MRI for the diagnosis of pulmonary embolism

Abstract

Pulmonary embolism (PE) is one of the most frequently encountered clinical emergencies. The diagnosis often involves multiple diagnostic tests, which need to be carried out rapidly to assist in the safe management of the patient. Recent strides in computed tomography (CT) have made big improvements in patient management and efficiency of diagnostic imaging. This review article describes the developments in magnetic resonance (MR) techniques for the diagnosis of acute PE. Techniques include MR angiography (MRA) and thrombus imaging for direct clot visualization, perfusion MR, and combined perfusion-ventilation MR. As will be demonstrated, some of these techniques are now entering the clinical arena, and it is anticipated that MR imaging (MRI) will have an increasing role in the initial diagnosis and follow-up of patients with acute PE. J. Magn. Reson. Imaging 2003;18:627–640. © 2003 Wiley-Liss, Inc.

VENOUS THROMBOEMBOLISM REMAINS ONE OF the most frequently encountered clinical dilemmas, with an estimated annual incidence of three to four suspected cases per 1000 in the general population in the Western world; proven cases of deep vein thrombosis (DVT) and pulmonary embolism (PE) of approximately 1.0 and 0.5 per 1000, respectively; and up to two million cases per year in the United States alone (1, 2). Approximately 10% of patients with a proven venous thromboembolic event will not survive the first hour. In the remaining patients, a subclassification can be made of massive (e.g., life-threatening), submassive, and nonmassive PE (3). Less than 5% will suffer from massive PE. In this situation, echocardiography is generally the best (and most speedy) approach, and treatment will consist of intravenous thrombolytics or (in rare instances) more invasive techniques such as embolectomy or interventional fragmentation (3). In the subgroup of patients with submassive PE, i.e., those with echocardiographic signs of right ventricular dysfunction, there is some evidence that more aggressive therapy may be indicated (4–6), but further studies are required. Thus, the vast majority of cases will have nonmassive PE and can undergo standard diagnostic tests.

Any diagnostic strategy has two main goals. First, the strategy must identify those patients at risk of a (potentially fatal) second event; these patients will require long-term anticoagulant therapy. Second, and of equal importance, the diagnostic approach needs to identify patients in whom anticoagulants can be safely withheld. These aims need to be assessed in the background of the risk of missed diagnosis and unnecessary treatment.

If the diagnosis is missed, there is a considerable risk of recurrence. Only one randomized trial was ever performed to assess treatment vs. placebo in patients with clinically diagnosed PE, which showed a fatal recurrence rate of 30% and a nonfatal second event in a further 30% of patients (7). Given this high incidence of secondary events, the efficacy of diagnostic strategies generally relies on follow-up of patients and assessment of mortality and morbidity from recurrent venous thromboembolism and bleeding complications.

Overall, only approximately 20% to 40% of patients who were initially suspected of PE will have their diagnosis confirmed (8–11). Thus, the group without disease is much larger than those with disease. This latter statement is particularly highlighted in view of the complications associated with anticoagulant therapy: bleeding complications in unfractionated heparin occur in approximately 3% to 4% of patients (9, 12), with similar figures for low-molecular-weight heparins (13, 14). Oral anticoagulant therapy is associated with a fatal bleeding risk of one per 100 treatment years and 4% to 16% nonfatal bleeding risk (15, 16).

Traditionally, the diagnostic management of PE was largely concentrated around lung scintigraphy. A normal perfusion lung scan result effectively rules out PE, while a high-probability lung scan carries a positive predictive value of 90% (8, 9, 17). Unfortunately, 40% to 60% of lung scan results are neither normal nor high-probability (the term nondiagnostic has been advocated).

Although conventional pulmonary angiography is (still) regarded as the reference method (18), it is invasive and carries an, albeit low, complication risk and cannot be performed in up to 20% of patients in whom the procedure is contemplated (19). Thus, developments have been aimed at establishing novel, noninvasive diagnostic modalities, which can offer a safe diagnostic outcome and will reduce the need for angiography. This includes plasma D-dimer (this is a breakdown product of fibrin) to help exclude the disease (11, 20), ultrasonography of the leg veins (based on the observation that 70% to 90% of patients with PE also have demonstrable deep leg vein thrombosis) to help prove the disease (21, 22), and, most prominently, the development of helical computed tomography (hCT) with CT-pulmonary angiography with or without extended CT-venography of the pelvis and leg veins (23, 24).

Several diagnostic management strategies have been employed, usually combining several diagnostic modalities (10, 11, 20–22, 25–28). These strategies have resulted in adequate diagnosis and management of patients with PE. However, they also have several disadvantages, which can be solved by the introduction of magnetic resonance imaging (MRI)-based techniques. Although MRI in the diagnosis of PE is still in its infancy, there are huge gains to be made in terms of its noninvasive nature, safe contrast agents, and (maybe most importantly) its inherent lack of ionizing radiation. This latter point has been emphasized both in general use and in pregnant patients (29, 30). The most recent calculations show that with careful selection of imaging parameters, CT-pulmonary angiography can be safely performed (29, 30); however, the additional use of CT-venography does carry a significant increase in radiation dose (31).

MRI of the chest has been developed relatively recently, when compared to other body areas. This was due to several difficulties, which needed to be overcome, such as the lack of protons within the chest, motion of both heart and lungs, and the susceptibility artifacts due to the interfaces between air and soft tissues. Contrast-enhanced MR angiography (MRA) enabled imaging of the large vessels. New sequences were developed to enable lung perfusion MRI, direct thrombus imaging of the entire venous system, and, more recently, the introduction of hyperpolarized 3-helium, which allows high-resolution lung ventilation imaging. The use of faster imaging sequences and the application of increased gradient-strength systems have made it feasible to image even breathless patients. This review will describe the current state of MRI in the diagnosis of venous thromboembolic disease (Table 1).

Table 1. Magnetic Resonance Imaging Techniques That May Be Applied for Venous Thromboembolic Disease

1. MR angiography

a. Non-Gadolinium enhanced techniques

i. Time of flight MRA

ii. 3D-VUSE

b. Gadolinium-enhanced 3D MRA

i. Gadolinium

ii. Blood pool agents

2. MR direct thrombus imaging

3. MR perfusion imaging

a. Gadolinium-enhanced

i. Gadolinium direct (first pass)

ii. Blood pool agents

b. Non-Gadolinium:

i. Arterial spin labelling

ii. HASTE-type

4. MR perfusion-ventilation imaging

a. Oxygen-enhanced ventilation

b. Hyperpolarized 3-Helium ventilation

5. Imaging directed at lower extremity thrombosis

a. Direct clot imaging

b. Blood pool venography

MRA

Techniques

Several methods may be employed to perform MRA, ranging from time-of-flight angiography to gadolinium-enhanced three-dimensional MRA. Time-of-flight MRA was applied with some success for imaging of the central pulmonary arteries (32–35). However, several problems were identified, including a lack of spatial resolution, insensitivity to slow flow, motion sensitivity (both respiratory and pulsation), and field distortion artifacts. Another time-of-flight technique used variable-angle uniform signal excitation during free breathing (three-dimensional-VUSE), but this was not attempted in patients with suspected PE (36).

The introduction of higher-gradient-strength systems (>20 mT/m) and the development of short TR three-dimensional gradient echo sequences allowed the development of single breath-hold three-dimensional contrast-enhanced MRA (37–39). Depending on the patient's ability to hold his or her breath, either monophasic protocols with a high spatial resolution and scan times in the range of 20–30 seconds (39–44) or time-resolved multiphasic imaging protocols with scan times of under 10 seconds (45, 46) can be applied. The use of shorter scan times allows the investigation even of severely breathless patients. While monophasic MRA protocols with a high spatial resolution achieve a diagnostic accuracy for the detection of PE comparable to helical CT (44), the diagnostic accuracy of time-resolved protocols in patients with PE remains unknown. The recent introduction of parallel MRI techniques, such as the simultaneous acquisitions of spatial harmonics (SMASH) (47) or the sensitivity encoding for fast MRI (SENSE) (48), will further improve breath-hold three-dimensional contrast-enhanced MRA (47–51). In contrast to conventional MRI, parallel MRI uses the spatial information inherent in the geometry of surface coil arrays to reduce the number of phase-encoding steps, thus leading to a faster acquisition time. In practice, parallel techniques allow a substantial improvement of the temporal and/or spatial resolution (Fig. 1). In contrast-enhanced MRA with the use of parallel acquisition techniques, the breath-hold duration of monophasic imaging protocols might be reduced at constant spatial resolution, or the spatial resolution of time-resolved imaging protocols might be improved at fixed breath-hold duration.

Figure 1. Maximum intensity projection of parallel imaging gadolinium-enhanced MRA technique.

Hardware and Software Requirements

High-performance gradient systems with amplitudes in excess of 20 mT/m and slew rates over 120 mT/m/msec have enabled the development of ultrafast, short breath-hold MRA. This has seen a significant reduction of imaging time from a full breath-hold (20 seconds or more) to less than five seconds.

A prerequisite for parallel MR techniques is the utilization of numerous surface or array coils (52). A major drawback of parallel MRI is that the time saving achieved over sequential MRI is accompanied by an increase of the image noise. From experimental studies it is known that the ultimate achievable signal-to-noise ratio (SNR) for parallel MR images is closely tied to the geometry and sensitivity patterns of the coil arrays (53–55). Dedicated coil arrays developed for parallel acquisitions might overcome this limitation.

Gadolinium should be injected using a power injector and is generally followed by a saline flush (45). This generally requires a 20-G to 22-G intravenous access needle. A dose of 0.2 mmol/kg was shown as optimal, while increasing the dose did not improve image quality (56, 57). Gadolinium concentration in the vessels of interest should be optimal at the time of central k-space acquisition, as this will determine the vascular signal intensity (58). Spherical and centric-elliptic phase-encoding ordering offer optimum contrast with conventional Cartesian k-space sampling (59), while interleaved spiral (51) and radial acquisitions (60) inherently oversample central k-space and are thus less sensitive to bolus passage timing. Time-resolved contrast-enhanced three-dimensional MRA methods (61) are the obvious extension whereby continuous three-dimensional data sets are required, thus capturing the arterial and venous phases and obviating the need for precise bolus synchronization. In sequences that require 20- to 30-second breath holds, a trial bolus may be useful to obtain optimal injection-imaging delay time (44). However, in ultrafast imaging sequences, this is no longer required in the majority of patients and a routine delay time of five seconds following the initiation of contrast injection will result in excellent-quality images in patients with suspected acute PE (62).

An alternative method of contrast involves the use of blood pool agents, several of which are under evaluation (Fig. 2). These agents circulate for a prolonged period, thus allowing prolonged imaging as well as opening the prospect of simultaneous imaging of pulmonary circulation and deep venous system of the lower extremities (63).

Figure 2. Normal angiogram using blood pool agent NC100150. Notice both arterial and venous systems are well depicted.

Images should be viewed on a workstation, which allows for imaging of source images as well as multiplanar reformatting, volume rendering, and maximum intensity projections. Thus, a true three-dimensional evaluation can take place with extended views of vessels (64).

Clinical Applications

Only limited clinical studies have evaluated the role of three-dimensional gadolinium-enhanced MRA for the diagnosis of PE in reasonable patient samples. The first study to compare MRA with pulmonary angiography obtained a sensitivity of 70% and a specificity of 100% in a group of 23 patients with PE (37). Another small study evaluated three-dimensional gadolinium-enhanced MRA in 13 patients, but these patients suffered from a variety of lung disorders other than PE (40).

A pioneering study investigated the performance of a coronal gadolinium-enhanced three-dimensional time-of-flight sequence during a single breath hold in 30 patients with suspected PE (43). A total of eight patients had PEs demonstrated by angiography. The study design included independent assessment of MRA by three radiologists and the use of pulmonary angiography as a reference standard. In 10% of patients the images were of insufficient quality to be diagnostic. The interobserver correlation was good to excellent (kappa values of 0.57–0.83), while sensitivity and specificity of MRA for PE for the remaining 27 patients varied between 75% and 100%, and 95% and 100%, respectively (43). Some of the criticisms of this study were possible selection bias (no subsegmental emboli were diagnosed) and the fact that many patients with suspected PE would not be able to hold their breath for 27 seconds.

An alternative approach is to use MRA in conjunction with other diagnostic tests. Thus, using perfusion lung scintigraphy to safely exclude PE seems a cost-effective and sensible approach in terms of availability of diagnostic tests. One study took place in 36 consecutive patients with intermediate- or low-probability lung scan result and high clinical suspicion (65). Patients underwent pulmonary angiography and MRA, and assessment of these investigations was independent. A total of 19 emboli were demonstrated in 13 patients by angiography, and MRA diagnosed 12 patients as PE (one false positive, specificity of 96%) and missed two cases (sensitivity of 85%). Both missed PEs were isolated and subsegmental in location.

A more recent study performed contrast-enhanced pulmonary three-dimensional MRA in 141 consecutive patients with an abnormal perfusion lung scintigram and compared the findings with pulmonary angiography (44). A double-contrast injection was employed with a 20-second breath hold for each lung. MRA was contraindicated in 13 patients (9%), while images were not interpretable in eight patients (6%). MRA could be performed in two patients in whom conventional pulmonary angiography was contraindicated. Thus, both MRA and pulmonary angiography were available in 118 patients (84%). Figure 3 shows a comparative study of pulmonary angiography and MRA in a patient with PE. The prevalence of PE was 30%, partially as a result of exclusion by normal scintigraphy as initial test. Images were read independently in 115 patients, and agreement was obtained in 105 cases (91%), with a kappa value of 0.75. MRA demonstrated 27 of 35 patients with confirmed emboli for an overall sensitivity of 77%. The sensitivities for isolated subsegmental, segmental, and central/lobar PE were 40%, 84%, and 100%, respectively. MRA demonstrated emboli in two patients with a normal angiogram for a specificity of 98%. These latter two studies demonstrate the difficulty of diagnosing isolated subsegmental PEs, while the sensitivity for segmental or larger emboli is high (44, 65). Isolated subsegmental emboli are relatively uncommon, occurring between 6% and 15% of cases (8, 44, 65, 66). These results are comparable to the early (single-slice helical) CT data in the literature (17, 23).

Figure 3. Conventional pulmonary angiogram of right (a) and left (b) pulmonary arteries in patient with PE. The corresponding multiplanar reformatted images of a gadolinium-enhanced three-dimensional MRA are shown for the right (c) and left (d) sides.

Most recently, the use of a short TR (1.64 msec) three-dimensional gradient echo sequence on a high-performance gradient system (1.5 T, amplitude of 40 mT/m, slew rate of 200 mT/m/msec) has resulted in a gadolinium-enhanced three-dimensional MRA sequence that can be performed in under four seconds per image data set, with five data sets acquired within a 19-second breath hold (45). The sequence has been tested in three healthy volunteers and eight dyspneic patients and demonstrated emboli in all four subsequently confirmed cases, while those without emboli were also adequately identified. A problem was that only two of the eight patients could hold their breath for 19 seconds, while all patients could hold their breath for eight seconds, during which time two data sets were acquired.

No reports using parallel acquisition techniques for three-dimensional contrast-enhanced pulmonary MRA in a large patient sample are available. Improved image quality and smaller degradation from artifacts were reported in a comparative study in 20 patients using time-resolved pulmonary MRA with SENSE and electrocardiogram (ECG)-gated MRA without SENSE (50). In another study in 12 patients with known or suspected pulmonary arterial hypertension, high-resolution MRA with a 512 matrix (0.8 × 1.0 × 1.5 mm voxel size) was acquired in a 20-second breath-hold time by using partially parallel acquisition techniques (51). With this technique an excellent visualization of subsegmetal vessels was possible in all cases, and the results of the MRA highly agreed with those of conventional pulmonary angiography. Despite these promising results, the diagnostic benefit over nonparallel MRA has yet to be established in a larger patient sample.

It is concluded that three-dimensional gadolinium-enhanced MRA may offer a fast, reliable test for the diagnosis of PE, but further studies will be required to confirm this. In particular, management studies will need to demonstrate the additional value of MRA, especially in terms of cost-effectiveness. The use of parallel MRI techniques might improve pulmonary MRA by improving the spatial end temporal resolution. The use of blood pool agents to image the entire venous system may be useful in the future.

MR DIRECT THROMBUS IMAGING

Direct thrombus imaging is based on the principle that blood undergoes predictable change as it clots. One intermediate product of this process is methemoglobin, which causes significant reduction in T1, and thus causes a high signal when imaged using a heavily T1-weighted sequence (67, 68). This endogenous contrast can be exploited to allow the detection of subacute thrombosis directly without the need for gadolinium-based contrast agents. This technique therefore holds promise for the detection of PE and DVT (69, 70). One advantage of the technique is that the presence of high signal is only seen with subacute thrombosis and therefore can help to differentiate between old and new clots. Initial pilot studies to detect DVT employed an MRI sequence consisting of a T1-weighted magnetization-prepared three-dimensional gradient echo sequence (71). A water-only excitation radio-frequency (RF) pulse was administered, and the effective inversion time was chosen to nullify the blood signal. Using a body coil, two imaging blocks resulted in coverage from ankle to inferior vena cava, with each block requiring 3½ minutes acquisition time (Fig. 4).

Figure 4. MR direct thrombus imaging. a: Normal findings in pelvis and proximal legs. b: High signal corresponding to thrombus in proximal deep venous system. c: Further acute clot demonstrated in the proximal calf veins (notice that this is a case of bilateral DVT).

Subsequently, a study applied this technique in 101 patients with suspected DVT and compared it with conventional contrast venography (72). Although the cohort was not a consecutive group of patients, the authors prevented selection bias by including all patients with a positive venogram and randomly selecting 25% of patients who had a normal venogram result. Images were reviewed independently, and sensitivity and specificity ranged from 94% to 96% and from 90% to 92%, respectively. The interobserver agreement was formally assessed, and kappa values of greater than 0.8 were obtained. As could be expected, the diagnostic accuracy for isolated calf vein thrombosis was slightly lower than that for ileofemoral thrombosis, yielding a sensitivity range of 83% to 92%.

MR direct thrombus imaging has also been applied to the direct detection of PEs (69). This has either employed a two-dimensional or three-dimensional breath-hold acquisition. Early studies displayed the feasibility of this technique and its accuracy when applied in the clinical setting (Fig. 5). A recently concluded investigation employed MR direct thrombus imaging in the legs and chest of patients suspected of PE. Patients were randomized to the MRI study; other management strategies included 1) lung scintigraphy, followed by ultrasonography of the deep venous system of the leg if the lungscan is nondiagnostic; 2) same as strategy 1, but with pulmonary angiography as final diagnostic test if ultrasonography is normal; and 3) helical CT-pulmonary angiography (73). Outcome rather than comparison with gold standard tests was used as the end point.

Figure 5. MR direct thrombus imaging demonstrating a central PE in the left main pulmonary artery.

The results of the study were recently presented, with 157 patients assigned to the MRI arm of the study (74). Twenty-one percent of patients did not undergo scanning. During follow-up, eight patients died (none due to thromboembolic disease) and no episodes of recurrent disease were encountered in the negative group, nor were there incidences of bleeding in the positive group. PE was detected directly in the positive group in 19 patients and by the presence of DVT alone in 13. The outcome of the MR direct thrombus imaging management protocol was similar to that of the other management strategies. If these results can be reproduced, it would show that MR direct thrombus imaging could replace other technologies in the management of venous thromboembolic diseases and provide a comprehensive imaging technique.

PERFUSION IMAGING

Lung perfusion imaging was first developed using phase-contrast techniques, but this only allowed quantification of the left and main pulmonary arteries and was prone to errors due to flow changes and motion (75). Subsequently, the technique was developed using a variety of other sequences, which were capable of probing smaller pulmonary vessels (76–80). First, one can use gadolinium-based contrast agents and obtain perfusion information by assessment of the first-pass effect (Fig. 6). Second, one can use blood pool agents, which circulate for prolonged periods, and perform repeated data acquisitions of the pulmonary arterial circulation. Third, several noncontrast techniques may be employed invariably, using inversion recovery pulses for arterial spin labeling, followed by half-Fourier single-shot fast spin echo sequences to image the inflowing blood (81, 82).

Figure 6. Left top to right bottom: Time-resolved MR subtraction angiography in a patient with chronic thromboembolic pulmonary hypertension. The first image is the subtraction image. Notice inflow and progressive perfusion with virtual absence of perfusion in the left lung. The final frame shows the corresponding gadolinium-enhanced three-dimensional MRA demonstrating absence of vasculature in the left lung.

The contrast-based perfusion technique is usually combined with three-dimensional MRA, giving a dynamic enhancement of the pulmonary vascular tree similar to pulmonary angiography, which can be time-resolved to demonstrate enhancement changes during the first pass of the contrast bolus. Thus, perfusion-deficient areas can be shown, and this can aid in the visualization of PE after reconstruction of the pulmonary arterial tree. A recent publication using gadolinium-enhanced perfusion MR of the lung showed that semiquantitative evaluation of regional pulmonary perfusion may be achieved for prediction of postoperative lung function (77). This technique used a 2-mL bolus injection followed by saline flush at a rate of 3 mL/second. Three dynamic imaging sets consisting of a multislice two-dimensional inversion recovery gradient echo of six to seven seconds each were obtained during a single breath hold. The initial imaging set was obtained prior to contrast injection, while the following two were obtained seven seconds after start of bolus injection (pulmonary arterial phase) and 13 seconds after start of bolus injection (aortic phase). The technique was applied to 20 patients with potentially surgically resectable lung cancer. For comparison, pulmonary perfusion scintigraphy and pulmonary function tests were obtained, and the patients' outcomes were observed. The results showed good correlation between MR perfusion and lung scintigraphy, while MR perfusion also seemed able to predict postoperative outcome. Another study in patients following lung transplant surgery or lung reduction surgery used an adapted three-dimensional time-of-flight technique, which was adapted for imaging of both lungs during gadolinium contrast passage (three-dimensional MRA with double VUSE RF pulses) (36, 83).

Several additional studies have been performed, which have added to the interpretation of perfusion MR studies. One study compared the interobserver variation of MR perfusion and scintigraphic perfusion-ventilation studies in 20 patients with PE, 11 patients with acute pneumonia, and 13 patients with chronic obstructive pulmonary disease (COPD) exacerbation (84). This study showed that the final results were closely matched, but that the interobserver variation was significantly better for MR than for lung scintigraphy.

The effects of breath holding itself has an influence on pulmonary perfusion as demonstrated in 16 patients, of which 12 were suspected to have acute PE (78). The authors compared a two-dimensional sequence with 1-second image acquisition time for 25 images with a three-dimensional sequence with five series of five seconds each. Patients were asked to hold their breath (10 patients) or breathe shallowly if they couldn't hold their breath (six patients). The findings were compared with perfusion lung scintigraphy. The results showed that in the group who could hold their breath, both sequences gave similar results, whereas in those who couldn't hold their breath, the two-dimensional sequence was more sensitive and specific for demonstration of perfusion defects. A study by the same group assessed the influence of injection rate on enhancement profiles of pulmonary parenchyma in 15 healthy volunteers (79). With increasing injection rates the peak enhancement occurred earlier, albeit that it did not affect the SNR. However, the enhancement in the aorta also occurred earlier, which could have ramifications in patients with pulmonary diseases and greater perfusion through the bronchial arterial system. It was possible to separate the pulmonary arterial perfusion from the systemic arterial perfusion using an injection rate of 3 mL/second.

Quantification of perfusion is of importance if one is to use this technique for patients' management and follow-up of treatment effects. A technique using an ultrashort TE inversion recovery two-dimensional turbo FLASH sequence was developed (85). Subsequently, this technique was shown to be successful in healthy volunteers and quantitatively validated against colored microsphere infusions in a porcine model (86).

The influence of gravity on lung perfusion is another important parameter to be considered. A study compared perfusion as obtained in six subjects in a coronal plane and six subjects in a sagittal plane to address this issue (87). The apex of the lung showed faster transit time of contrast and decreased blood volume and blood flow compared with central portions of the lung. Furthermore, the transit time was greater in gravity-dependent lung regions.

Despite these promising results of first-pass contrast-enhanced perfusion MRI of the lungs, several limitations are obvious. As the lung, unlike other organs, is characterized by a very short circulation time, perfusion MRI requires rapid imaging in order to visualize the peak enhancement of the lung parenchyma. In addition, MRI of the lung perfusion also requires a high spatial resolution and large anatomic coverage in order to reliably visualize small perfusion defects in the lung periphery, as observed in patients with PE. Although major improvements of the gradient hardware have been achieved over the last few years, previous studies dealing with MRI of lung perfusion were mostly limited to two-dimensional MRI (85, 86, 88). Whenever three-dimensional MRI has been attempted, the temporal and/or spatial resolution was insufficient to resolve subtle temporal changes in blood flow (76–79). Similar to contrast-enhanced MRA, it can be expected that parallel MRI will improve contrast-enhanced perfusion MRI of the lungs in terms of both spatial and temporal resolution (Fig. 7) (76–80).

Figure 7. Three-dimensional perfusion MRI with parallel imaging techniques in a patient with a central lung cancer. Left top: Coronal HASTE. Right top: Conventional planar radionuclide perfusion scintigraphy. Bottom: Coronal and sagittal three-dimensional perfusion MR images. A perfusion defect due to the bronchus carcinoma is demonstrated.

Blood Pool Agent Techniques

Several types of blood pool agents have been developed, such as macromolecules that remain in the circulation or ultrasmall particles of iron oxide (USPiO). Although these agents have the advantage of prolonged imaging data acquisition, the disadvantage is that there is more overlap of arterial and venous structures. One of the first attempts in animal studies used gadolinium-DTPA-polylysine (88), and this showed good visualization down to sixth-order vessels and an increase in SNR of 120% (89). Other compounds that were tested consisted of albumin-binded gadolinium and different polymers of gadolinium, but none of these compounds were tried in human experiments (90–93).

More recently, two compounds have been studied in animals and subsequently in phase III clinical trials for cardiovascular MRA: NC100150 and MS-325 (94–97). One of these compounds was applied to perform complete venographic MR of pulmonary arteries and the deep venous system of the legs with good results in a small number of patients (98).

Noncontrast Techniques

Arterial spin labeling techniques enable us to distinguish flowing blood from stationary tissue and thus give information on lung perfusion. Two methods have been developed for pulmonary perfusion studies, those that use precursory inversion recovery RF pulsed sequences followed by single-shot fast spin echo imaging (81, 82), and those that use the concept of adiabatic fast passage with precursory pulsed field gradients to introduce continuous arterial spin labeling (99, 100). The RF pulsed labeling technique has recently been used in healthy volunteers (101) and more recently in a group of nine healthy volunteers and three patients with proven PE (102). This latter study also showed the effects of the respiratory cycle, demonstrating greater SNR at end-expiration. Furthermore, a comparative study in rabbits of first-pass gadolinium MR lung perfusion and the FAIRER technique showed similar results, albeit that the gadolinium technique provided a higher SNR (103). Another study used a single-shot half-Fourier turbo spin echo sequence (HASTE) in healthy volunteers to assess perfusion (82). This sequence applied an inversion pulse to the right ventricle and pulmonary artery, and after an inflow period, the HASTE sequence was run. More peripheral perfusion can be demonstrated by increasing the inflow time. This technique has not yet been used for the demonstration of PE. Contrast variation in pulmonary vessel intensity has also been observed in non-inversion recovery half-Fourier single-shot fast spin echo imaging (104). To minimize cardiac-dependent signal losses caused by the half-Fourier time lag between excitation and the sampling of the center of k-space, it was found that imaging should be performed after systole but before rapid filling of the ventricles. Finally, an ECG-gated fast spin echo sequence has been developed and tried in dogs with experimental pulmonary arterial and bronchial obstruction (103).

The gradient spin labeling technique was applied in eight healthy volunteers (100). This technique demonstrated the capability to distinguish between resting and postexercise blood flow changes in the pulmonary circulation. One patient with chronic obstructive pulmonary disease showed flow abnormalities.

PERFUSION-VENTILATION IMAGING

Ventilation lung MRI has become feasible using two techniques: through the use of inhaled contrast agents, such as aerosolized gadolinium and hyperpolarized noble gases, or through the use of subtraction oxygen-enhanced MRI. The advantage of ventilation imaging is not only related to direct lung function assessment, but it has also become possible to correlate perfusion and ventilation of the lung. Thus, PE is an area of particular interest, as its direct effect is a (segmental) reduction of perfusion without directly affecting ventilation (physiological responses and infarction will have a secondary effect).

Aerosolized gadolinium was combined with intravenous blood pool agents in rats (105). The model consisted of experimental PEs or large airway obstruction, and the technique was able to distinguish between them. The pulmonary parenchymal signal intensity increased by 70% due to the aerosol and a further 300% by the blood pool contrast agent. Another study used pigs and a small-particle generator, and demonstrated a signal-to-noise increase in excess of 100%, but with a time reduction of 30% (106). Finally, a study in dogs used an ultrasonic nebulizer to produce gadolinium aerosol ventilation MRI followed after 10 minutes by gadolinium-enhanced perfusion MRI (107). This study confirmed findings of those mentioned before. Initial human experiments in normal healthy volunteers were recently presented, which demonstrated homogeneous ventilation (108).

The introduction of ventilation imaging using hyperpolarized noble gases, such as 3-helium, has introduced another method to perform perfusion-ventilation imaging. One recent overview gives a good outline for technical requirements and the initial results in animals and human studies (109). Another overview focuses on the potential role within the field of lung functional imaging (110).

The feasibility of combined hyperpolarized 3-helium MRI and proton MR perfusion imaging of the lungs was initially shown in a rat model (111). A study by the same group used hyperpolarized 3-helium MRI in combination with intravenous superparamagnetic iron oxide nanoparticle perfusion MRI in a PE/airway obstruction rat model (112). Relative pulmonary blood volume maps could be superimposed on the ventilation images to demonstrate perfusion-ventilation mismatch in rats, which had PEs induced through air injection. Another study applied hyperpolarized 3-helium MR and both gadolinium-enhanced perfusion MR and gadolinium-enhanced MRA in a pig model with PE or large airway obstruction (113). It was possible to demonstrate perfusion defects with PEs, combined perfusion-ventilation defects with airway obstruction, and embolus location using MRA (Fig. 8). Very recently, a study combining hyperpolarized 3-helium MR and inversion recovery arterial spin labeling was performed in three normal volunteers, two patients following single-lung transplant for emphysema and one patient with PE (114). This showed perfusion and ventilation abnormalities in native emphysematous lungs and perfusion defects with normal corresponding ventilation in the patient with PEbolism.

Figure 8. Hyperpolarized 3-helium MR static ventilation image (a) and corresponding gadolinium-enhanced three-dimensional MRA (b) in a patient with chronic thromboembolic disease. Notice normal ventilation in presence of extensive obstructive pulmonary vascular disease.

Oxygen-enhanced ventilation imaging was developed around the same time as the application of hyperpolarized noble gases (115). The technique was subsequently successfully combined with gadolinium-enhanced MR perfusion imaging in a pig model of airway obstruction and/or PE (116). Oxygen-enhanced ventilation and arterial spin labeling perfusion imaging of the lungs was applied in 20 healthy volunteers, with the combination of both techniques in 10 volunteers (117). This demonstrated the feasibility of ventilation-perfusion MRI in humans. More recently, the technique was used to study 16 patients with a variety of lung diseases, including nine with PE (76). The results were encouraging, and regional perfusion deficits without ventilation abnormalities were shown in all patients with PE.

CONCLUSIONS

MRI of the chest, and in particular PE, is rapidly evolving, and the inherent advantages of noninvasiveness, nonionizing radiation requirement, the use of safer (or complete absence of) contrast agents, and the versatile sequences employed to assess various tissues in the chest render it of potential value for the future. However, it will take additional studies before one can advocate introduction in a clinical context of PE. In addition to technical considerations, the availability of MRI remains problematic in most hospitals. Thus, extension of services to include an acute disease like PE will only be possible with additional MRI time becoming available. However, as more systems are coming on-line, and with new laws on the medical use of ionizing radiation, it seems essential to develop this technology for use in the short term for those that have adequate MR resources and in the longer term for those that will have more MR capacity later. It should be emphasized that all work thus far has been performed on high-gradient, high-field-strength (1.5 T) systems. Additional work on low-field systems and the wider introduction of 3-T body systems is expected with interest. The use of gadolinium and hyperpolarized 3-helium could result in implementation of low-field-strength systems, while initial results on 3-T systems has shown capability of improved functional imaging within the chest. Furthermore, the introduction of blood pool agents and novel contrast techniques may change the way in which venous thromboembolic disease is imaged in the future.