. Values>1 on det(JT) refer to expansion after registration. Density correction affected the posterior base (pink arrows) and apical chest wall (orange arrows).

Figure 2 Ventilation defects segmented from 3He images (right lung in green, left lung in yellow) and corresponding PSE map from OE-MRI (orange arrows). Large Ventilation defects mismatched in subject #7 (a vs. b) but aligned well in subject #6.

Table 2 Severity ranks based on OE-MRI and whole lung VDP from HP-MRI



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1626