Fast 3D quantitative 1H ventilation imaging of the human lung at 1.5T with SSFP
Orso Pusterla1,2, Grzegorz Bauman1,2, Mark Wielpütz3,4, Claus Heussel3,4, and Oliver Bieri1,2

1Radiological Physics, Dep. of Radiology, University of Basel Hospital, Basel, Switzerland, 2Department of Biomedical Engineering, University of Basel, Basel, Switzerland, 3Department of Diagnostic and Interventional Radiology with Nuclear Medicine, Thoraxklinik at University Hospital Heidelberg, Heidelberg, Germany, 4Department of Diagnostic and Interventional Radiology, University Hospital of Heidelberg, Heidelberg, Germany

Synopsis

Monitoring lung ventilation is of great interest to assess pulmonary function and disease progression. Here, a novel, fast, and simple three-dimensional (3D) quantitative in vivo 1H imaging method is introduced, reflecting regional ventilation information. To this end, typically five ultra-fast balanced steady state free precession (ufSSFP) scans are repetitively performed in breath-hold from which a respiratory index map, $$${\gamma}$$$, is derived. The new measure $$${\gamma}$$$ shows high reproducibility in healthy volunteers and high sensitivity to respiratory defects, such as in patients with COPD.

Introduction

Detection of impaired pulmonary function is of high importance for both the diagnosis of pulmonary diseases as well as for monitoring of disease progression or response to treatment. Here, a novel native proton-based MRI technique is introduced to map human lung ventilation in 3D with ultra-fast balanced steady-state free precession (ufSSFP1). Regional ventilation information is reflected by a new quantitative measure, termed respiratory index $$$γ$$$, derived from local proton density modulations of lung parenchyma, as observed for different respiratory volumes. The new method is evaluated in healthy volunteers and in patients with respiratory disease.

Methods

Theory
It has recently been shown that ufSSFP is able to catch native proton signal intensity ($$$SI$$$) modulations in the human lung at different inspiratory levels2. The observed signal modulation (see Fig. 1a), is in excellent agreement with the adapted sponge model3, where the $$$SI$$$ is assumed to scale inversely proportional to the total lung volume $$$V_L$$$: $$SI(\vec{x},V_L)=\alpha \cdot V_L^{-\gamma(\vec{x})}+β \Rightarrow γ(\vec{x})=-\frac{\partial \log(SI(\vec{x},V_L))}{\partial \log(V_L)} \qquad\text{[1]}$$where $$$α$$$ is a scaling factor, $$$β$$$ the signal intensity in the limit of infinite volume, i.e. noise, and $$$γ(\vec{x})$$$ maps regional lung expansion and changes in pulmonary blood volume during breathing, referred to as respiratory index.
The model expects $$$γ(\vec{x})$$$ to be sensitive in the detection of the local lung function impairment. If the impaired lung regions expand less or if no change of pulmonary blood volume occurs, then a smaller change in $$$SI$$$ is measured and thus lower respiratory index values are expected.

Measurement protocol
3D ultra-fast SSFP imaging of the lung was performed in supine position at 1.5T (Siemens MAGNETOM Avanto) in breath-hold (< 17 sec) in the tidal respiratory region with an isotropic resolution of 3.1-3.5 mm3 using a TE/TR = 0.47/1.19 ms and a flip angle α = 23º (technical details available in Ref.2). Scanning was repeated five times, naturally leading to small variations in the breath-hold positions, and thus in the respiratory volumes.
The feasibility of the technique was evaluated in five healthy volunteers, in two patients with chronic obstructive pulmonary disease (COPD), in one patient with pancoast tumor, and in a patient with chronic pulmonary embolism (CPE). One volunteer was scanned three times at three consecutive days to assess reproducibility.

Image post-processing and analysis
The lung volume ($$$V_L$$$) was determined for each of the five datasets using a 3D fast-marching algorithm4.
Image registration was performed with a mass preserving 3D deformable algorithm implemented in elastix5. Subsequently, registered datasets were processed using a median filter to reduce signal noise and remove the vasculature overlaying the pulmonary tissue6. Finally, $$$γ(\vec{x})$$$ was extracted pixel-wise from Eq. [1] by linear least-squares estimation.

Results

The model is confirmed by Fig. 1a, which shows the measured mean parenchyma $$$SI$$$ as the function of the lung volume (from forced expiration to forced inspiration) on a double logarithmic scale. Exemplar $$$γ$$$-maps obtained in two healthy volunteers are presented in Fig. 1b, revealing homogeneity on isogravitational planes (coronal view) and a gradient from frontal to dorsal lung. In the reproducibility test (not shown), mean $$$γ$$$-values of 1.04, 1.02 and 1.01 in the segmented lung volume were found (coefficient of variation cv=1.5%).
Overall, data acquisition was successfully performed in all patients.
Fig. 2 shows morphological CT, DCE-MRI and the respiratory maps obtained in a 53 years old male COPD patient (GOLD III). The location of regions characterized by the structural emphysematous destruction visible on CT images and perfusion defects on DCE MRI images correlate with functional impairment on the respiratory maps.

Discussion and Conclusions

In this work, we have presented a novel proton-based MRI method for 3D lung ventilation mapping in healthy volunteers and patients. In patients, emphysematous destruction and obstructed lung areas lead to low regional $$$γ$$$-values, as expected.
Since the respiratory index $$$γ$$$ is independent of the inspiratory phase and breathing amplitude, it may serve as a reproducible and normative measure, i.e. to monitor disease progression. In future studies, the reproducibility and clinical value of the method will be further investigated.

Acknowledgements

No acknowledgement found.

References

[1] Bieri O, Ultra-fast steady state free precession and its application to in vivo (1)H morphological and functional lung imaging at 1.5 tesla, Magn Reson Med. 2013 Sep;70(3):657-63. doi: 10.1002/mrm.24858.
[2] Pusterla O et al., How volume affects the pulmonary MRI signal: Investigations with 3D ultra-fast balanced Steady-State Free Precession, Proc. Intl. Soc. Mag. Reson. Med. 23 (2015), 1481.
[3] Staring M et al., Towards local progression estimation of pulmonary emphysema using CT, Med Phys. 2014 Feb;41(2):021905. doi: 10.1118/1.4851535.
[4] Maleike D et al., Interactive segmentation framework of the Medical Imaging Interaction Toolkit, Comput Methods Programs Biomed. 2009 Oct;96(1): 72-83. doi:10.1016/j.cmpb.2009.04.004.
[5] Klein S et al., elastix: a toolbox for intensity-based medical image registration, IEEE Trans Med Imaging. 2010 Jan;29(1):196-205. doi: 10.1109/TMI.2009.2035616.
[6] Bieri O, A Method for Visualization of Parenchyma and Airspaces from 3D Ultra-Fast Balanced SSFP Imaging of the Lung at 1.5T, Proc. Intl. Soc. Mag. Reson. Med. 22 (2014), 2300.

Figures

(a) Measured mean $$$SI$$$ of the parenchyma and the fit of the sponge model as a function of the lung volume ($$$V_L$$$) in double logarithmic scale. (b) Exemplary coronal and axial views of respiratory $$$γ$$$-maps overlaid on the 3D-ufSSFP morphological images for two healthy volunteers.

(a) CT morphological, (b) DCE-MRI and (c) respiratory $$$γ$$$-maps for a COPD patient. The arrows indicate functional defects corresponding to emphysematous regions on CT.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2917