Pulmonary ventilation imaged by magnetic resonance: at the doorstep of clinical application

Contrast agents: intravenous administration

The most widely employed technique used for MRA requires the administration of i.v. contrast agents, such as Gd­chelates. They all reduce the longitudinal relaxation time (T₁‐time) of blood, resulting in an increase in signal intensity, but are disadvantaged by removal from the circulation. Thus, imaging is performed best during the first pass of the contrast agent. This requires fast imaging, synchronization of the MRI acquisition with bolus arrival, and breath­hold techniques. This technique delineates the pulmonary arterial vessels down to 5th–6th order 27. It has been used in a clinical setting to study pulmonary embolism (both acute and chronic emboli) with reasonable success (fig. 2⇓), although clinical management studies have not yet been performed 23, 28, 29. Blood pool agents (ultra small super­paramagnetic iron oxide particles, coated Gd­complex compounds, albumin­binding agents) enable a more prolonged investigation time, greater spatial resolution or coverage and imaging of slow or complex blood flow 30.

Perfusion of lung parenchyma is one of the most important parameters in the determination of gas exchange complementing the assessment of lung ventilation. Strategies have been developed which make use of the first pass of a concentrated contrast bolus or blood pool agents 22, 31, 32. Another technique uses specialized MR image sequences to “tag” the flow of blood (arterial spin tagging) but is limited by insufficient spatial resolution, low signal­to noise, and long measurement times 33. Most recently, an animal study showed the combined use of ventilation MRI (using hyperpolarized gases) and perfusion MRI, resulting in an over­lay technique showing V′/Q′ correlation 34. The main advantages of MRI over scintigraphy in this respect are the greater spatial resolution, the incorporation of functional and anatomical data, visualization of the pulmonary arteries by MRA and the lack of ionizing radiation.

Contrast agents: administration by inhalation

In an attempt to directly visualize the pulmonary airways, several groups have tried to administer the widely used Gd­chelates by inhalation. In the beginning, the results were not convincing because very little contrast agent reached the small airways and airspaces. Even uses of different kinds of nebulizers, mixtures, dilutions and animal models were associated with a lot of difficulties. First studies in rats were performed in 1992, and an increase of ∼70% was observed 35. The administration of aerosolized Gd­DTPA was successfully complemented by a perfusion study using an intravascular contrast agent for investigations of pulmonary embolism and bronchial obstruction in an animal model 35. In a different study, commercially available Gd­DTPA was mixed with mannitol and a surface­active detergent. This modified contrast agent was aerosolized (particle size 0.5–5 µm) and administered to small animals. It gave high signal intensity from the alveolar space 36. The preparation yielded a significantly higher increase in lung signal intensity than usually obtained from Gd­DTPA, since the osmotic effect of mannitol increased the density of the protons. At the same time the detergent reduced surface tension and lead to smaller droplets. In a recent study using a porcine model, more promising results have been obtained: Gd­DTPA was aerosolized and yielded a homogeneous distribution (fig. 3⇓), and an average signal increase of 118% 37.

Oxygen enhancement

Oxygen can also be used as a contrast agent in H‐1 MRI, since its paramagnetic properties can affect the lung tissue and blood by direct contact. Dissolution of oxygen causes T₁‐time shortening in tissue and blood, resulting in a signal increase in affected areas as shown in a T₁‐time­weighted image. By taking the difference of two images, one with the subject breathing room air and the other when breathing 100% oxygen, ventilation can be imaged. The typical increase in signal intensity is in the order of 15% 38, 39. The imaging sequence used is called inversion recovery 40, where variation of the inversion time can weight the image for V′ or Q′. Image quality and signal can be further improved by suppressing the signal from muscle and fat 41. In combination with oxygen­enhancement, high signal visualization of the pulmonary parenchyma has been achieved as shown in figure 4⇓. Since the signal increase is mainly determined by oxygen in the capillaries and the lung veins, there was a higher increase in signal in the lung cortex than in the medulla 41.

In summary, ventilation imaging simply by changing the inspiratory oxygen concentration is an attractive principle compared with sophisticated and complex techniques like hyperpolarized noble gases (see later). With appropriate postprocessing algorithms, this imaging technique can be easily implemented into a clinical routine (e.g. as a complementary step in the MR­diagnosis of acute pulmonary embolism 42). Further applications include studies of regional oxygen uptake in obstructive and interstitial lung disease. It is unclear whether the individual contributions of regional changes of ventilation, diffusion (oxygen uptake) or perfusion to the overall gas exchange can be separated. One limitation of oxygen­enhanced imaging compared to using hyperpolarized gases, is the inability to perform dynamic imaging of ventilation, as images need to be taken at equilibrium of breath mixtures.

Hyperpolarized noble gases

Utilizing hyperpolarized noble gases with MRI is a recent approach for ventilation imaging. In contrast to the proton­based techniques, a dedicated gas is used as a “contrast agent” to directly visualize the gas, rather than relying on indirect effect (e.g. oxygen on blood or tissue T₁‐time). Normally, the density of gases is too low to produce a detectable signal. This drawback is overcome by artificially increasing the amount of polarization per unit volume (hyper­polarization) using an optical pumping technique. Optical pumping (polarization) of inert noble gases, ³He or ¹²⁹Xe, was developed in the 1960s when physicists started experiments to elucidate the structure of the neutron 43, 44. Two techniques for generating hyperpolarization have been established: 1) spin exchange: indirect transfer of angular momentum from a laser source to the nuclei of ³He or ¹²⁹Xe using an alkali metal such as rubidium; and 2) metastability exchange: direct transfer of angular momentum from laser light to ³He nuclei via a radio frequency discharge. A detailed review of advantages and disadvantages of either technique is beyond the scope of this article. The reader is referred to reviews and original articles in the literature 45, 46.

The two inert noble gases have different properties: 1) ³He is a very rare isotope, which is a by­product from tritium (Hydrogen‐3 (³H)) decay. It is particularly suited for ventilation studies since it has no known deleterious side­effects. Additionally, ⁴He, the stable counterpart of ³He with similar physical properties, is widely used for decompression in deep­sea diving, special PFTs (measurement of FRC), and jet ventilation. 2) ¹²⁹Xe occurs more commonly, existing in quantities of 26% in naturally abundant Xe. Blood, fat, and tissue absorb Xe, making it promising for perfusion and gas exchange studies. An impediment to such studies is that in high concentrations and during prolonged exposure, Xe acts as an anaesthetic.

For the specific purposes of ventilation imaging, ³He provides a number of advantages over ¹²⁹Xe. The polarization is higher and the gyromagnetic ratio of ³He is approximately three times higher than ¹²⁹Xe, yielding a signal advantage of almost an order of magnitude 47. Current techniques for ³He yield polarization rates of 30–40% using spin­exchange 48, 49 and 35–50% using metastability exchange 50, whereas for ¹²⁹Xe they are only 5–7% 51–53. Finally, the lack of anaesthetic effects of ³He has made clinical trial studies easier to expedite. The main disadvantage of ³He is the limited availability compared to naturally abundant Xe.

Most developments of hyperpolarized gas imaging have taken place on “standard” 1.5 T MR systems. Since the polarization is generated by optical pumping rather than by the magnetic field of the MR scanner, this technique is ideally suited for high­quality imaging in an open, low­field system 54. Using such systems would lower the costs of the investigation, while the open concept would increase patient comfort, especially if dyspnoea is a problem. One of the drawbacks of low­field systems is that proton imaging is generally complimentary, both for anatomical detail as well as perfusion studies. Unfortunately, these studies become more difficult to perform in low‐field systems. Another interesting application is to perform MR in a physiological setting (i.e. with the patient in the upright position). Such a system would have significant implications for imaging of ventilation, as it would more closely match the physiological state.

Gas density imaging in animals

Initial efforts in imaging of hyperpolarized noble gases were focused on visualizing gas in the lungs during breath­hold. Such images are termed gas density because no special weightings or breathing techniques are used. One of the key differences between imaging hyperpolarized gas and conventional equilibrium protons is the lack of signal recovery with the gas. Because the gas is externally hyperpolarized, it is in a nonequilibrium state, so the application of radiofrequency pulses destroys the signal irreversibly. Thus, careful consideration of image sequences and parameters is required. The first demonstration of hyperpolarized ¹²⁹Xe imaging was in excised mouse lungs 55. This was followed by ³He lung imaging in guinea­pigs ex vivo 47, and in vivo 45, as well as in vivo ¹²⁹Xe images in rats 56. Using multiple breath­holds, high­resolution images of the airspaces can be acquired (fig. 5⇓).

Gas density imaging in humans

As with the animal experiments, first studies in humans were gas density images of airways and alveolar airspaces in an inspiratory breath­hold, using either ³He or ¹²⁹Xe 58–61. A typical image series requires a breath­hold period of <10 s to obtain ∼10 coronal images covering the whole lung. Spatial resolution is significantly better than in nuclear medicine (2.5×2.5×10 mm with 2–5 mm gap between slices). Assessment of the ³He signal intensities revealed preferential ventilation of the posterior lung areas in supine position 50. The same data can also be used to measure the volume of the ventilated airspaces, revealing a good correlation (r=0.88) with the results of PFTs 62. Absolute lung volumes were estimated after introduction of additional correction factors.

Initially, investigators studying gas density imaging concluded that normal ventilation is represented by an almost complete and homogeneous distribution of the ³He signal within the lung 63. However, some small transient ventilation defects (order of 2 cm) have been observed in healthy nonsmokers. Most of them were in the posterior lung fields and have been attributed to posture 64. These defects are regarded as physiological, and illustrate the sensitivity of ³He MRI. More widespread ventilation defects or inhomogeneities indicate impaired regional ventilation 63. The exact threshold between physiological variations and pathological findings has yet to be determined.

Transient ventilation defects in the anterior and posterior lung fields were also observed in subjects with a history of asthma and seasonal allergies, but without noticeable pulmonary symptoms. In some instances, the small peripheral defects seemed to migrate between examinations on different days. These defects were attributed to mucous plugging or bronchospasm 48, 65. The clinical experience with different lung diseases imaged by ³He gas density techniques is reviewed below.

Helium‐3 diffusion imaging in animals

A unique difference between conventional proton­MRI and hyperpolarized gas imaging, is the degree of diffusive motion. All liquids and gases have an inherent property of random microscopic molecular movement, called Brownian motion. This property is represented by the diffusion coefficient. Typically, gases have diffusion coefficients that are 10⁴–10⁵ times greater than those of liquids. The diffusion coefficient can be used to calculate a diffusion length; the average distance a molecule travels per unit time. The diffusion coefficient is not directly related to convection and diffusion by which air reaches alveoli during breathing. In that case, diffusive gas movement follows a concentration gradient. When hyperpolarized gas molecules are in the lung airspaces, the diffusive Brownian motion is measured by MRI during breath­holding without a concentration gradient. Since the airspaces are of an order smaller than the diffusion length of ³He, the diffusive motion is restricted and a smaller apparent diffusion coefficient (ADC) is measured. The diffusive motion can be used as a contrast parameter in MR images by using specialized pulse sequences. A magnetic field gradient is used to coherently disrupt the signal. However, the same strength field gradient applied in the opposite direction will reconstitute the signal. If the molecules have moved due to diffusion in between the application of these field gradients, the signal from that region will be attenuated. For instance, using a heavily diffusion­weighted sequence will result in total destruction of signal in the large airways, but signal will still remain in the distal airspaces. Using a series of diffusion­weighted images, an ADC map can be calculated. From the diffusion map, it has been shown that the ³He diffusive motion can change as much as a factor of 10 in a healthy guinea­pig lung 70. The ADC measured in the trachea was 2.4 cm²·s⁻¹, whereas the average ADC in the alveolar spaces was 0.16 cm²·s⁻¹. This technique has been applied to a rat model of emphysema, showing significant differences between healthy and diseased lungs 49. Differences in alveolar size detected using MR diffusion techniques were in the order of 20 µm, as verified by histology.

Helium‐3 diffusion imaging in humans

In a pilot study of four volunteers, Brookeman et al. 71 investigated the potential of MR imaging with ³He to assess regional and age­related distribution patterns of gas diffusion. The diffusion maps of ³He in the healthy human lungs had uniformly distributed ADC values with a mean value of 0.2–0.3 cm²·s⁻¹. However, older subjects, had higher ADC values with larger standard deviations (SDs). These trends are consistent with an increase in alveolar size, as a result of gradual loss of interstitial lung tissue. An anteroposterior gradient in ADC values was measured in healthy volunteers, corresponding to smaller airspaces in the posterior lung regions (supine position during the examination) 72. This effect will be minimal at full inspiration and more obvious on expiratory scans. It has also been demonstrated that ADC measurements are highly reproducible with an SD <0.03 cm²·s⁻¹ 72. In further studies, the ADC values of normal healthy volunteers where compared with patients with COPD 73 and patients with severe emphysema 74. In the trachea, the ADC was measured between 0.9–1.0 cm²·s⁻¹, and reduced values were found in the alveolar spaces of healthy lungs (0.2–0.4 cm²·s⁻¹). However, in COPD and emphysema patients, alveolar ADC values ranged from 0.4–0.9 cm²·s⁻¹ indicating an increase of airspace size. Broadening of the histogram represented the heterogeneity of emphysematous destruction. The typical apical predominance of centrilobular emphysema could also be demonstrated 73.

ADC mapping in emphysema patients, however, is limited by the concomitant ventilation defects. Hanisch et al. 75 determined ADC values of ³He in the trachea (0.67 cm²·s⁻¹), in normal parenchyma (0.13 cm²·s⁻¹) and in lung fibrosis with honeycombing (0.35 cm²·s⁻¹). In normal lungs, diffusive gas movement was isotropic, whereas in fibrosis and emphysema, diffusion was anisotropic. This may indicate a nonspherical change in geometry of the alveoli and a preferential direction for diffusive gas movement 76.

These measurements reveal a new approach for differentiating normal from diseased lung. Higher ADC values seen in severe emphysema are consistent with increased alveolar size as a result of alveolar wall destruction. Diffusion imaging of the human lung may aid in the assessment of emphysema and other diseases that alter alveolar size or may be associated with reduced elastic recoil of the alveoli. Further investigations of preferential directions for diffusive gas movement (anisotropy) might lead to the evaluation of local compliance. Diffusion imaging might be an important complement of gas density imaging to increase the specificity of ³He MRI, potentially leading to several important clinical applications. First, it will be possible to perform cross­sectional population studies to determine the normal ageing pattern of the lung, which will determine a normal range of values. With the standard range, the influence of exogenous factors, such as smoking and industrial exposure to substances, can be studied. Also, the normal range can be used as a metric for early identification of emphysema in patients at risk and test industrial compensation. ADC measurements can also be used as a research tool for early treatment of emphysema. As new drugs become available, both the efficacy and the long­term effects could be studied using ADC values. Finally, ADC measurement could contribute in the determination of patients suitable for lung volume reduction surgery as well as for follow­up of patients with α‐1 antitrypsin deficiency receiving substitution therapy.

Helium‐3 dynamic imaging in animals

Another unique aspect of gas imaging is the ability to directly image dynamic ventilatory function. With the appropriate imaging sequences, cine imaging can show continuous visualization of the respiratory cycle, including inspiration, distribution of ³He within the alveolar space, and expiration. Distribution analysis of normal and abnormal ventilated regions and corresponding time constants becomes feasible.

In animal studies, timing of breathing must be regulated by mechanical ventilation. By synchronizing the excitation pulse with influx of gas, it is possible to image the gas as it enters into the lungs 77, 78. There are several parameters that can be varied to obtain dramatically different types of images: the flip angle, the length of the acquisition window, and the placement of the acquisition window. These parameters have been explored in a study by Chen et al. 79, which showed airway branching down to the 5th order. More sophisticated development of pulse sequences have resulted in “movies” of ventilation with temporal resolutions of 50 ms 80.

Helium‐3 dynamic imaging in humans

In human studies, real­time techniques have been developed to visualize the distribution of ventilation. High temporal resolution was achieved (40–130 ms), but with reduced spatial resolution 81, 82. Images have been obtained in the coronal or transverse plane, but it is still unclear which orientation is best suited for distribution analysis. In healthy volunteers, the distribution of ³He was observed in the trachea (time­to­peak 260 ms), the mainstem bronchi, the peripheral airways and in alveolar space (time­to­peak 910 ms) 82.

Gierada et al. 83 recently investigated the distribution of ³He in healthy volunteers and emphysema patients. Patients with smoking­related centrilobular emphysema had severe ventilation defects predominately in the upper lobes. Distribution during the first inhalation of ³He was characterized by sequential filling of nonsegmental lung regions interspersed with unfilled regions. The signal distribution became more homogeneous during rebreathing of the ³He gas from a bag. This observation was attributed to collateral ventilation and represents different intrapulmonary time constants. The authors demonstrated that severely diseased regions (as determined from CT) correlated with delayed or absent ventilation at ³He MRI. Washout was significantly prolonged in emphysema patients with signal persisting in some regions, most likely due to gas retention and low oxygen concentration (increased T₁‐time).

There are several potential clinical applications for dynamic ventilation imaging. Typically, impaired distribution with delayed ventilation in lung regions distal to diseased airways will precede the development of significant ventilation defects. Thus, dynamic ³He MRI may help in early detection of obstructive airway disease, when the disease is still reversible. Another application for dynamic imaging is to determine impaired ventilation patterns in different types of emphysema. Emphysema caused by smoking tends to result in diffuse changes, whereas congenital or nonsmoking related emphysema have more focal changes. Since the latter group is more likely to benefit from lung volume reduction surgery, regional data about impaired ventilation distribution may be helpful in surgical planning.

Helium‐3 measurements of intrapulmonary oxygen concentration in animals and humans

The partial pressure of oxygen (P_O2) in ventilated airspaces of the lung varies due to local imbalances between V′/Q′. Routinely, alveolar V′/Q′ is measured indirectly in end­expiratory gas at the mouth, and compared to the P_O2 in arterial and venous blood. From these global measurements, physiological models allow calculation of compartmental P_O2 values in the lungs. However, these values do not give the true P_O2 present at a specific time and location in the alveolar space. Regional imbalances of V′/Q′ ratios, however, may be an important indicator for potentially reversible lung diseases which are compensated by physiological regulation mechanisms, such as hypoxic vasoconstriction, and may support planning of thoracic surgery.

The T₁‐time of ³He in the lung is ∼20 s. Within the airspaces, the irreversible polarization loss of ³He and the continual loss of signal is mainly caused by radiofrequency pulses and relaxation due to paramagnetic molecular oxygen 84, 85. There is a linear dependency between the concentration of oxygen and the relaxation time of hyperpolarized ³He as has been demonstrated in vitro 84. This property has been used to compute the intrapulmonary oxygen concentration in live pigs 86. By measuring the relaxation times of the hyperpolarized ³He gas within the animal lungs during two different imaging series, Eberle and coworkers 86, 87 were able to make use of this phenomenon to compute the intrapulmonary oxygen concentration in vivo from ³He MRI images. The values sampled from large alveolar regions correlated significantly with global end­expiratory values from conventional respiratory gas analysis (r=0.88). Deninger et al. 88 utilized this relaxation effect to determine the local oxygen partial pressure and to quantify time­dependent changes of P_O2 in human volunteers. The temporal evolution of P_O2 during apnoea was found to be linear, both in an animal model and a human volunteer. These preliminary results yielded physiologically plausible measurements for initial P_O2 and rates of P_O2 decrease with time (R). The apparent correlation between P_O2 with the decrease rate R, indicates the local matching of ventilation and perfusion and reflects the relative contributions of alveolar space and peripheral anatomical deadspace to a volume of interest.

In conclusion, measurements of P_O2 give indirect evidence about lung perfusion. This unique information is clinically important in identifying regions of V′/Q′ mismatch in diseased lungs, which cannot be obtained by any laboratory method or clinically applicable technique measuring true P_O2 regionally. Patients suffering from lung diseases with V′/Q′ imbalance, like pulmonary embolism or bronchiectasis, may have a method for early detection by noninvasive ³He MRI measurements of P_O2.

Xenon‐129: gas­ and dissolved­phase studies

Xe is another inert noble gas that can be imaged using MR techniques (fig. 9⇓). One of the main differences between ¹²⁹Xe and ³He is the solubility of Xe in blood and lipid­rich tissue, thus dissolved­phase imaging is possible 55. Using inhalation techniques, only a small fraction (∼2%) of Xe is dissolved into lung parenchyma or blood. Once Xe enters the blood stream it is distributed throughout the body by circulation. Thus, it is possible to perform chemical shift imaging in the lung, kidney, and brain, as has been demonstrated in rats 52, 89. The resonance of Xe is strongly influenced by its environment and thus, a chemical shift occurs. The dissolved­phase resonances of Xe are shifted by ∼200 parts per million from the gas­phase resonance. Spectra measured in humans and rats have shown resonances that correspond to fat, tissue, and red blood cells. The collection of gas­ and dissolved­phase signal is possible within a single acquisition. In 1997, Mugler et al. 61 achieved combined imaging and spectroscopic approach in two human volunteers. Comparing ¹²⁹Xe and H‐1 images, they found good correlation between the gas­space signal void in the proton images and the gas­space signal in Xe­images. Combined imaging of the gas­phase and dissolved­phase ¹²⁹Xe MRI may lead towards simultaneous V′/Q′ studies of the lung. Further studies into the dynamics of Xe in the chest have been studied in the dog 53. After inhalation, the signal from tissue (or red blood cells) was destroyed. The signal decay in the gas­phase was observed, providing a measure of the absorption of Xe into the tissue. Time constants of 61 ms were measured for tissue saturation with Xe and 70 ms for red blood cell saturation.

The high solubility of Xe into the blood stream makes it possible to use it as a diffusible tracer for measurements of Q′ of different organs, such as lung and brain. However, there are a number of drawbacks to using Xe as a Q′ agent. Xe has anaesthetic properties, which limit its amount and hence signal, available for a study. The deficiency in signal is further compounded by low polarization rates. Early studies were limited to 2% 60, while more recent studies have reported levels up to 5–7% 52, 53. Using isotopically enriched ¹²⁹Xe gas (71–79%) improves the signal, but is about 20 times more expensive than naturally abundant ¹²⁹Xe.

Fluorine­containing compounds as magnetic resonance­contrast agents for ventilation imaging in animals

The low spin density of the intrapulmonary gases can be compensated for not only by hyperpolarization of the spin system, but also by using either a high number of signal averages 90, 91, or by increasing the spin density by filling in a fluid compatible with oxygen or carbon dioxide gas exchange. These two approaches have shown to be feasible when performed with fluorinated substances. Since the MR­visible isotope ¹⁹F has both a high natural abundance as well as high gyromagnetic ratio, a high sensitivity of ¹⁹F MRI can be achieved when compared to H‐1 MRI.

Magnetic resonance imaging in clinical algorithms of patient­work­up

The developments of MR for V′ imaging have now reached preliminary clinical studies. The charm of these new techniques is that they can be used to obtain information about ventilation, perfusion, anatomical and functional data in a nonionizing radiation environment. This allows for repeated studies to assess treatment response. Further development of MR ventilation imaging will involve demonstration of superiority over nuclear medicine procedures (while some competition with CT also exists). Higher sensitivity and specificity for recognized diseases will need to be shown, as well as demonstration of improved clinical significance for well­recognized indications. Finally, the new ability to incorporate more refined functional information should show improved definition of disease states, which should change the selection of more refined drug and surgical therapies (dependent on the type and state of the disease process). Thus, early detection and treatment of obstructive lung disease could result in reversibility of the disease process, which could lead to improved prognosis of the illness. Furthermore, the treatment regime could be based on more individual parameters leading to improved quality of life, changed treatment modalities and could influence the handling of riskful situations, such as anaesthesia. This can be achieved by demonstrating impaired distribution of V′, changes of diffusive gas movement in the peripheral airways or by different regional oxygen concentration. The forthcoming years will show which MRI­techniques will finally prevail or whether several methods will be established in conjunction with different clinical indications.