Advances in functional and structural imaging of the human lung using proton MRI

Abstract

The field of proton lung MRI is advancing on a variety of fronts. In the realm of functional imaging, it is now possible to use arterial spin labeling (ASL) and oxygen-enhanced imaging techniques to quantify regional perfusion and ventilation, respectively, in standard units of measurement. By combining these techniques into a single scan, it is also possible to quantify the local ventilation-perfusion ratio, which is the most important determinant of gas-exchange efficiency in the lung. To demonstrate potential for accurate and meaningful measurements of lung function, this technique was used to study gravitational gradients of ventilation, perfusion, and ventilation-perfusion ratio in healthy subjects, yielding quantitative results consistent with expected regional variations. Such techniques can also be applied in the time domain, providing new tools for studying temporal dynamics of lung function. Temporal ASL measurements showed increased spatial-temporal heterogeneity of pulmonary blood flow in healthy subjects exposed to hypoxia, suggesting sensitivity to active control mechanisms such as hypoxic pulmonary vasoconstriction, and illustrating that to fully examine the factors that govern lung function it is necessary to consider temporal as well as spatial variability. Further development to increase spatial coverage and improve robustness would enhance the clinical applicability of these new functional imaging tools. In the realm of structural imaging, pulse sequence techniques such as ultrashort echo-time radial k-space acquisition, ultrafast steady-state free precession, and imaging-based diaphragm triggering can be combined to overcome the significant challenges associated with proton MRI in the lung, enabling high-quality three-dimensional imaging of the whole lung in a clinically reasonable scan time. Images of healthy and cystic fibrosis subjects using these techniques demonstrate substantial promise for non-contrast pulmonary angiography and detailed depiction of airway disease. Although there is opportunity for further optimization, such approaches to structural lung imaging are ready for clinical testing.

Keywords: arterial spin labeling; functional lung imaging; oxygen-enhanced pulmonary MRI; respiratory triggering; specific ventilation imaging; structural lung imaging; ultrashort echo time; ventilation-perfusion ratio.

Figure 1

ASL-FAIRER in the healthy supine lung. A blood-bright (control) image and a blood-dark (tag) image are acquired approximately 5 seconds apart. Subtraction yields an image of pulmonary blood delivered to the slice during the delay time between the magnetic tag and the image acquisition, encompassing one systolic ejection period. These are images of a 15 mm sagittal slice through the middle of the right lung. The HASTE pulse sequence collects 72 phase-encoding lines of a 128×256 matrix with ∼3.0×1.5 mm in-plane resolution, using an inter-echo time of 4.5 ms. Reproduced with permission from [26].

Figure 2

The time course of Fluctuation Dispersion (see text for details), both FD_Global (panel D) and FD_Local (panel E), as subjects were switched from breathing air to one of three different gas mixtures: hypoxic (blue), normoxic (black), and hyperoxic (red). Note that both hypoxia and hyperoxia resulted in changes in fluctuation dispersion, indicating alterations in pulmonary vascular tone from both atypical gas mixtures compared to breathing air. Reproduced with permission from [27].

Figure 3

The time course of recovery following methacholine challenge. The leftmost panel shows the map of specific ventilation obtained before challenge with a uniform pattern of ventilation. The remaining panels show SV maps obtained at the indicated times after challenge. The development of ventilation defects due to bronchoconstriction (especially in the dependent caudal lung) and subsequent recovery of ventilation to these regions can be clearly seen from left to right in the figure.

Figure 4

Example images of (A) lung density, (B) alveolar ventilation, (C) perfusion, and (D) V˙_A/Q˙ ratio in a sagittal slice of the right lung in a healthy subject positioned in the supine posture. Voxels with larger conduit blood vessels are excluded from the calculation of regional perfusion and V˙_A/Q˙ ratio since they do not represent perfusion and incorrectly map as regions of shunt. Reproduced with permission from [46].

Figure 5

Effect of prone versus supine positioning on the gravitational gradient of the ventilation-perfusion ratio in healthy subjects. In prone posture the vertical distribution of regional V˙_A/Q˙ ratios is more uniform. Reproduced with permission from [46].

Figure 6

(a) Balanced SSFP pulse sequence combined with a 3D radial UTE acquisition. The readout gradient direction varies for each excitation to evenly sample the interior of a sphere in k space. To avoid nerve stimulation, the gradient slew rates are usually smaller than the hardware limit, and a slightly longer ramp time is often used for the first (readout) lobe. Standard 0°/180° phase alternation is applied to the excitation RF pulses. (b) A sphere in k space, showing a fully sampled spoke-radial k-space trajectory corresponding to a 25×25×25 Cartesian matrix. The dots on the surface of the sphere indicate the endpoints of the radial spokes. The equivalent Cartesian matrix for an actual, high-resolution lung scan is much larger than the one used for this example, resulting in much denser sampling than shown here. (c) Respiratory triggering of the segmented acquisition. The triggering threshold is set in advance by the operator, and is expressed as a percentage of the normal diaphragm excursion between inspiration and expiration. A 40% threshold setting is depicted here.

Figure 7

Elements of the respiratory triggering procedure. (a) Sagittal planning image through the middle of the right lung, showing the position of the navigator window (superimposed rectangle). (b) Navigator trace showing the craniocaudal position of the diaphragm as a function of time during free breathing. (c) Example coronal images from two separate, low-resolution 3D scans, each triggered at a different part of the respiratory cycle.

Figure 8

Three-dimensional structural imaging in cystic fibrosis. The upper row of images show 2mm-thick coronal views reconstructed from a non-triggered 3D scan, performed while the subject was freely breathing. The lower images are 2mm-thick views at the same slice positions, reconstructed from an equivalent respiratory triggered, segmented acquisition. Although the actual data acquisition time was the same for both scans (2.5 min), the total scan time was considerably longer for the respiratory triggered acquisition (7 min), since k-space data was acquired during only ∼35% of the respiratory period. Arrows point to thickened airways viewed in cross section, and the rectangular inset shows a cluster of even smaller airways. Many of these fine details are not well resolved in the non-triggered scan. Pulse sequence parameters included: flip angle = 15°; TR/TE = 1.18/0.09 ms; voxel size = 2.0 mm isotropic; FOV= 392 mm; equivalent Cartesian matrix = 196×196×196; 121,952 radial spokes; trigger threshold = 50%; readout gradient amplitude = 17.5 mT/m, ramp time = 230 μs, flat top time = 110 μs.

Figure 9

Maximum intensity projections (MIPs) in health and disease. (a-c) 20mm-thick MIPs constructed from a respiratory triggered 3D scan of a healthy subject, showing the pulmonary vasculature. (d) Single 2mm axial slice at the same location as the MIP in (b). Arrows indicate lobar fissures. (e-g) 20mm-thick MIPs constructed from the respiratory triggered scan of the cystic fibrosis subject shown in Figure 8. Mucous and inflammation in the airways, which run alongside the pulmonary arteries, obscure much of the vasculature in this subject. Despite the lack of cardiac synchronization, there are no obvious artifacts of heart motion in these 3D radial acquisitions. Pulse sequence parameters for the scan of the healthy subject included: flip angle = 15°; TR/TE = 1.19/0.09 ms; voxel size = 2.0 mm isotropic; FOV= 512 mm; equivalent Cartesian matrix = 256×256×256; 207,057 radial spokes; trigger threshold = 25%; readout gradient amplitude = 17.6 mT/m, ramp time = 240 μs, flat top time = 100 μs. Total scan time was 12 minutes; actual data acquisition occupied 4.1 minutes.