Free-breathing multi-slice ultra-fast SSFP acquisitions for multi-volumetric morphological and functional lung imaging
Orso Pusterla1,2, Grzegorz Bauman1,2, and Oliver Bieri1,2

1Department of Radiology, Division of Radiological Physics, University of Basel Hospital, Basel, Switzerland, 2Department of Biomedical Engineering, University of Basel, Basel, Switzerland


Multi-volumetric breath-hold imaging was recently proposed for respiratory $$$\alpha$$$-mapping, a novel quantitative measure of pulmonary ventilation. In this work, we evaluate the feasibility to derive 3D pulmonary functional maps from morphological lung data which are reconstructed from a time-series of multi-slice 2D ultra-fast SSFP scans acquired in free-breathing (4D-MRI).


Recently $$$\alpha$$$-mapping was introduced as a novel proton-based ventilation-weighted quantitative measure1, derived from the naturally occurring signal intensity (SI) modulations of the lung at diverse inspiratory phases. In contrast to e.g. Fourier decomposition2, respiratory α-mapping takes into account not only breathing-related SI modulations, but also associated lung volume changes. From a physical lung model relating the observed SI changes to the lung volume1, the derived α-index becomes virtually independent on the inspiratory phase, breathing amplitude and regional spin-density. For respiratory α-mapping, volumetric ultra-fast steady-state free procession (ufSSFP3) scans are acquired at different breath-hold (BH) positions. This technique has shown promising results for reproducible and sensitive detection of respiratory defects in COPD and cystic fibrosis patients. Heavily respiratory compromised patients, infants or young children, however, might be unable to reliably perform breath-holding maneuvers. Therefore, we propose to derive 3D pulmonary functional maps from multi-volumetric data which are reconstructed from a reordered time-series of multi-slice 2D ufSSFP scans acquired in free-breathing (referred to as 2D-FB).

Materials and Methods

Measurement protocol

All measurements were performed on a 1.5T clinical MR-scanner using a 12-channel body array and a 24-channel spine receiver coil in a healthy volunteer positioned supine.
A contemporary 2D-ufSSFP multi-slice sequence2 was modified for coronal multi-slice imaging using a TE/TR = 0.72/1.50 ms, a flip angle = 40º (maximal due to SAR constrain), a readout bandwidth of 1953 Hz/pixel, a field-of-view = $$$450\times450\times196$$$ mm3 (in-plane spatial resolution of $$$3.5\times3.5$$$ mm2, interpolated to $$$1.75\times1.75$$$, slice-thickness 7 mm), a parallel imaging factor of 2, yielding 8 images/s (125 ms per slice). In order to let the magnetization to recover between each slice acquisition, two interleaved groups of 28 slices each were acquired coronally. The acquisition was repeated 48 times in free breathing, resulting in a nominal scan time of 5 min 30s. Neither prospective gating nor fixed-slice navigator acquisitions were used.

For comparison, the same 2D multi-slice acquisition was acquired also in breath-hold (2D-BH) with an acquisition time of 7 seconds per chest-image, as well as in 3D in 15 seconds with isotropic resolution of 3.1 mm3 [3D-BH, see Ref. (3) for details]. For respiratory α-mapping, BH scans were performed at 5 different inspiratory levels comprised in the tidal breathing region.

Image post-processing and analysis

Reconstruction of multi-volumetric data from free-breathing scans was performed with a retrospective respiratory gating. To this end, the diaphragm position was extracted for each slice from a navigator window (cf. Fig. 1). Subsequently, images were binned as a function of the retrospectively derived diaphragm position and recombined to form multi-volumetric datasets at different inspiratory phases (4D-MRI). Finally, respiratory α-maps were derived as previously described1, using $$\alpha\left(\vec{x}\right)=-\frac{\partial\log\left(\textrm{SI}\left(V,\vec{x}\right)\right)}{\partial\log\left(V\right)}\quad[1]$$

after lung volume segmentation, 3D-deformable image-registration (Elastix4), and median filtering5. Generally, $$$\alpha\left(\vec{x}\right)=1$$$ indicates that a region-of-interest (ROI, at position $$$\vec{x}$$$) expands equally to the whole lung mean expansion, whereas $$$\alpha\left(\vec{x}\right)<1$$$ indicates that the ROI expands less than the whole lung mean expansion [for $$$\alpha\left(\vec{x}\right)>1$$$, more].

In addition, from the vector-transformations resulting from image co-registration, lung deformation fields and Jacobian-based estimates of lung regional ventilation were calculated.


A comparison between 3D-BH, multi-slice 2D-BH and multi-volumetric data reconstructed from multi-slice 2D-FB scans is presented in Fig. 2 in sagittal and coronal views and for a respiratory volume in the middle of the tidal respiratory phase. Overall, despite of the simplistic reordering scheme, only residual misalignments of vessels and diaphragm are recognizable, particularly in the dorsal part of the chest.
A side-by-side comparison of respiratory α-mapping resulting from the three acquisition-modalities is presented in Figure 3. Respiratory α-maps are characterized by iso-gravitational homogeneity and a physiological ventral-to-dorsal respiratory-gradient, which is less pronounced for 2D acquisition, especially in FB; we theorize that this effect is caused primarily by non-perfect image reconstruction (reordering) and in second place by physiological effects.
Figure 4 shows respiratory α-maps, deformation field and Jacobian determinant-maps resulting from 2D-FB acquisitions. A moderate visual similarity is seen for the three functional imaging measures, i.e. grater expansion in the inferior lobes.

Discussion and Conclusions

In this work, we have shown the feasibility to obtain 3D functional pulmonary mapping ($$$\alpha$$$-mapping, deformation fields / motion-dynamics and Jacobian-expansion) from multi-volumetric morphological images of the lung which are reconstructed from retrospectively binned multi-slice 2D-FB scans. The reconstructed lung volumes still suffer from small residual misalignments, to be addressed in future investigation, e.g. according to (6).

In conclusion, with an overall acquisition time of approximately 5 minutes, the proposed scheme seems promising for multi-volumetric morphological imaging and for functional respiratory mapping in infants or pediatric patients, as well as in subjects unable to perform breath-holding maneuvers.


This work was supported by a grant from the Swiss National Science Foundation (Grant No. 320030_149576).


[1] Pusterla O, Bauman G, Wielputz MO, Nyilas S, Latzin P, Heussel CP, Bieri O. Rapid 3D in vivo 1H human lung respiratory imaging at 1.5 T using ultra-fast balanced steady-state free precession. Magn. Reson. Med. 2016.

[2] Bauman G, Bieri O. Matrix pencil decomposition of time-resolved proton MRI for robust and improved assessment of pulmonary ventilation and perfusion. Magn Reson Med. 2016 Jan 12.

[3] Bieri O. Ultra-fast steady state free precession and its application to in vivo 1H morphological and functional lung imaging at 1.5 tesla. Magn. Reson. Med. 2013;70:657–663.

[4] Klein S, Staring M, Murphy K, Viergever MA, Pluim JPW. Elastix: A toolbox for intensity-based medical image registration. IEEE Trans. Med. Imaging 2010;29:196–205.

[5] Bieri O. A Method for Visualization of Parenchyma and Airspaces from 3D Ultra-Fast Balanced SSFP Imaging of the Lung at 1.5T. In: Proc. Intl. Soc. Mag. Reson. Med. 22; 2014. p. 2300.

[6] Tong Y, Udupa JK, Ciesielski KC et al. Retrospective 4D MR image construction from free-breathing slice Acquisitions: A novel graph-based approach. Med. Image Anal. 2017;35:345–359.


Figure 1. (a) Coronal slice acquired with multi-slice ufSSFP in FB; retrospectively, a navigator window (red-dashed line) is positioned on the right hemisphere of the diaphragm to depict its position. (b) Signal intensity in a vertical line of a coronal slice (yellow line on the left figure) as a function of the image number after the reordering process. The pink arrow indicates the diaphragm position in tidal inspiration, the yellow arrow in expiration (functional residual capacity).

Figure 2. Sagittal (maximum intensity projection, MIP=10 mm) and coronal (MIP=20 mm) views of three different ufSSFP imaging. (a) 3D-ufSSFP in BH (15s scan-time per 3D volume), (b) 2D-multislice in BH (7s per 3D volume), (c) 2D-multislice in FB (5 min scan time for 48 3D volume). Images are representatively shown for a lung volume in the middle of the tidal respiratory phase. To notice the high signal from within the lung parenchyma reached by ufSSFP imaging.

Figure 3. Sagittal and coronal view of respiratory α-maps derived from multi-volumetric datasets using (a) 3D-BH, (b) multi-slice 2D-BH, (c) multi-slice 2D-FB. There is visual similarity among the three derived respiratory-maps, but due to non-perfect multi-volume reconstructions and physiological effects, slightly lower α-values are found for 2D-FB.

Figure 4. (a) Respiratory α-maps, (b) deformation field and (c) Jacobian expansion resulting from expiration-to-inspiration (functional residual capacity to tidal-inspiration, lung volumes 2.4-3.4 L) co-registration/transformation of 3D images reconstructed from multi-slice 2D-FB acquisitions. The respiratory index α represents the expansion of each voxel as a function of the mean lung expansion (cf. ‘Method’ section). The deformation field shows the direction and strength of motion-dynamics from expiratory-to-inspiratory phases. The determinant of the Jacobian (Jac) is a voxel-wise estimation of lung expansion/compression: if Jac>1, the local tissue expands (e.g. for Jac=2, its size is doubled from expiration-to-inspiration), Jac<1 represents tissue contraction.

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)