Published ahead of print on November 21, 2002, doi:10.1164/rccm.200208-899OC
© 2003 American Thoracic Society
Pulmonary EmbolismComparison of Angiography with Spiral Computed Tomography, Magnetic Resonance Angiography, and Real-Time Magnetic Resonance ImagingDepartment of Diagnostic Radiology, University of Technology, Aachen; and Philips Research Laboratories, Hamburg, Germany Correspondence and requests for reprints should be addressed to Patrick Haage, M.D., Department of Diagnostic Radiology, University of Technology, Pauwelsstrasse 30, D-52057 Aachen, Germany. E-mail: haage{at}rad.rwth-aachen.de
In the last decade, spiral computed tomography (CT) and magnetic resonance (MR) angiography (MRA) have become a viable alternative to conventional angiography in the diagnosis of acute pulmonary embolism. However, patients with dyspnea are often unable to hold their breath for a longer time and thus image degradation is frequently observed. Consequently, an imaging sequence that allows free breathing is desirable. The aim of this animal study was to compare contrast-enhanced spiral CT, MRA and a real-time MR sequence, the latter without breath-hold, with pulmonary angiography as reference gold standard. Nine pigs with artificially induced pulmonary embolism underwent this multimodality comparison. All images were independently evaluated for the presence of pulmonary emboli by two reviewers. Forty-three filling defects were detected by conventional angiography on lobar and segmental levels. Sensitivity of CT images was 72.1 and 69.8% for Readers 1 and 2, respectively, and sensitivity of MRA images was 79.1 and 81.4%. With real-time MR imaging, however, the detection rate was 97.7% for both readers. We conclude that, under experimental conditions, real-time MR imaging without the use of radiation or iodinated contrast material is comparable with angiography in the detection of pulmonary emboli.
Key Words: pulmonary embolism • angiography • spinal computed tomography • magnetic resonance imaging Pulmonary embolism (PE) is a frequently observed cause of patient morbidity and mortality. Because this condition has nonspecific signs and symptoms, its adequate diagnosis relies on imaging techniques. In case of clinically suspected PE, four radiologic modalities may be employed: conventional digital subtraction angiography (1), contrast-enhanced spiral computed tomography (CT) (2), magnetic resonance (MR) angiography (MRA) (3), and real-time MR imaging with radial k-space scanning (4). The reference gold standard, conventional pulmonary angiography, is invasive and carries a risk of major complications of 1% and a mortality risk of 0.5% (5) and has thus in many institutions been replaced by spiral CT, which is more readily available with lower procedure-associated risks (6). Sensitivity of spiral CT is in the order of 90% for central, lobar, or segmental pulmonary emboli (7, 8). MR imaging uses even safer contrast agents than CT and does not involve radiation exposure. In addition, MR imaging allows for the depiction of perfusion and ventilation, which may further aid in the (differential) diagnosis of PE and thus appropriate patient care and treatment (9, 10). This diagnostic technique is improving constantly, and by contrast-enhanced breath-hold techniques, pulmonary vessel status can be visualized effectively in most of the cases (11). Although faster scanning techniques have permitted MRA to be performed in shorter time periods than before, MRA does not yield optimal image quality in people who are unable to hold their breath. A motion-compensated projection reconstruction sequence with radial k-space scanning may offer potential in these patients (12). Nevertheless, before applying the latter in a clinical setting, its effectiveness in the diagnosis of PE compared with accepted modalities has to be evaluated and confirmed. This was the purpose of the study presented herein; to the best of our knowledge, a comparative analysis of all available modalities including radial real-time MR imaging has not been performed yet.
The study protocol was approved by the Animal Care and Use Committee. Handling and caring of the laboratory animals for this study was performed according to state and institutional guidelines.
Thrombus Preparation
PE Model The procedures were performed in nine female pigs with a mean body weight of 40 kg. To avoid infections of the respiratory tract, each animal was treated with a total of four injections of amoxicillin (Duphamox; AHP AG, Zug, Switzerland) before study initiation. Preceding the procedure, the veterinarian involved performed a physical examination of the animals. The animals were fasted for 24 hours before the intervention. General anesthesia was induced with intravenous injection of 0.2 ml/kg body weight azaperone (Stresnil; Janssen-Cilag, Neuss, Germany) and 0.1 ml/kg body weight ketamine hydrochloride (Ketavet; Sanofia-Ceva, Düsseldorf, Germany) after intramuscularly applied premedication of 0.05 ml/kg body weight atropine (Atropin; Braun, Melsungen, Germany). Pentobarbital solution (Narcoren; Rhone-Merieux, Laupheim, Germany) diluted 1:3 with saline was injected as needed. The animals were intubated and mechanically ventilated with a volume-controlled ventilator (Sulla 808V; Dräger Medizintechnik GmbH, Lübeck, Germany) using a mixture of oxygen, nitrous oxide, and halothane (Halothan ASID; Rüsch, Böblingen, Germany). Respirator settings provided a rate of 12 breaths/minute and a tidal volume of 500 ml/breath. Throughout the procedure heart rate and blood oxygenation levels were continuously monitored. Under sterile conditions, a 16-F introducer sheath (Kimal, Middlesex, UK) was inserted into the right jugular vein and advanced into the proximal superior vena cava over a 0.035-inch J-shaped guidewire with movable core (Cook, Bjaeverskov, Denmark) after access via a surgical cutdown. A total of 50 artificial emboli were then injected through the sheath (range 4–6/animal), which subsequently embolized into the pulmonary arteries.
Pulmonary Angiography
Contrast-enhanced Spiral CT
Magnetic Resonance Angiography
Real-Time MRA with Radial Scanning
Image Interpretation and Data Analysis The accuracy of the reviewers' diagnoses for all three imaging techniques (spiral CT, MRA, and real-time MRA) was calculated. This incorporated detection of embolic defects that were not present (false positive) or their failure to identify emboli that were present (false negative). These results were matched up to the conventional angiography results, which served as the reference method. The sensitivity and positive predictive value of each imaging modality, as interpreted by each reader, was calculated. Because it is not possible to reliably measure the total number of unaffected vessels, the specificity (percentage of true negatives) was not determined. Assessment of differences in the detection rate for thrombi between spiral CT, MRA, and real-time MRA was done by the unpaired two-tailed Student t test. Probability (p) values were calculated for each comparison with a significance level of 0.05. Interobserver agreement was evaluated with the use of the kappa statistic, where k is 1 in case of perfect agreement, k is 0 if there is no agreement better than chance, and k is negative when agreement is worse than chance (16).
The surgical and experimental protocol was technically successful in all nine animals. Except for one case of transient tachycardia directly after PE induction, no significant changes in vital signs were observed. Forty-three out of the 50 emboli migrated from the superior vena cava into the pulmonary arterial tree, and 7 got trapped in the right atrium. The average duration of anesthesia was 6 hours. No episodes of adverse contrast medium reactions were noted. All conventional angiography (Figure 2) and CT (Figure 3) images were of diagnostic quality and could be included in the evaluation. Lung vessels could also be effectively depicted with the MRA sequence (Figure 4) and the real-time radial MR sequence (Figure 5) down to segmental and, to some extent, subsegmental levels.
Artificial thrombi were found in lobar and segmental arteries in all pigs. Forty-three affected vascular territories were counted by conventional angiography. Corresponding to the size of the emboli, all filling defects were contained on the lobar or segmental level. Subsegmental vessels were patent and free of thrombus due to their smaller diameter. In all, 51.2% of the thrombi were localized on the right side and 48.8% on the left. Sixteen of the 43 emboli (37.2%) were detected in lobar arteries; the remaining 27 had advanced into segmental arteries. The sensitivity of spiral CT images was 72.1% for Reader 1 and 69.8% for Reader 2 (mean 71.0%), and the sensitivity of MRA images was 79.1% for Reader 1 and 81.4% for Reader 2 (mean 80.3%) (Table 1) .
The analogous value for the real-time MR images was 97.7% for the two readers, which detected all emboli except for one left-sided segmental embolus. False positive diagnoses were not observed with either modality yielding a positive predictive value of 100%. Differences in the detection rate for emboli between spiral CT and MRA (p = 0.06) and between MRA and real-time MRA (p = 0.14) to determine the level of confidence were not statistically significant. Only the difference between spiral CT and real-time MRA was of statistical significance (p = 0.01). The inter-rater coefficient k was 0.86 for CT, 0.84 for MRA, and 1.0 for real-time MRA.
The results of the presented study show that in this pig model, real-time MRA with radial scanning is comparable with conventional pulmonary angiography with better performance than contrast-enhanced spiral CT and MRA for the visualization of pulmonary emboli. All investigated modalities have demonstrated good to excellent interobserver agreement. Clinically, requirements of an acceptable PE assessment are that it is feasible in most patients and leads to the proper diagnosis and therefore the appropriate therapy. In this context, angiography has long been recognized to be the most exact modality with a sensitivity and specificity of 95 and 100%, respectively (5, 17). Thus, it is considered the reference standard against which all other techniques have to be measured. Advantages include the option of inducing a local fibrinolysis-therapy via the angiographic catheter already in place. However, angiography is not possible in up to 20% of patients and due to its associated invasiveness, morbidity, and sporadic mortality it is nowadays being employed less frequently and replaced by spiral CT and recently MRA, both of which are faster, less invasive, less operator-dependent, and are associated with a smaller number of complications (5, 18). Another advantage of spiral CT is its widespread availability. Conversely, MRA has advantages over spiral CT because it does not involve ionizing radiation and it makes use of smaller amounts of gadolinium chelates, which have an even better safety profile than iodinated contrast agents (19, 20). Over the last few years, many of the limitations of earlier MR techniques were dealt with. High signal-to-noise ratios and high contrast between flowing blood and pulmonary filling defects can now be depicted with the development of contrast-enhanced MRA in a single breath-hold, during "first-pass" imaging with gadolinium-containing contrast media. The key aspects that restrict MRA application are nondiagnostic quality of images and alterations that avert a patient from entering a magnetic field, which together are in the order of 15% (21). These difficulties are a weakness of MRA, particularly when compared with spiral CT. Furthermore, the necessity to eliminate the need for breath-holding for the detection of pulmonary emboli in patients with dyspnea is beyond question for any imaging modality. Improved gradient performance, ultrafast reconstruction capabilities (22), and lately developed software tools for "on-the-fly" interactive MR scanning (23) now offer the potential for real-time MR imaging with satisfactory contrast and spatial resolution for pulmonary MRA. Herein, radial MRA showed a high contrast of the pulmonary artery lumen. Enhanced motion artifact suppression was achieved by combination of the real-time data acquisition with radial k-space filling, which offers better motion artifact suppression when compared with Cartesian readouts (14, 24). An additional advantage of radial k-space filling is an option for undersampling of radials, thereby further enhancing temporal resolution without sacrificing spatial resolution (25). Consequently, in contrast to Cartesian k-space filling, the radial acquisition spatial resolution is not negatively affected by timesaving undersampling. In addition, improved temporal resolution was achieved applying the sliding window technique (26). The present study was designed to assess the applicability of real-time MRI techniques in an established animal model, wherein the technique combination allowed for high-quality pulmonary vessel visualization, exploiting the continued shortening of longitudinal relaxation time due to the injected contrast agent. Bolus tracking for timing of the arterial contrast medium passage was not necessary. Mean image acquisition time was 9 minutes, whereby uncertain vascular territories, contrary to spiral CT and MRA, could be repeatedly evaluated. The high frame rate permitted data acquisition during free breathing and without cardiac triggering. Albeit the results show that real-time MRA is comparable with pulmonary angiography for the detection of pulmonary emboli in this animal model, some restrictions as to experimental setup must be reflected on before the results can be transposed to a clinical setting. These include differences between species that might influence lung anatomy as well as the use of artificial methyl methacrylate thrombi rather than "real" thrombi, which can only approximate the clot seen in clinical PE. The pulmonary anatomy and vessel size of the pig are not identical to the human anatomy, with one large main pulmonary arterial trunk and many smaller arteries rather than the dichotomous branching configuration of the human pulmonary vascular tree (27). The emboli were prepared so that they would migrate and lodge in branches of the major artery that are approximately the same size as human subsegmental pulmonary arteries (13). The sensitivity values should not have been affected by differences in anatomy because there is no rationale that these differences between pigs and humans should lead to improved accuracy of one technique. Principally, the use of real blood clots as thrombi would have been an ideal thrombosis model, but the comparability of different modalities requires constant size and position of the filling defects over at least several hours, which would not have been possible with real thrombi. The lower sensitivity values of spiral CT in our setup will certainly improve with the advent of multislice spiral CT (28). At this time, due to its availability and speed, spiral CT should still be considered the preferred diagnostic modality in suspected PE (29), in particular if or as long as patients in the intensive care unit are treated with non-MR–compatible equipment. In conclusion, it is not proposed that real-time MRA can be used to definitely rule out PE. The results however demonstrate its advantages and potential in comparison with CT and "conventional" MRA. Real-time radial MR scanning allows promising image acquisition of the pulmonary vasculature without respiratory control. Especially, patients with severe dyspnea may profit from this imaging modality because real-time imaging quality is not degraded by respiratory motion. Furthermore, technical improvements and clinical evaluations are warranted before applying this technique as an alternative to already established modalities in the diagnostic approach to PE. Received in original form August 20, 2002; accepted in final form November 12, 2002
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
