Introduction

Due to its mostly unspecific clinical presentation, pulmonary embolism (PE) is often referred to as the great masquerader and remains a diagnostic challenge. PE is a common disorder occurring in some 600,000–630,000 patients per year in the United States [1–3]. It has been estimated that the diagnosis of PE will not be established in approximately 70% of patients who survive the initial thromboembolic event, accounting for 400,000 episodes per year. Of these patients, 120,000 will die from the disease (Fig. 1). The clinical scenario is crucial for assessing the likelihood of PE. Accordingly, distinct diagnostic algorithms are needed to assist clinical assessment and optimize the use of diagnostic tests, especially in an emergency department setting.

Fig. 1
figure 1

Incidence of pulmonary embolism in the United States (after [1]).

Full size image

A chest radiograph is usually the first imaging study performed in patients with suspected PE [4]. However, conventional X-rays are of limited value in this respect [5]. For a long time, pulmonary catheter angiography has been advocated as the gold standard in the diagnostic work-up of suspected PE. Although the accuracy of classical pulmonary angiography is beyond dispute, only a small percentage of patients undergo study—even patients in whom diagnosis of PE was not adequately confirmed or excluded after ventilation/perfusion (V/Q) scintigraphy [6, 7]. Practically, this standard of reference has not been widely accepted [8].

Nuclear medicine V/Q scintigraphy has for long been the clinical mainstay for the evaluation of suspected PE. This traditional imaging test technique is noninvasive, inexpensive, and readily available in the acute care setting. It has a high sensitivity, which is why it remains a primary PE screening technique [9, 10]. However, the high sensitivity of perfusion scanning is associated with a relatively low specificity [11]. Furthermore, V/Q scintigraphy has a poor spatial resolution and provides only indirect evidence for PE [12]. Venous thromboembolism is not directly visualized but rather its effects on perfusion and ventilation [13]. These problems cause the need for probability criteria, categorized as high, intermediate, low, or very low probability and normal [14]. The main practical problem is evident in the large group of patients with intermediate or indeterminate probability when the classic PIOPED criteria are applied [15].

Single-slice CTA

Spiral CTA emerged in the 1990s as a new diagnostic technique. The basis of PE assessment on CTA is the direct visualization of clot material in the pulmonary arteries. The first CTA protocols with 5-mm-thick sections were suitable for visualization of the central (and segmental) pulmonary arteries. However, initial data suggested that this technique was not ideally suited as a first-line imaging tool: Drucker et al. [16] found an interobserver variation of 53–60% (sensitivity) and 81–97% (specificity) for diagnosing acute PE using pulmonary catheter angiography as the standard. Other CTA studies also had less than ideal sensitivity rates, especially on the subsegmental level [17, 18]. It is now known that 5-mm-thick sections are inaccurate for PE diagnostics [19, 20].

Despite a variety of scanning parameters, sophisticated protocols optimize the trade-off between effective slice thickness and the area covered. Thin collimation spiral CTA is mandatory for assessing peripheral PE [21]. However, only the central 10–15 cm of the pulmonary vasculature can be assessed within a single inspiration breath-hold acquisition.

Remy-Jardin et al. [22, 23] reported a higher sensitivity and specificity for thinner slices (3 vs 5 mm) using a high-pitch protocol (1.7). The main drawback of a 5-mm collimation is the partial volume effect, which may limit detection of subsegmental and segmental filling defects [24]. Therefore, the central portion of the thorax was examined within a 20-s inspiration breath-hold using 3-mm collimation protocols at 1-s tube rotation times (level of inferior pulmonary veins to aortic arch, z-coverage 10 cm). On the basis of a small prospective patient study with separate analysis of 20 different arterial territories, 3-mm collimated helical CTA provided good interobserver agreement for the main, lobar, and segmental level. Dedicated analysis of the subsegmental territories, however, had a high incidence of nonvaluable branches and poor interobserver agreement [25].

Use of subsecond single-slice spiral CT scanners (tube rotation times 0.75 s) further reduced the partial volume effect [24]. Selecting a 2-mm collimation and a pitch of 2.0 significantly improved evaluation of subsegmental pulmonary arteries on CTA examination [23].

In addition, radiation exposure of helical CT was lower than that of classical pulmonary angiography [26]. Resten et al. [27] performed measurements of radiation dose using a predefined standardized CTA acquisition protocol and compared the results with a “standard” selective pulmonary arteriogram, using an anthropomorphic phantom and thermoluminescent dosimetry. Average dose was 4.375 times lower for CT (6.4±1.5 mGy vs 28±7.6 mGy).

It has been suggested [12, 28–30] that spiral CTA should replace ventilation–perfusion radionuclide scanning as the initial test for screening patients with suspected PE by using advanced single-slice CT examination protocols with 2- and 3-mm-thick sections. In 1999, Anderson and Wells [31] reported spiral CTA to be both a more sensitive and specific test (at least for central PE), although there are much more data on the use of diagnostic algorithms with lung scanning. A number of comparative studies led to a new definition on the impact of spiral CTA in this respect. Both techniques are rated equally highly in the literature [11, 32–34]. Due to limitations for accurate diagnosis of isolated peripheral PE with single-slice spiral CT, CTA has not yet become accepted as the reference standard [35]. Patients should therefore undergo either lung scanning or single-slice spiral CT imaging, depending on availability and expertise (Fig. 2). However, CTA as a second procedure after an abnormal V/Q scintigraphy had limited value in the clinical setting and vice versa [36, 37]. A survey by Schibany et al. [38] on equipment availability in Austria showed that CTA was the imaging tool of choice in 56% of hospitals, with V/Q scintigraphy in second place (43%). However, scintigraphy was not available in many hospitals (scintigraphy 19%, CTA 54%). Where available, it was used for a large number of patients each year.

Fig. 2
figure 2

Diagnostic algorithm including single-slice spiral CT (after [11]).

Full size image

Multislice CTA

The introduction of multislice spiral CT was a milestone in CTA technology. Dual-section CT [39] and four-slice multi–detector row computed tomography (MDCT) [40–45] show promising results regarding PE diagnostics. Today, 16-slice MDCT is gaining in popularity because it allows a comprehensive evaluation in a short period of time [35]. Compared with single-slice CT, MDCT can more precisely delineate clots down to the subsegmental level: third subsegmental branches can be assessed with 4×2.5-mm collimation [40], and delineation of arteries down to the 5th and 6th order can be done with 4×1-mm collimation [42]. Using thin sections (approx. 1 mm) significantly decreases the number of arteries classified as indeterminate (approx. 70% less) [43] and improves interobserver agreement in detection of PE (1.25 vs 2.5-mm collimation) [44]. Combining such a dedicated examination protocol with additional 3D shaded-surface display reconstruction images allows precise anatomic analysis of peripheral pulmonary arteries [46].

Four-slice MDCT with 0.5-s rotation time and simultaneous acquisition of four slices already yields an up to eightfold increase in performance compared with 1.0-s single-slice spiral CT, which may be distributed over volume, time, and axial resolution [47]. Therefore, thin collimation CTA of the entire chest can now be done within a single breath-hold even in critically ill patients [45].

The overall image quality was rated as superior for four-slice MDCT compared with subsecond single-slice CT, allowing accurate interpretation down to the subsegmental arteries more often even in otherwise healthy patients (65% vs 10%) as well as in patients with underlying respiratory disease (47% vs 15.5%). Recently, using advanced tomographic imaging techniques, SPECT V/Q scintigraphy and four-slice MDCT both yielded an excellent and absolutely comparable diagnostic accuracy down to the subsegmental level [48]. CTA has been recommended as an initial imaging method, even for nonmassive PE. Technical parameters are becoming a more important issue in diagnostic algorithms (single-slice vs multislice) [49]. Effective dose equivalents according to ICRP 60 were calculated as 3.3 mSv (male patients) and 4.2 mSv (female) for CTA exams using four-slice MDCT technology (4×1-mm collimation, 120 kVp, 100 mAseff, z-coverage 293±26.8 mm for men, 282±27.8 for women) [50]. These dose equivalents are in accordance with another study performed by Kuiper et al. [51]. Using a 4×1-mm collimation protocol (120 kVp, 90–160 mAseff), effective dose equivalents were in the range of 2.2–6.0 mSv, based on the CTDI. In comparison to the dose–area product for pulmonary arteriograms, radiation dose was slightly lower for MDCT (average 4.2 vs 7.1 mSv, range 3.3–17.3 mSv).

In principle, retrospective ECG gating of the whole chest can now be done using MDCT during a single breath-hold. For PE assessment, reduction of motion artifacts might be advantageous, especially in the pericardiac arteries. However, no additional relevant diagnostic information was gained for four-slice MDCT. Longer scan times (35–42 s) and breath-holds make the procedure more difficult for the patient [52]. Even with eight-slice MDCT, ECG gating does not influence the diagnostic accuracy and interobserver agreement [53]. Furthermore, ECG-gated acquisitions require a considerably reduced (overlapping) pitch factor and usually higher tube–current time settings, which will result in a higher radiation dose to the patient. Therefore, ECG gating cannot currently be recommended for routine assessment of acute PE.

In daily clinical practice, standard MDCT with thin collimation—e.g., 4×1, 4×2.5, 8×1.25, 16×0.75 mm—has practically become the method of choice for imaging of the pulmonary circulation in patients suspected of having PE (Figs. 3 and 4) [35]. According to the updated guidelines of the British Thoracic Society, no further examination or treatment is needed for patients with a high-quality negative MDCT CTA [3, 49]. Moreover, the fast data acquisition and real-time image reconstruction allow for a distinct examination protocol. Because PE and deep venous thrombus are two different aspects of the same disease, a combined examination may be a valuable addition to the initial diagnostic algorithm [37]. Indirect CT phlebography (CTP) can therefore be an option for complete assessment of venous thromboembolism (Fig. 5). Several authors have stressed that the uncritical use of indirect CT venography may increase radiation exposure to the population [26, 50, 54]. CT venography should therefore only be performed in addition to CTA of the pulmonary arteries, if this was positive for PE and demonstration of the extent and localization of deep venous thrombosis has an impact on clinical management, as radiation exposure is quite high for MDCT (up to 9.3 mSv) [50, 54, 55].

Fig. 3
figure 3

Small subsegmental PE in the left anterior lobe artery. Small clot material can be delineated on CTA (white arrow), with corresponding areas of decreased aeration in the adjacent lung tissue (black arrows) (120 kVp, 100 mAseff; 16×0.75 mm; rotation time 0.5 s; reconstructed effective slice thickness 1 mm).

Full size image
Fig. 4
figure 4

Small subsegmental PE in the left anterior lobe artery. From the same raw data set, different effective slice thicknesses were reconstructed: a 0.75 mm, b 1 mm, c 2 mm, d 3 mm, e 4 mm. The embolus is more easily distinguished from partial volume artifact on the thin sections.

Full size image
Fig. 5
figure 5

Combined examination with a PE being displayed at the central and lobar level on CTA. b Additional indirect CT phlebography revealed venous thrombosis in the right external iliac vein.

Full size image

Accuracy

The clinical impact of isolated subsegmental PE has been a matter of controversy. On the one hand, small emboli may indicate a risk for recurrent, more significant emboli. Also, patients with underlying cardiorespiratory disease may be impaired. On the other hand, otherwise healthy patients with isolated subsegmental PE may not be at significantly increased risk for morbidity and mortality. The number of isolated subsegmental PEs falls in the range of 6–30% [56–58].

There is a recent trend to analyze the overall clinical accuracy of diagnostic tests. Although data on sensitivity and specificity will always be important, the clinical outcome of patients with negative diagnostics is even more important. A number of studies has shown that the risk of PE after an initial negative CTA is about 1%. Of 3,713 patients with a minimum clinical follow-up of 3 months, 44 recurrent cases of PE were documented. Interestingly, the examination technique itself did not have a significant impact in this regard. Studies were done with EBCT [59], single-slice CT [28, 60–70], and multislice CT [45, 71]. Secondary imaging rarely added clinically important findings, even for high-risk oncologic patients with initially negative CTA results [37].

Quantitative CT scores for PE are based on the overall amount of clot burden within the pulmonary vasculature and were adapted from angiographic indices. The degree of pulmonary artery obstruction may be relevant for patient treatment [72]. There is a good correlation with the clinical severity of PE [73–77]. The vascular obstruction index introduced by Qanadli et al. [74] was highly correlated with pulmonary artery obstruction on pulmonary angiography (r=0.867, p=0.0001; assessed by the Miller index [78]). It scores the level of obstruction (according to the anatomic subdivisions involved) and adds a weighting factor for the degree of vascular obstruction. This quantitative specific CT index showed an excellent concordance (CT: r=0.944, p<0.0001; pulmonary angiography: r=0.904, p<0.0001). Collomb et al. [77] suggested measuring four main criteria for assessment of the hemodynamic severity of PE: the vascular obstruction index, the diameter of the central pulmonary artery, the minimum diameter of the left ventricle, and the right ventricle to left ventricle ratio, respectively.

A pulmonary artery diameter of >30 mm indicates pulmonary artery pressure >20 mmHg [79], which is defined as acute pulmonary hypertension in this respect [80].

In spite of a direct demonstration of intravascular thrombi, secondary findings are common, such as wedge-shaped opacities within the adjacent lung parenchyma. CT can also identify characteristic heart changes, e.g., acute dilatation of the right heart as well as abnormal and characteristic interventricular septal shift, as the transseptal pressure gradient is reversed and the septum is bulged convex toward the left ventricle (Fig. 6) [81–84]. The consequence of severe PE on the heart can be accounted for in the following way: sudden obstruction of the pulmonary arteries results in increased vascular resistance, causing acute pulmonary hypertension and right ventricle afterloading [77, 85]. Collomb et al. [77] considered the right ventricle to left ventricle ratio as an easy assessable secondary sign for the severity of PE in everyday practice.

Fig. 6
figure 6

a Acute and subacute PE on the central, lobar, and segmental levels. b Distal to a, a pleural-based consolidation is seen (arrow). c Sixteen-slice MDCT also identified characteristic heart changes, with acute dilatation of the right heart, and abnormal and characteristic interventricular septal shift with bulging of the septum toward the left ventricle in this four-chamber view. d Additional indirect CT phlebography depicted bilateral femoral vein thrombosis (arrows).

Full size image

Depiction of cardiac and pulmonary function in combination with the quantification of pulmonary obstruction helps to grade the severity of PE for further risk stratification and to monitor the effect of thrombolytic therapy.

Demonstration of ancillary findings [86] and suggestion of the true alternate diagnoses are clinically extremely beneficial (Fig. 7) [54, 87–90]. Garg et al. [61] reported that CT provided either additional findings or alternate diagnoses in more than half of the examinations interpreted as negative for PE.

Fig. 7
figure 7

CTA for suspected PE: in addition to exclusion of PE, an extensive mediastinal tumor with pleural effusion is displayed in a axial, b coronal, and c sagittal orientation (arrows). A non-Hodgkin’s lymphoma was finally proven by 18-G CT-guided core biopsy.

Full size image

Recently, more generally applicable protocols have been introduced. CTA has been evaluated more comprehensively toward a one-stop shopping strategy, including assessment of deep venous thrombosis using indirect CT phlebography [40, 50, 54, 91–93], right-heart geometrical assessment [84], and interactive analysis of nearly isotropic data sets using advanced postprocessing methods [94–96].

Lung scintigraphy remains the first imaging modality where CTA is not available and for patients with contraindications for iodinated contrast media (anaphylaxis, renal failure), as well as in pregnant patients [3]. Real-time and contrast-enhanced magnetic resonance angiography [97–99] may represent a valuable alternative in this respect.

CTA is an appropriate initial test in patients with intermediate and high clinical suspicion of PE under emergency conditions [100]. In patients with a low clinical probability of PE, the most cost-saving strategy involves plasma D-dimer assessment, a degradation product of cross-linked fibrin [101, 102]. Due to its high sensitivity, D-dimer ELISA has been advocated as the first-line test for ruling out PE in outpatients, provided the assay has been validated in an outcome study [33]. D-dimer is very specific for fibrin. However, the specificity of fibrin for venous thromboembolism is poor. Fibrin is produced in a wide variety of conditions, such as cancer, inflammation, infection, and necrosis. Additionally, it is not useful in elderly patients or inpatients, as it is normal in less than 10% of patients [33].

From a radiological point of view, CTA has effectively become the de facto first imaging test in clinical routine, as patients with a high-quality negative CTA do not require further examination or treatment for suspected PE (Fig. 8). The 32-, 40-, and 64-slice MDCT, which are already available, will further improve the overall image quality, especially in the pericardial region (Figs. 9 and 10). We are likely to see further technical developments in CT technology in the near future.

Fig. 8
figure 8

Diagnostic algorithm including multislice spiral CT (after [3, 49]).

Full size image
Fig. 9
figure 9

Axial, sagittal, and coronal multiplanar reformats (MPRs) from a 64-slice MDCT CTA data set with central non-occluding saddle embolism and segmental and subsegmental occluding PE (online tube-current modulation; 100 kVp, 140 ref. mAs; 64 (2×32)×0.6 mm; resolution 0.4 mm, tube rotation time 0.33 s).

Full size image
Fig. 10
figure 10

Maximum intensity projection (MIP) from a standard 64-slice MDCT data set without ECG pulsing: Note the sharp and distinct anatomic detail of the pericardial region, without cardiac motion artifacts (120 kVp, 100 mAseff; 64 (2×32)×0.6 mm; resolution 0.4 mm, tube rotation time 0.5 s; total time for data acquisition 7 s for 350 mm).

Full size image

Nevertheless, our clinical partners who refer patients for diagnostic testing must still be convinced. Hopefully, not only the British Thoracic Society [49], but many others will create new standards of practice and release updates of existing guidelines.