Paradoxical Embolism: Role of Imaging in Diagnosis and Treat- Ment Planning1

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights. PARADOXICAL EMBOLISM

1571 Paradoxical Embolism: Role of Imaging in Diagnosis and Treat- ment Planning1

Farhood Saremi, MD Neelmini Emmanuel, MD Paradoxical embolism (PDE) is an uncommon cause of acute Philip F. Wu, BS arterial occlusion that may have catastrophic sequelae. The pos- Lauren Ihde, MD sibility of its presence should be considered in all patients with an David Shavelle, MD arterial embolus in the absence of a cardiac or proximal arterial John L. Go, MD source. Despite advancements in radiologic imaging technology, Damián Sánchez-Quintana, MD, PhD the use of various complementary modalities is usually necessary to exclude other possibilities from the differential diagnosis and Abbreviations: DVT = deep venous thrombo- achieve an accurate imaging-based diagnosis of PDE. In current sis, IVC = inferior vena cava, PDE = paradoxical embolism, PFO = patent foramen ovale practice, the imaging workup of a patient with symptoms of PDE usually starts with computed tomography (CT) and magnetic RadioGraphics 2014; 34:1571–1592 resonance (MR) imaging to identify the cause of the symptoms Published online 10.1148/rg.346135008 and any thromboembolic complications in target organs (eg, Content Codes: stroke, peripheral arterial occlusion, or visceral organ ischemia).

1From the Departments of Radiology (F.S., Additional imaging studies with modalities such as peripheral N.E., P.F.W., L.I., J.L.G.) and Cardiovascular venous Doppler ultrasonography (US), transcranial Doppler US, Medicine (D.S.), University of Southern Califor- echocardiography, and CT or MR imaging are required to detect nia, USC University Hospital, 1500 San Pablo St, Los Angeles, CA 90033; and Department of peripheral and central sources of embolism, identify cardiac and/ Human Anatomy, University of Extremadura, or extracardiac shunts, and determine whether arterial disease is Badajoz, Spain (D.S.Q.). Recipient of a Cer- tificate of Merit award for an education exhibit present. To guide radiologists in selecting the optimal modalities at the 2012 RSNA Annual Meeting. Received for use in various diagnostic settings, the article provides detailed January 3, 2013; revision requested April 4 and received July 13; accepted July 19. For this information about the imaging of PDE, with numerous radiologic journal-based SA-CME activity, the authors, and pathologic images illustrating the wide variety of features that editor, and reviewers have disclosed no relevant may accompany and contribute to the pathologic process. The relationships. Address correspondence to F.S. (e-mail: [email protected]). roles of CT and MR imaging in the diagnosis and exclusion of PDE are described, and the use of imaging for planning surgical SA-CME LEARNING OBJECTIVES treatment and interventional procedures is discussed. After completing this journal-based SA- ©RSNA, 2014 • radiographics.rsna.org CME activity, participants will be able to: ■■Describe the causes of PDE and sequelae in target organs. ■■Discuss the specific uses of various im- Introduction aging modalities in the diagnosis of PDE. Paradoxical embolism (PDE) is usually definitively diagnosed at ■■Recognize CT and MR imaging fea- autopsy or at radiologic imaging when a thrombus that crosses an tures that are pertinent for the diagnosis of PDE and for posttreatment evaluation. intracardiac defect is seen in the setting of arterial embolic damage in end organs (eg, stroke). Imaging evaluation of patients in whom See www.rsna.org/education/search/RG. the presence of PDE is suspected usually necessitates the use of more than one modality. Peripheral Doppler ultrasonography (US) and echocardiography are well-established methods for assessing thromboembolic processes. Although echocardiography is the prime modality for depicting a shunt across a patent foramen ovale (PFO), no single modality can cover the whole spectrum of findings in the imaging workup of PDE. 1572 October Special Issue 2014 radiographics.rsna.org

The article outlines the optimal imaging approach in various clinical settings and the value Table 1: Essential Elements of PDE contributed by each imaging modality for accu- Systemic embolism confirmed by clinical, an- rate diagnosis of PDE. The current roles of com- giographic, or pathologic findings without an puted tomography (CT) and magnetic resonance apparent source on the left side of the heart or in the proximal arterial tree (ascending aorta) (MR) imaging in identifying cardiac and extra- Embolic source within the venous system cardiac abnormalities known to contribute to the development of PDE and detecting sequelae in Abnormal intracardiac or intrapulmonary communication between the right and left circulations target organs are emphasized, and the utility of Pressure gradient that promotes a right-to-left shunt supplemental US studies is reviewed. Strategies at some point during the cardiac cycle for treating PDE, including interventional techniques, also are described.

Historical target organs. Additional imaging studies, includ- Background and Definitions ing peripheral venous Doppler US, transcranial In 1877, Cohnheim (1) reported the first case Doppler US, echocardiography, and CT, are used of PDE by describing the path of an embolus to detect peripheral and central sources of embo- through a septal defect in the heart. In 1881, lism, arterial disease, and cardiac or extracardiac Zahn (2) reported an autopsy study in which shunts. Further diagnostic testing often includes thrombosis of the uterine vein, multiple systemic continuous long-term electrocardiographic re- emboli, and a branched thrombus within a PFO cordings, blood chemistry panels, and coagula- were seen in the same cadaver. Later, in 1885, he tion tests. used the term paradoxical embolism to describe a condition in which emboli derived from the ve- Types of Embolism nous system reached the systemic arterial system Thrombi from tributaries of the IVC are the through an abnormal communication between major sources of embolism, but emboli of fat, the heart chambers (3). air, amniotic fluid, and tumor tissue have also Four essential elements contribute to the devel- been described (10–15). Fat embolism syndrome opment of PDE: systemic embolism, an embolic is primarily a pulmonary disease (10). Shunt- source, a right-to-left shunt, and a pressure gradi- ing of fat or other material across a PFO can be ent across the shunt (Table 1) (3–6). The diagnosis precipitated by increased right atrial pressure for of PDE is considered definitive when it is based on a variety of reasons, including changes in body a finding at autopsy or at imaging of a thrombus position, breathing patterns, and intrathoracic that crosses an intracardiac defect in the setting of pressure. Paradoxical air embolism can lead to an arterial embolus (4). A diagnosis of PDE in the cerebral lesions in scuba divers (11). Cerebral absence of these findings is considered presump- air embolism can occur through central venous tive (4,6). The triad of systemic embolism, venous catheters (12). Patients undergoing neurosurgery thrombosis, and intracardiac communication in a sitting position have a risk for paradoxical defines the clinical diagnosis of PDE and allows air embolism (13). In these cases, preoperative treatment with a high level of confidence (7,8). detection of PFO and additional monitoring and The diagnosis of PDE is termed “possible” if an special care during surgery are advised. Amni- arterial embolus and PFO are detected; many phy- otic fluid embolism can rarely be complicated sicians treat patients on the basis of a diagnosis of by PDE resulting from increased pressure in the “possible PDE” (9). right side of the heart due to the release of vaso- Most early case reports of paradoxical embolus active substances when amniotic fluid enters the were based on autopsy findings (4). Later, an in- pulmonary circulation (14). tracardiac right-to-left shunt was demonstrated in Imaging findings of PDE complications in the a living patient when dye injected into the inferior brain are probably similar for different types of vena cava (IVC) appeared earlier than expected embolism, and the clinical history is important at the left brachial artery (6). Limited catheter- for final diagnosis. Air emboli absorb quickly and ization of the right side of the heart was proposed are best depicted in an early stage at CT. as a method for excluding an intracardiac shunt in patients with coexistent venous thrombosis or Peripheral Sources pulmonary embolism and arterial embolism. of Venous Thromboembolism In current practice, the imaging workup of a Venous thrombosis in the legs may be the most patient for PDE usually starts with CT and MR common source of embolus. Approximately 90% imaging. These modalities are used to diagnose of symptomatic pulmonary emboli arise from thromboembolic sequelae of arterial embolism in thrombi located in the leg veins (8,16). In most RG • Volume 34 Number 6 Saremi et al 1573 studies, the prevalence of deep venous thrombo- venography for evaluation after a stroke, DVT sis (DVT) in patients with acute pulmonary em- was found within 3.25 days after the occurrence bolism appears to be higher than that in patients of a stroke in 27% of those with cryptogenic with a cryptogenic stroke and PFO (16,17). In brain ischemia and an interatrial communica- many cases of PDE, the source of the embolus in tion, half of the thrombi being isolated within a peripheral veins cannot be found (8). The report- calf or pelvic vein (25). In a related multicenter edly low rate of DVT in patients with a PFO and study, pelvic DVT was found at MR venography cryptogenic stroke may be an effect of the delay performed within 3 days after the occurrence between the initiation of anticoagulation therapy of a cryptogenic stroke in 20% of 46 patients and the imaging evaluation, complete thrombus with a PFO or atrial septal defect (26). With migration, inability to detect residual thrombus, the use of MR venography, Kiernan et al (27) or undetected thrombosis in a calf or pelvic vein found pelvic venous thrombosis (May-Thurner (10%) (8,18). Another possibility is that the syndrome) in 6.3% of patients who underwent embolic source remains undetected in the upper- PFO closure after a cryptogenic stroke. Eighty extremity veins (19). percent of the patients were female, and 54% Duplex US is the most common method for of the female patients were receiving oral con- evaluating DVT. Most US studies of the lower traceptive therapy. Overall, the results of the extremity are limited to veins at or above the level preceding studies show that (a) PDE from the of the popliteal veins, which may lead to underes- lower extremity and possibly the pelvis is one timation of the true incidence of venous throm- mechanism that accounts for ischemia related bosis (20). to systemic embolization in a subset of patients US is more accurate than venography for de- and (b) pelvic CT or MR imaging may be useful picting peripheral DVT but is much less accurate for determining whether pelvic DVT is present for showing central (ie, pelvic) DVT (20,21). A in patients in whom findings are negative for small proportion (2%–7%) of thrombi that can DVT of the lower extremities. be diagnosed at venography or CT venography Upper-extremity sources of PDE include are limited to the pelvic veins or vena cava and spontaneous DVT (Paget-Schroetter syndrome) may therefore remain undetected at US (22). and catheter-related DVT. The occurrence of Contrast material–enhanced MR venography PDE as a complication of Paget-Schroetter seems to be more accurate than color Doppler syndrome is rare, but at least one case has been US in depicting a central (toward the pelvis) reported (28). Catheter-related thrombosis ac- extension of DVT (23). Nonenhanced balanced counts for approximately 80% of cases of upper- steady-state free precession MR venography extremity DVT (29). Thrombogenesis associated is more accurate than US for the diagnosis of with catheters has been well documented, with lower-extremity DVT and is capable of depicting an incidence ranging from 2% to 67%, depend- greater central extension of the thrombus (24). ing on the catheter type and location, diagnostic Nonenhanced MR venography can be performed criteria, and population studied (29,30). The when intravenous administration of gadolinium- published literature about catheter-associated based contrast material is contraindicated. paradoxical thromboembolus is limited to case The reported incidence of DVT associated reports of coronary arterial, limb, or brain in- with PFO and PDE ranges widely between dif- volvement (31,32). ferent patient series (8,17,25), depending on the imaging modality used, anatomic location Sequelae of of the venous thrombus, time interval between PDE in Target Organs the onset of symptoms and imaging, and dura- Although PDE is an uncommon cause of acute tion of anticoagulation therapy before imaging. arterial occlusion, it can have catastrophic se- For example, Stöllberger et al (8) reported that quelae, and the possibility that it is present DVT was found at lower-extremity venography should be considered in all patients with an arte- performed within 90 days after symptom onset rial embolus in the absence of a cardiac or proxi- in 57% of patients with a PFO and arterial em- mal arterial source. PDE is frequently associated boli without evident arterial or cardiac sources. with cryptogenic stroke and peripheral embolism By contrast, in a study by Lethen et al (17), (33) (Fig 1). Uncommon complications include venography depicted DVT in only 10% of pa- brain abscess (34), decompression sickness in tients with a PFO as the sole identifiable cardiac underwater divers (11), myocardial infarction risk factor for PDE; most of those patients had (35), and mesenteric infarction (7). Hypoxemia undergone heparin therapy before venography. due to a transient right-to-left shunt is also pos- In a recent study of 37 patients who underwent sible. In Loscalzo’s (7) study based on findings duplex US of the lower extremity and pelvic MR in 30 patients, the five sites of arterial emboli 1574 October Special Issue 2014 radiographics.rsna.org

Figure 1. Axial diffusion-weighted MR images demonstrate paradoxical embolic infarcts. (a) Multiple bilateral nonterritorial subcortical infarcts (light gray foci) are seen in a patient with tetralogy of Fallot and an anomalous connection of a left-sided superior vena cava with the left atrium. (b) A single territorial infarct (arrow) that origi- nated from right subclavian venous thrombus is depicted in a patient with a PFO. (c) A large lobar infarct (light gray–whitish area) is evident in the right temporoparietal region in a patient with an atrial septal defect.

were peripheral (49%), cerebral (37%), coronary are cryptogenic in origin (38). The cause of cryp- (9%), renal (1%), and splenic (1%). Among cases togenic stroke remains undetermined in most of PDE reported by Travis et al (9), the most cases because the event is transitory or reversible, frequent clinical manifestations were (in order of investigators cannot look for all possible causes, decreasing frequency) lower-extremity ischemia, and some causes remain unknown. The detection upper-extremity ischemia, respiratory distress, of a PFO in a patient with a confirmed stroke does cerebral infarction or amaurosis fugax, and ab- not necessarily mean that the cause of the stroke dominal and/or flank pain. has been identified. Establishing a causal relationship between the presence of a PFO and the oc- Cryptogenic Stroke currence of a stroke remains the crucial point in Ischemic strokes can be classified into two ma- the diagnosis of PDE. The four criteria described jor categories: (a) those due to a known cause earlier for the diagnosis of PDE may not always such as large-artery atherosclerosis, intracardiac be met. The presence of other contributing fac- thrombus, or small-artery occlusion and (b) those tors, such as the morphologic characteristics of the due to an undetermined cause or cryptogenic PFO and associated structures, may increase the infarction (36,37). One-third of ischemic strokes probability that PDE is present (37,38). RG • Volume 34 Number 6 Saremi et al 1575

Table 2: Potential Routes of PDE (Right-to-Left Shunt) Intracardiac PFO Iatrogenic connection (baffle defect, Fontan conduit, Rashkind device) Enlarged thebesian veins (ie, interatrial muscle bundle) Congenital anomaly (atrial septal defect, unroofed coronary sinus, ventricular septal defect, atrioventricular septal defect) Extracardiac Pulmonary arteriovenous malformation (congenital, secondary to cavopulmonary shunts) Systemic to pulmonary venous communication (congenital, acquired) Arterioarterial communication (patent ductus arteriosus) or venovenous communication

Embolic and Nonembolic Infarcts to PDE. However, intracardiac causes are more PFO is thought to be an important causal mecha- common; of these, most arise from the presence nism of embolic stroke in young patients (38,39). of a PFO. In Loscalzo’s (7) series of cases with a Some investigators believe that various imaging clinical diagnosis of PDE, 72% had a PFO, and patterns can support a diagnosis of PDE (40). the remaining potential routes included atrial sep- Theoretically, paradoxical emboli are expected to tal defect, pulmonary arteriovenous malformation, cause brain infarcts with an imaging appearance and ventricular septal defect. Some shunts, such resembling that of brain infarcts due to other as those produced by muscular and membranous (cardiac or arterial) embolic causes. At brain ventricular septal defects, may be small and found imaging, the occlusion of a superficial arterial incidentally at clinical and imaging examinations branch or the presence of a large infarct involv- (Fig 2). ing more than one lobe is strongly suggestive of A PFO has been known to be a common find- embolic infarction (40). Scattered lesions or cor- ing since 1930, when Thompson and Evans (4) tical-subcortical territorial lesions also are indica- identified a “probe patent” foramen (0.2–0.5 cm tive of embolic infarction (Fig 1). Multiple acute in diameter) in 29% of unselected autopsy cases infarcts, especially those that are bilateral and and a “pencil patent” defect (0.6–1.0 cm in di- affect various networks of cerebral circulation, are ameter) within the atrial septum in 6%. strong indicators of a proximal embolic source or Hagen et al (46) found a PFO in 27% of 965 a systemic cause, and diffusion-weighted imaging autopsied hearts. The prevalence of PFO and the is an excellent MR imaging technique for depict- size of the defect did not differ significantly ac- ing multiple small infarcts (40,41). cording to sex but varied significantly with age: Patients with a large PFO are more likely to 34% of PFOs were found in those who had died demonstrate embolic infarcts after a cryptogenic in the first 3 decades of life; 25% of PFOs, in stroke than are patients with a small or no PFO those who had died in the 4th to 8th decades of (42). Patients with a medium or large PFO more life; and 20% of PFOs, in those who had died frequently have occipital and infratentorial (pos- in the 9th or 10th decade of life. The size of the terior circulation) strokes than do patients with a PFOs seen in the cadavers ranged from 1 to 19 small PFO (57% versus 27%) (42,43) or patients mm (mean, 4.9 mm), increasing progressively with a history of atrial fibrillation; they also tend from a mean of 3.4 mm in those who had died in to have multiple infarcts (44). Cryptogenic stroke the 1st decade of life to 5.8 mm in those who had with an “embolic” pattern is more common when died in the 10th decade of life, perhaps because PFO and atrial septal aneurysm coexist (45). smaller PFOs close spontaneously with age. Although the presence of hemorrhagic transfor- PFO has been implicated in the pathogenesis mation is a strong indicator of embolic infarction, of many diseases (11,37). The precise frequency published data do not demonstrate an association with which PDE complicates PFO is unknown; between PFO and hemorrhagic infarcts (42). PDE occurs in a minority of patients with venous thromboembolic disease who also have a Anatomic and Physiologic Con PFO. This is thought to be because the foramen siderations in Patients with a PFO ovale is normally closed by the higher left-to- Potential routes of PDE are classified in Table 2. right atrial pressure gradient. Case control stud- Both intracardiac and extracardiac shunts can lead ies that demonstrate a higher prevalence of PFO 1576 October Special Issue 2014 radiographics.rsna.org

Figure 2. Still MR images from a cine sequence depict interventricular membranous septal aneurysm and defect. LV = left ventricle. (a) Coronal view shows an aneurysm of the interventricular membranous septum (arrow). (b, c) Three-chamber views obtained during systole (b) and diastole (c) show the same aneurysm (arrow in b), along with a linear region void of signal in diastole (arrow in c). The latter finding is suggestive of a small ventricular septal defect, a potential cause of PDE. The ventricular portion of the membranous septum may become aneurysmal, usually bulging toward the right. The protruding aneurysm may limit intracardiac shunting, with resultant spontaneous closure of the ventricular septal defect.

among patients with cryptogenic strokes led to people with a PFO may be as low as 0.1% (49). the acceptance of PFO as a potential risk factor This observation suggests that other factors con- for stroke (38,39). However, whether a PFO has tribute to the increased risk of stroke in people a direct causal role in the occurrence of stroke with a PFO. Diagnostic studies performed with or whether the relationship is merely an associa- contrast agent–enhanced echocardiography while tion remains controversial. In support of a causal the patient performed the Valsalva maneuver dem- relationship, case reports have described direct onstrated pressure-dependent shunts in 50%–60% visualization of thrombus in migration through of all patients with a detectable PFO (50). Con- a PFO tunnel (47) (Fig 3). The proportion of current risk factors for venous thromboembolism, cryptogenic strokes due to PDE is believed to be such as trauma, recent surgery, use of oral contra- around 20% (48). Obviously this proportion may ceptives, and various hypercoagulable states, also vary, depending on patient age and the presence influence the clinical relevance of a PFO. of predisposing factors (Table 3). A transient spontaneous physiologic reversal of the pressure gradient between the left and Hemodynamic Parameters That In- right atria is present during early diastole and fluence PFO-related Right-to-Left Shunt during isovolumetric contraction of the right The etiology of a right-to-left shunt through a ventricle during each cardiac cycle (51,52). This PFO despite normal intracardiac pressures and so-called reversal gradient may increase when normal or near-normal pulmonary function has the patient performs physiologic maneuvers not yet been completely elucidated. The results of leading to increased right atrial pressure, such population-based studies suggest that the annual as postural changes, inspiration, vigorous cough- risk of cryptogenic stroke in otherwise healthy ing, or the Valsalva maneuver. The reversal gra- RG • Volume 34 Number 6 Saremi et al 1577

Figure 3. Thrombus in migration through a PFO tunnel (black arrows) in a 43-year-old woman. (a) CT angiography of the chest demonstrates multiple pulmonary emboli (arrowheads). (b) The presence of a thrombus (arrow) was confirmed at transthoracic four-chamber echocardiography, which provided better depiction. LA = left atrium, RA = right atrium.

from the posterior of the right atrium and is Table 3: Predisposing Factors for Right-to-Left Shunt through a PFO or for Intracardiac Shunt directed upward and backward through the flap valve of the fossa ovalis (Fig 4). The eustachian Increased pressure in the right side of the heart valve plays a crucial role in deflecting the flow Physiologic cause (eg, Valsalva maneuver, through the PFO. Horizontal reorientation of coughing) the plane of the interatrial septum may facilitate Chronic obstructive pulmonary disease flow from the IVC directly into the left atrium Extensive pulmonary embolism through the PFO. This reorientation of the sep- Primary pulmonary hypertension tum has been observed in patients with a right Intracardiac anatomic factors pneumonectomy, aortic aneurysm, or large pleu- Chiari network ral effusion (54,55). Septal aneurysm (atrial, interventricular) In patients with a PFO, pulmonary embolism Large eustachian valve is thought to be associated with a small but defi- PFO morphologic features nite risk for PDE and with findings of silent brain Short flap length infarcts at MR imaging (56). Patients with a PFO Large opening and hemodynamically important pulmonary em- Gross anatomic position bolism are more likely to experience an ischemic Supine more than sitting stroke (13% vs 2.2%), peripheral arterial embolism (15% vs 0%) (57), and arterial hypoxemia possibly due to PDE (58). dient also may be increased in the presence of a Effect of PFO Morph- pathologic condition that results in high pulmo- ology on Right-to-Left Shunt nary vascular resistance (eg, acute pulmonary The functional and morphologic characteristics embolism, hypoxemia due to obstructive sleep of a PFO are closely related. The magnitude of apnea, severe chronic obstructive pulmonary a shunt through a PFO depends not only on he- disease, right ventricular infarction, and positive modynamic parameters but also on the anatomy end-expiratory pressure during neurosurgical of the PFO (Fig 5). These include the size of the procedures performed with the patient in a sit- opening into the right atrium, length of the PFO ting position) (13,51,52). tunnel, and extent of excursion of the flap mem- Another mechanism that helps explain a brane. The magnitude of a shunt through a PFO transient right-to-left shunt is preferential di- appears to be directly related to the degree of risk rectionality of blood flow from the IVC toward for a first stroke (42). Medium to large PFOs are the interatrial septum (53). The IVC flow enters found more often in patients with cryptogenic 1578 October Special Issue 2014 radiographics.rsna.org

Figure 4. IVC flow and the eustachian valve. (a) Short- axis CT image shows preferential IVC flow toward the fossa ovalis (FO) and eustachian valve (EV). In the posterior half of the right atrial (RA) chamber, blood enters from the IVC, flowing upward through the flap valve of the fossa ovalis and backward through the eustachian valve. The eustachian valve plays a crucial role in deflecting blood flow toward a PFO.LA = left atrium, SVC = superior vena cava. (b) Axial CT image demonstrates a prominent eustachian valve (EV) guarding the anterior ostial margin of the IVC. RV = right ventricle. (c, d) Axial MR images acquired at end diastole (c) and systole (d) show dynamic movement of the eustachian valve (arrow) during the cardiac cycle.

infarcts than in those with infarcts with a known 3–9 microbubbles; a moderate shunt, by 10–30 cause (26% vs 6%) (42). Moreover, the risk of microbubbles; and a severe shunt, by more than stroke appears to be increased in the presence 30 microbubbles. Patients with a cryptogenic of structures that direct flow toward a PFO (eg, stroke had a larger PFO with a more severe right- a prominent eustachian valve) or hemodynamic to-left interatrial shunt than did patients with a changes that increase right-sided pressure (eg, a stroke of determined cause (59). large pulmonary embolism) (Table 3). PFO tunnel length is another important de- Homma et al (59) characterized PFOs in pa- terminant of the presence and severity of a right- tients with findings of cryptogenic stroke to as- to-left shunt (60,61). In a study performed by sess morphologic factors that might be conducive Natanzon and Goldman (62), the magnitude of to the development of PDE. PFO size was mea- the right-to-left shunt in patients with a stroke was sured as the maximum separation of the septum greater than that in control subjects, but no signifi- primum and septum secundum in millimeters. A cant difference was found between the two groups PFO was classified as large when the separation with regard to entry zone diameter (mean ± stan- was 2 mm or greater and small when the separa- dard deviation, 2.5 mm ± 2.0 vs 1.9 mm ± 1.6 for tion was less than 2 mm. The severity of a shunt patients vs control subjects). This finding led Na- was classified as mild, moderate, or severe on the tanzon and Goldman to speculate about whether basis of the number of microbubbles appearing in other measures could also affect the presence the left atrium during an agitated saline study: a and magnitude of a shunt through a PFO. They mild shunt was characterized by the presence of reported a shorter flap length in patients with a RG • Volume 34 Number 6 Saremi et al 1579

Figure 5. CT angiography demonstrates components of the interatrial septum in three different patients. Magnified view for each case is pre- sented in the right column. (a, b) Four-chamber views show a small fossa ovalis (FO) with fatty infiltration of the interatrial groove and the atrioventricular sandwich (AVS). IVS = interventricular septum, mAV = muscular atrioventricular septum, MV = mitral valve, P = posterior margin of the fossa ovalis, RA = right atrium, SI = septal isthmus, STV = septal leaflet of the tricuspid valve. (c–f) Short-axis images show a PFO (c, d) and a large fossa ovalis (FO) with a short interatrial groove (e, f). The septum secundum forms the interatrial groove, which cov- ers the superior (S), posterior, and inferior (I) margins of the fossa ovalis. Blue line denotes the septum primum. LA = left atrium, RA = right atrium.

cryptogenic stroke than in those with incidental PDE is more common in patients with a bi- detection of a PFO at transesophageal echocar- directional shunt through a PFO than in those diography (7.5 mm ± 3.4 vs 9.9 mm ± 6.0). We with a right-to-left shunt only (Fig 6). However, have observed similar findings at cardiac CT (Fig a bidirectional shunt is a relatively uncommon 6). In a recent study performed with multidetector finding (50).When the flap length is very short, CT in asymptomatic individuals with a PFO, 92% a bidirectional shunt is more probable. Patients of shunts occurred in the presence of a PFO tun- with an atrial septal aneurysm also have a very nel length of 6 mm or less (60). Such information short PFO tunnel length. In a recent postmor- may be important for determining the feasibility of tem study, Ho et al (61) described two types of percutaneous closure of a PFO. PFO: valve competent and valve incompetent. 1580 October Special Issue 2014 radiographics.rsna.org

Figure 6. Valve-incompetent PFO at CT angiography. (a) Short-axis CT image shows the free flap of the PFO valve (arrowheads), which is too short to cover the superior rim of the septum secundum and form a PFO tunnel (black arrow). Note the free flow of contrast material through the PFO (white arrows). (b) Short-axis CT image from another patient shows an atrial septal aneurysm (arrowheads) and a very short PFO tunnel (black arrow) causing a left- to-right shunt (white arrow). LA = left atrium, RA = right atrium.

PFOs with a short overlapping flap in the pres- of patients (64). This feature was more frequently ence of an atrial septal aneurysm were classified seen in patients with a PFO (83%) and a right-to- as incompetent, with a high likelihood of bidirec- left shunt (55%) than in subjects within the con- tional flow. Similar morphologic features of valve trol group. A Chiari network was also frequently incompetence observed at multidetector CT in associated with an atrial septal aneurysm, which patients with a PFO included a short PFO tunnel was seen in 24% of patients. Fine Chiari networks length and an atrial septal aneurysm (Fig 6) (60). may be difficult to visualize at CT or MR imaging. However, a dedicated multidetector CT study of Effect of Structures Occur- the right side of the heart may produce artifact- ring in Association with a PFO free images of the right atrium that depict a Chiari The presence of one or more aberrant ana- network (Fig 7b, 7c). tomic structures in association with a PFO can increase the probability that PDE will occur. Atrial Septal Aneurysm.—An atrial septal aneu- These abnormal structures include a Chiari net- rysm is another important anatomic feature to work, an atrial septal aneurysm, and a persistent consider when evaluating PFO. An atrial septal eustachian valve. aneurysm is defined as a bulge that protrudes more than 15 mm beyond the plane of the atrial Chiari Network.—The Chiari network consists of septum (65). Pearson et al (66) used the Han- coarse or fine fibers within the right atrium, arising ley diagnostic criteria (65) to classify findings of from the eustachian or thebesian valve and con- atrial septal aneurysm into two groups on the necting them with the crista terminalis, right atrial basis of the direction and timing of the protru- wall, or interatrial septum (Fig 7a). This structure sion. Generally, right atrial protrusion is the most is a remnant of the embryonic right valve of the common (76%) direction and is followed by tran- sinus venosus (63) and should be differentiated sient motion toward the left atrium during systole from a large eustachian valve by looking care- or with the Valsalva maneuver (65) (Fig 8). In fully for attachments to other parts of the right creased interatrial septal mobility is believed to atrium. A Chiari network has been reported in increase the probability of PDE by mechanically 2%–4% of autopsy studies (63) and is generally redirecting blood flow from the IVC through the thought to have no clinical significance. In rare PFO into the left atrium. instances, however, a Chiari network may be the The detection rate for atrial septal aneurysm site of thrombus formation (64) (Fig 7b, 7c). In a is 4.6%–10% at transesophageal echocardiogra- large patient series, a Chiari network diagnosed at phy (67,68). An atrial septal aneurysm associ- transesophageal echocardiography was seen in 2% ated with a PFO has a prevalence of 30%–60% RG • Volume 34 Number 6 Saremi et al 1581

Figure 7. Chiari network. (a) Photo graph of a cadaveric heart provides a four-chamber view of the right atrium (RA), with the Chiari network (*) in the anatomic region of the eustachian valve, anterior to the IVC and extending to the ostium of the coronary sinus (CS). (b, c) Axial (b) and two-chamber (c) CT angiograms of the right ventricle (RV) in a live patient show rounded and bandlike structures (arrows) attached to the walls of the IVC and coronary sinus (CS) ostium at the inferior cavoatrial junction. This finding was confirmed at echocardiography, which showed a possible thrombus covering the Chiari network. The patient had a history of right ventricular endocardial pacemaker. FO = fossa ovalis, LA = left atrium, TV = tricuspid valve.

(67,68) and is most likely associated with an ter birth and thereby indirectly contribute to the increased rate of embolic events (67). Among development of PDE. A prominent eustachian patients with normal patency of the carotid arter- valve is a common finding at cardiothoracic CT ies, atrial septal aneurysm is more prevalent in and MR imaging (Fig 9b). In echocardiographic those with cerebral ischemia (28%) than in those studies performed by Schuchlenz et al (69), a without cerebral ischemia (10%) (67). An atrial persistent eustachian valve was seen in 57% of septal aneurysm can easily be assessed at CT and patients, with a mean valve diameter of 1.0 cm MR imaging. In one study performed with mul- ± 0.4 (range, 0.5–2.0 cm). Seventy percent of tidetector CT, an atrial septal aneurysm was seen patients with a eustachian valve also had a PFO. in 4% of patients, and 63% of patients with an A persistent eustachian valve was more common atrial septal aneurysm were found to have a left- in patients with presumed PDE than in control to-right shunt (60) (Fig 6). subjects (68% vs 33%), but there was no significant difference in the size of the eustachian valve Persistent Eustachian Valve in Adults.—The between the two groups. eustachian valve, which guards the anteroinferior aspect of the IVC, is a remnant of the embryonic Imaging-based Diag- right valve of the sinus venosus (Fig 9a). During nosis of Intracardiac Shunts embryonic development, the eustachian valve Various imaging modalities can be used to diag- directs oxygenated blood from the IVC through nose an intracardiac shunt either directly or in- the PFO into the systemic circulation (69). By directly (Table 4). Echocardiography is the most directing the blood from the IVC toward the popular modality for this purpose; it is widely interatrial septum, a persistent eustachian valve available, noninvasive, accurate, and relatively in- may prevent spontaneous closure of the PFO af- expensive. An intracardiac shunt can be directly 1582 October Special Issue 2014 radiographics.rsna.org

Figure 8. Atrial septal aneurysm. LA = left atrium, RA = right atrium. (a, b) Short-axis CT angiograms demonstrate an atrial septal aneurysm (arrow) protruding into the right atrium in diastole (a) and toward the left atrium in systole (b). Right atrial protrusion of an atrial septal aneurysm is the most common morphologic feature, whereas motion toward the left atrium during systole or with the Valsalva maneuver is transient. (c–e) Transesophageal echocardiograms obtained during an agitated saline contrast- enhanced CT study show the atrial septal aneurysm (arrow in c and e) with right-to- left shunting visible in d and e.

assessed with echocardiography and MR imaging; shape, and location of the PFO and its spatial to evaluate extracardiac shunts, CT or MR imag- relationship to other structures (70). Mohrs et al ing is commonly performed. However, given the (70) found a good correlation between dynamic widespread use of cardiac CT for other indica- contrast-enhanced MR angiography and trans- tions, CT is increasingly relevant for assessments esophageal echocardiography in the grading of of intracardiac shunt and PFO as well. An intra- PFO shunts. In their study, a right-to-left shunt cardiac shunt may also be indirectly diagnosed on through the PFO was demonstrated by compar- the basis of findings at transcranial Doppler US. ing time-intensity curves for contrast material MR imaging enables direct flow quantification arrival in the left atrium and in a pulmonary and provides valuable information about the size, vein. However, contrast-enhanced MR angi- RG • Volume 34 Number 6 Saremi et al 1583

Figure 9. Eustachian valve or ridge. (a) Photograph of a cadaveric heart shows the right atrium, with a large eustachian valve or ridge (*) extending between the IVC and the coronary sinus (CS) ostium. C = crista terminalis, FO = fossa ovalis, TV = tricuspid valve. (b) Long- axis two-chamber black-blood MR image of the right ventricle (RV) shows a large eustachian valve (arrows) in the anteroinferior ostial margin of the IVC. RA = right atrium. ography may be inferior to contrast-enhanced taneous injections of moderately concentrated transesophageal echocardiography for detecting a contrast material into the antecubital and femoral right-to-left shunt and identifying an atrial septal veins (73). aneurysm (71). Transcranial Doppler US of the middle ce- Electrocardiographically gated multidetector rebral artery performed during the injection of CT is a fast and easy method that may obviate contrast material can be used as an alternative invasive imaging studies by depicting a com- method for indirect detection of a PFO or shunt pletely closed interatrial septum. Demonstration (Fig 10). If both echocardiography and tran- of a PFO tunnel at multidetector CT may help scranial Doppler US are performed when the predict the presence of a shunt. Only a limited presence of a PFO is suspected, the PFO detec- number of studies have involved comparison of tion rate is higher than that with either method multidetector CT with transesophageal echocar- used alone (74). Results of comparative studies diography (60,72). Current CT techniques for with transesophageal echocardiography suggest coronary angiography are capable of showing a that the sensitivity of transcranial Doppler US is left-to-right shunt. This can be important because higher than 90% but the specificity is low (ap- the demonstration of a left-to-right shunt, partic- proximately 65%–90%). The low specificity of ularly when a short flap valve length or atrial sep- transcranial Doppler US may be due to technical tal aneurysm exists, is indicative of an incompe- limitations of the study or the presence of extra- tent valve mechanism and a high likelihood that cardiac shunts. the shunt is bidirectional. No provocative test is necessary to demonstrate a left-to-right shunt at Imaging for Percutaneous PFO Closure multidetector CT. Table 5 lists the anatomic information required For CT evaluation of intracardiac shunts, data to prepare for percutaneous PFO closure. Cur- must be collected for the entire cardiac cycle. rent techniques for placement of PFO closure This requires radiation-intensive retrospec- devices rely on fluoroscopic landmarks combined tive electrocardiographic gating. Detection of a with either transesophageal or intracardiac echo- small right-to-left shunt at the level of the PFO cardiographic guidance. Transesophageal echo- indicates a need for dedicated retrospectively cardiography for device closure usually involves gated right atrial CT angiography, preferably conscious sedation but may require general an- performed while the patient performs the Valsalva esthesia in patients with airway issues or a body maneuver. The acquisition of high-quality right habitus that makes transesophageal access dif- atrial CT angiograms is challenging and requires ficult. CT has been used for localization of the homogeneous enhancement of the right atrium, fossa ovalis to aid in transseptal catheterization correct scanning timing, a small field of view lim- (75). Given that a PFO is a three-dimensional ited to the atria, and a heart rate of less than 70 structure with dynamic opening and closing, beats per minute. Homogeneous enhancement as well as a channel-like structure in some pa- of the right atrium can be obtained with simul- tients, one-dimensional measurements may not 1584 October Special Issue 2014 radiographics.rsna.org - - - - - Limitations transesophageal echocar diography for detecting a right-to-left shunt through a PFO for the echocardiography detection of a right-to-left inferior to trans PFO shunt; esophageal echocardiogra of for the identification phy atrial septal aneurysms must perform patient dows; to al maneuver Valsalva the of a shunt identification low intrapelvic thrombi intracardiac shunts may with an in patients be low extracardiac shunt Radiation dose; inferiority to dose; Radiation Inferior to transesophageal limited acoustic win Invasive; sensitivity for detection of Low Specificity for detection of - - - Advantages and pelvis that may exclude predisposing factors may and pelvis that and other information provides about PFO causes of embolic events; which is importantinadvertent puncture of the to avoid size, septum secundum with resultant extracardiac perforation; complete closure of the interatrial septum or a PFOshows of the probability an intracar findings predictive tunnel, more obviate diac shunt (presence of complete closure may cardiac and extracardiac shows imaging studies); invasive for preprocedural transatrial intervention pathways anatomic depicts acute pulmonary assessment; embolism exclu allows venous drainage; depicts anomalous rect flow; sion of predisposing factors of possible and differentiation detailed information provides on causes of embolic events; to other structures relationship and spatial shape, size, when used in conjunction with transcranial Doppler US, the is higher PFO detection rate it can be a reliable alternative method for detection material, when used in conjunction with of PFO or intracardiac shunt; is higher the PFO detection rate echocardiography, Noninvasive, rapidly performed study of the chest, abdomen, abdomen, rapidly performed study of the chest, Noninvasive, Used to assess cardiac or extracardiac shunts and quantify di Best modality for depicting a left atrial appendage thrombus; Easiest means for directly depicting thrombi in the extremities When performed contrast with an injection of microbubble - -

Optimal Uses diac and extracardiac direct depic shunts, tion of thrombus direct cardiac shunts, imagingthrombus direct thrombus imaging sources shunt Detection of intracar- Intracardiac and extra Intracardiac shunt, Extracardiac embolic of Indirect evaluation

- Imaging Workup organ damage (ie, periph- organ damage (ie, eral emboli and infarc supplemental tion); studies brain organ damage (ie, emboli and infarction) Initial assessment of target Initial assessment of target Supplemental Supplemental Supplemental

echocardiography Doppler US Doppler US Table 4: Comparison of Different Imaging Modalities for Clinical Workup of PDE 4: Comparison of Different Imaging Modalities for Clinical Workup Table Modality Multidetector CT MR imaging Transesophageal venous Peripheral Transcranial RG • Volume 34 Number 6 Saremi et al 1585

Figure 10. (a, b) Transcranial Doppler US was performed for PFO in two patients after injection of agitated saline and during the Valsalva maneuver. (a) Doppler waveform from the first patient shows a positive test result, with a mild to moderate shunt (“shower” effect). (b) Doppler waveform from the second patient shows a moderate to severe shunt (“curtain” effect). At transcranial Doppler US, passage of a single bubble leads to an instantaneous increase in signal amplitude. A diagnosis of PFO is based on the number of Doppler signals detected and the time elapsed between the end of the contrast material injection and the appearance of signal.

Table 5: Imaging Analysis before Percutaneous PFO Closure

Imaging Features Sought and Assessed Figure(s) Showing the Feature PFO morphology: size of opening, tunnel length, Figures 5, 6 patency Thickness and length of superior interatrial groove Figure 11 (septum secundum) Dimensions of fossa ovalis on short-axis and four- Figures 5, 11 chamber views Presence of atrial septal aneurysm Figures 6, 8 Distance between the PFO and both the vena cava Figures 5, 11 and the aortic root Presence of atrial septal and sinus venosus defects … Integrity of pulmonary venous anatomy seen on … axial CT or MR images Presence of an anatomic abnormality that could Figures 4, 7, 9 interfere with device placement (eg, eustachian valve, Chiari network) Presence of thrombus in the left atrial appendage …

be accurate. With CT, detailed three-dimensional of the PFO closure device and may increase the information can easily be obtained (Fig 11a). CT likelihood of complications such as dislodgment can demonstrate the relationship of important of the closure device or injury of adjacent struc- cardiac structures to the PFO and allows a deter- tures such as the aortic root. mination of the length of the interatrial groove, The criteria for PFO closure are not standard- distance to the aortic root, anomalous coronary ized (76). Some authors recommend PFO clo- arterial course, and location of the coronary sinus sure in the presence of a large PFO, a permanent ostium. Knowledge of the length of the interatrial right-to-left shunt (versus a Valsalva maneuver– groove and distance to the aortic root is crucial induced shunt), or an atrial septal aneurysm (76). when planning the placement of a closure de- In this regard, cardiac CT can provide important vice (60). The presence of a markedly thickened information about the morphologic characteris- (eg, in lipomatosis) or unusually short interatrial tics of the PFO; a permanent right-to-left shunt groove may interfere with appropriate placement is less likely to be present when the PFO tunnel 1586 October Special Issue 2014 radiographics.rsna.org

Figure 11. Percutaneous closure of PFO. (a) Short-axis preprocedural reformatted CT image of the IVC at the level of the fossa ovalis (fo) shows the most important parameters of mea- surement (bidirectional arrows): the length of the superior interatrial groove (s) and the length of the membranous flap covering the fossa ovalis.AA = ascending aorta, LA = left atrium, RA = right atrium. (b) Postprocedural transesophageal echocardiogram shows septal occlusion (Amplatzer Septal Occluder; AGA Medical, Golden Valley, Minn) (arrows). LA = left atrium. (c, d) Short-axis (c) and four-chamber (d) postprocedural CT images show occlusion of a PFO (Cardioseal; NMT Medical, Boston, Mass) (arrows) in a patient with an atrial septal aneurysm. Inset in d shows an axial view of the occlusive device. LA = left atrium.

is well formed and narrow (Fig 5). CT and MR maker leads create a predisposition to thrombus imaging are complementary to transesophageal formation (79). Patients with transposition of the echocardiography for assessing the effective- great arteries and a concomitant baffle leak may ness of a percutaneously placed occlusive device also have an increased risk of PDE (Fig 12). For (77,78) (Fig 11b, 11c). However, a small residual this reason, patients who require pacemaker im- shunt is common after placement of an occlusive plantation typically undergo a thorough prepro- device, and complete endothelialization of the cedural imaging evaluation to determine whether device takes 3–6 months, so early postprocedural a baffle leak is present (79). Potential pathways imaging may not be relevant. of interatrial communication that may be seen at cardiac CT are enlarged thebesian veins pass- Congenital Heart Malfor- ing along the superior interatrial muscle bundle mations and the Risk of PDE (Bachmann bundle) between the right and left The presence of an intracardiac shunt due to an atrial appendages (73). Pulmonary arteriovenous atrial or ventricular septal defect can increase the communications are a known complication after risk of thromboembolism especially in patients some types of cavopulmonary anastomoses be- with a permanent pacemaker, because the pace- cause of the diversion of normal hepatic venous RG • Volume 34 Number 6 Saremi et al 1587

Figure 12. Mustard procedure for transposition of the great arteries in a 24-year-old man. PDE-related embolic events caused infarc- tions in the spleen, brain, and left kidney. (a) Short-axis MR image shows a large baffle defect (arrow). RA = right atrium, SVC = superior vena cava. (b) Axial abdominal CT image shows splenic infarction (arrow). (c) Axial CT image shows the cardiac appearance after closure of the baffle defect with an occlusive device (Amplatzer Septal Occluder) (arrow).

flow from the pulmonary circulation, which may venous shunt). PDE can develop with stenosis, lead to additional complications, including PDE. thrombosis, or absence of the left brachiocephalic vein and severe stenosis or occlusion of the supe- Extracardiac Causes rior vena cava (81). Collateral venous pathways of Right-to-Left Shunt frequently extend between the left brachiocephalic vein and left atrium, through an arcade Abnormal Venous Return comprising the left superior intercostal vein or The source of cryptogenic stroke remains unex- left vertical vein and a pulmonary vein, to the left plained in approximately 50% of patients who atrium. These collateral pathways are best appre- undergo imaging to determine whether PDE is ciated at CT (Fig 14). present (80). One important role of CT or MR imaging in the evaluation of PDE is to exclude any Arteriovenous Malformations anomalous venous drainage that may have gone Paradoxical neurologic complications can occur undetected at previous echocardiography. Among in patients with a pulmonary arteriovenous mal- the possible congenital causes of a right-to-left formation, a defect most often associated with shunt, anomalous drainage of the left superior Osler-Weber-Rendu disease or hereditary hem- vena cava into the left atrium and fenestrated orrhagic telangiectasia (82) (Fig 15). These pa- coronary sinus should be investigated. In less than tients should be screened for lung arteriovenous 10% of patients with a persistent left superior vena malformations by using MR imaging or CT. It is cava, that vein drains into the left atrium either inevitable that some of these patients will have a directly or through an unroofed coronary sinus, pulmonary arteriovenous malformation, with or creating a right-to-left shunt (81) (Fig 13). without an associated PFO (which may be the Large collateral mediastinal veins may cause a sole cause of PDE), and that the malformation visible right-to-left shunt (systemic-to-pulmonary may remain a source of continued embolization 1588 October Special Issue 2014 radiographics.rsna.org

Figure 13. Extracardiac causes of PDE in a patient with repaired tetralogy of Fallot and multiple episodes of PDE-related brain infarction. (a) MR angiographic image obtained after contrast material was injected into the patient’s left arm demonstrates a right-to-left shunt due to direct communication of the left superior vena cava (arrow) with the left atrium (LA). (b) Transesophageal echocardiogram obtained after injection of agitated saline into the patient’s left arm shows bubbles (arrows) entering the left atrium (LA), a finding that confirms the presence of the right-to-left shunt. Images obtained after an agitated saline injection into the patient’s right arm showed no evidence of a PFO. Emboliza- tion of the left-sided communication was performed with a coil device. RA = right atrium. even after PFO closure (82). Echocardiography would not be adequate for identifying such abnormalities, and CT might be more reliable.

Differential Diagnosis When evaluating patients in whom the presence of PDE is suspected, it is important to determine whether other possible causes of embolic events are present. CT or MR imaging is helpful for identifying predisposing conditions such as aortic atherosclerotic disease, aortic dissection, intracardiac masses, and intracardiac thrombi. An associated acute pulmonary embolism is an important Figure 14. Coronal CT image obtained after contrast finding that is easily detected at CT. Cardiac material injection into the left arm of a 17-year-old sources of non-PDE-related embolic events are female patient with hemoptysis, hypoxemia, and mild generally located in the left side of the heart. Em- cyanosis after a Fontan procedure for a double-outlet bolic sources in the left side of the heart include right ventricle demonstrates an extracardiac cause of a left atrial appendage thrombus associated with PDE: partial occlusion of the left brachiocephalic vein atrial fibrillation, left ventricular mural clot oc- and extensive mediastinal venous collateral (Collat.) formation. A relatively large network of collateral vessels curring in the setting of myocardial infarction, connected to the right superior pulmonary vein (RSPV) mitral or aortic valve vegetation, and tumor (83). causes early filling of the left atrium(LA), an indication Embolic phenomena are indicated by present- of a right-to-left shunt. Venous thrombosis combined ing symptoms in as many as 20% of patients and with the right-to-left shunt led to the development of may be caused by detached tumor fragments or PDE. RPA = right pulmonary artery. thrombi (83). Atherosclerotic plaques of the aortic arch are an important source of extracardiac thromboembolism. The highest risk is associated (85). A left atrial appendage thrombus might with proximal arch plaques with a maximal diam- occur in patients of any age, and every imaging eter of more than 4 mm (84). study obtained for evaluation of PDE should be Nonvalvular atrial fibrillation is a common inspected carefully. At present, for the exclusion cause of embolic brain infarction in the elderly of thrombus in the left atrial appendage, MR RG • Volume 34 Number 6 Saremi et al 1589

cases. MR imaging is not an ideal modality for assessing a PFO. In patients with a large intracardiac shunt, MR imaging and echocardiography play a major role and can be used to quantify the shunt and demonstrate its exact location. In patients with an extracardiac shunt, the use of echocardiography is limited; in this setting, CT and MR angiography are better choices because they provide valuable data about extracardiac lesions such as anomalous venous return and pulmonary arteriovenous malformation. MR angiography can provide additional functional information about vascular malformations (eg, the direction, amount, and rapidity of flow through the vessel) Figure 15. Axial CT image shows an extracardiac cause of PDE in a patient with Osler-Weber-Rendu without exposing the patient to iodinated con- disease: multiple pulmonary arteriovenous malforma- trast agents and ionizing radiation. tions (arrows). In summary, the following algorithm may be followed for the diagnostic imaging evaluation of patients in whom the presence of PDE is imaging and CT cannot replace transesopha- suspected. First, initial studies are performed to geal echocardiography (86). However, both MR detect target lesions: MR imaging or CT of the imaging and CT may be attractive noninvasive brain, heart, abdomen, or extremities may be used, alternatives if transesophageal echocardiography depending on the clinical manifestations. Second, is technically unfeasible or is declined by the predisposing intracardiac abnormalities are sought patient (77,86). If the left atrial appendage has a by performing contrast-enhanced echocardiogra- normal appearance at CT (a finding with a nega- phy (preferably with a transesophageal approach) tive predictive value of >95%), transesophageal to identify PFO, intracardiac shunt, left-heart echocardiography may not be needed (86). In- thrombus or mass, or aortic or mitral valve veg- complete mixing of contrast material with blood etation. Cardiac MR imaging or CT should be in the left atrial appendage, especially in patients performed if echocardiography is not feasible or if with atrial fibrillation, may result in many false- the echocardiographic findings are not convincing. positive findings at CT. However, in a study in Third, if the presence of DVT is suspected, Dop- which transesophageal echocardiography was pler US of the lower or upper extremity is per- compared with CT, a two-phase CT protocol (an formed. Fourth, if the presence of arterial lesions early arterial phase and a phase delayed after 30 is indicated, additional studies with Doppler US, seconds) was found to have high sensitivity and CT, or MR angiography of the aorta and carotid specificity for the diagnosis of left atrial append- arteries are performed. Last, if there is evidence age thrombus (87). of a possible extracardiac predisposing cause, CT is performed to allow the detection of anomalous Conclusion venous return and arteriovenous malformation. Imaging assessment for PDE usually requires complementary use of different modalities to al- References low an accurate diagnosis excluding various dif- 1. Cohnheim J. Thrombose und embolie. In: Vorlesun- ferential possibilities. No single imaging modality gen über allgemeine pathologie. Vol. 1. Berlin, Ger- many: Hirschwald, 1877. is capable of depicting the entire spectrum of 2. Zahn FW. Thrombose de plusieurs branches de la possible findings in PDE. veine cave inférieure avec embolies consécutives MR imaging plays a primary role in the di- dans les artères. Rev Med Suisse Rom, 1881;1: agnosis of target lesions in the brain, and CT 227–237. can easily show these lesions in other parts of 3. Zahn FW. Beiträge zur geschwustlehre. Dtsch Z Chir 1885;22(1-2):1–35. the body. An embolic source in the peripheral 4. Johnson BI. Paradoxical embolism. J Clin Pathol veins can be detected with high sensitivity with 1951;4(3):316–332. US, whereas MR imaging and CT are better for 5. Thompson T, Evans W. Paradoxical embolism. Q J the diagnosis of central venous thrombosis. An Med 1930;os-23(90):135–150. intracardiac shunt through a PFO, and the as- 6. Swan HJ, Burchell HB, Wood EH. The presence of venoarterial shunts in patients with interatrial com- sociated anatomic details, is best depicted with munications. Circulation 1954;10(5):705–713. transesophageal echocardiography. Electrocardio- 7. Loscalzo J. Paradoxical embolism: clinical presenta- graphically gated CT also can show the anatomic tion, diagnostic strategies, and therapeutic options. details of a PFO and may be helpful in selected Am Heart J 1986;112(1):141–145. 1590 October Special Issue 2014 radiographics.rsna.org

8. Stöllberger C, Slany J, Schuster I, Leitner H, Win- Doppler sonography. AJR Am J Roentgenol 1996; kler WB, Karnik R. The prevalence of deep venous 167(4):971–975. thrombosis in patients with suspected paradoxical 24. Lindquist CM, Karlicki F, Lawrence P, Strzelczyk embolism. Ann Intern Med 1993;119(6):461–465. J, Pawlyshyn N, Kirkpatrick ID. Utility of balanced 9. Travis JA, Fuller SB, Ligush J Jr, Plonk GW Jr, steady-state free precession MR venography in the Geary RL, Hansen KJ. Diagnosis and treatment diagnosis of lower extremity deep venous thrombosis. of paradoxical embolus. J Vasc Surg 2001;34(5): AJR Am J Roentgenol 2010;194(5):1357–1364. 860–865. 25. Cramer SC, Maki JH, Waitches GM, et al. Paradoxi- 10. Nijsten MW, Hamer JP, ten Duis HJ, Posma JL. cal emboli from calf and pelvic veins in cryptogenic Fat embolism and patent foramen ovale. Lancet stroke. J Neuroimaging 2003;13(3):218–223. 1989;1(8649):1271. 26. Cramer SC, Rordorf G, Maki JH, et al. Increased 11. Gerriets T, Tetzlaff K, Hutzelmann A, et al. Associa- pelvic vein thrombi in cryptogenic stroke: results of tion between right-to-left shunts and brain lesions the Paradoxical Emboli from Large Veins in Isch- in sport divers. Aviat Space Environ Med 2003;74 emic Stroke (PELVIS) study. Stroke 2004;35(1): (10):1058–1060. 46–50. 12. Han SS, Kim SS, Hong HP, Lee SY, Lee SJ, Lee BK. 27. Kiernan TJ, Yan BP, Cubeddu RJ, et al. May-Thurner Massive paradoxical air embolism in brain occurring syndrome in patients with cryptogenic stroke and after central venous catheterization: a case report. J patent foramen ovale: an important clinical associa- Korean Med Sci 2010;25(10):1536–1538. tion. Stroke 2009;40(4):1502–1504. 13. Engelhardt M, Folkers W, Brenke C, et al. Neuro- 28. Aoyama T, Suehiro S, Shibata T, Sasaki Y, Fujii H. surgical operations with the patient in sitting position: Paradoxical cerebral embolism in a patient with analysis of risk factors using transcranial Doppler Paget-Schroetter syndrome. Ann Thorac Cardiovasc sonography. Br J Anaesth 2006;96(4):467–472. Surg 2005;11(6):429–431. 14. Kumar S, Wong G, Maysky M, et al. Amniotic fluid 29. Rooden CJ, Tesselaar ME, Osanto S, Rosendaal FR, embolism complicated by paradoxical embolism and Huisman MV. Deep vein thrombosis associated with disseminated intravascular coagulation. Am J Crit central venous catheters: a review. J Thromb Hae- Care 2010;19(4):379–382. most 2005;3(11):2409–2419. 15. Horowitz MB, Carrau R, Crammond D, Kanal E. 30. Ong B, Gibbs H, Catchpole I, Hetherington R, Risks of tumor embolization in the presence of an Harper J. Peripherally inserted central catheters and unrecognized patent foramen ovale: case report. upper extremity deep vein thrombosis. Australas AJNR Am J Neuroradiol 2002;23(6):982–984. Radiol 2006;50(5):451–454. 16. van Rossum AB, van Houwelingen HC, Kieft GJ, 31. Petrea RE, Koyfman F, Pikula A, et al. Acute stroke, Pattynama PM. Prevalence of deep vein throm- catheter related venous thrombosis, and paradoxical bosis in suspected and proven pulmonary embo- cerebral embolism: report of two cases. J Neuroim- lism: a meta-analysis. Br J Radiol 1998;71(852): aging 2013;23(1):111–114. 1260–1265. 32. From AH, Wang Y, Eliot RS, Edwards JE. Coronary 17. Lethen H, Flachskampf FA, Schneider R, et al. arterial embolism in persistent truncus arteriosus: Frequency of deep vein thrombosis in patients with report of a case following cardiac catheterization. N patent foramen ovale and ischemic stroke or tran- Engl J Med 1965;272:1204–1207. sient ischemic attack. Am J Cardiol 1997;80(8): 33. Hausmann D, Mügge A, Becht I, Daniel WG. Di- 1066–1069. agnosis of patent foramen ovale by transesophageal 18. Cogo A, Lensing AW, Prandoni P, Hirsh J. Distribu- echocardiography and association with cerebral and tion of thrombosis in patients with symptomatic peripheral embolic events. Am J Cardiol 1992;70 deep vein thrombosis: implications for simplifying (6):668–672. the diagnostic process with compression ultrasound. 34. Dethy S, Manto M, Kentos A, et al. PET findings Arch Intern Med 1993;153(24):2777–2780. in a brain abscess associated with a silent atrial 19. Erharhaghen J, Bartz M, Di Giovanni S, Melms A, septal defect. Clin Neurol Neurosurg 1995;97(4): Haarmeier T, Sieverding L. An unusual location of 349–353. deep venous thrombosis associated with ischemic 35. Wöhrle J, Kochs M, Hombach V, Merkle N. Preva- stroke and persistent foramen ovale. Case Rep lence of myocardial scar in patients with cryptogenic Neurol 2011;3(2):160–164. cerebral ischemic events and patent foramen ovale. 20. Wells PS, Lensing AW, Davidson BL, Prins MH, JACC Cardiovasc Imaging 2010;3(8):833–839. Hirsh J. Accuracy of ultrasound for the diagnosis of 36. Adams HP Jr, Bendixen BH, Kappelle LJ, et al. deep venous thrombosis in asymptomatic patients Classification of subtype of acute ischemic stroke: after orthopedic surgery: a meta-analysis. Ann In- definitions for use in a multicenter clinical trial— tern Med 1995;122(1):47–53. TOAST. Trial of Org 10172 in Acute Stroke Treat- 21. Kassaï B, Boissel JP, Cucherat M, Sonie S, Shah ment. Stroke 1993;24(1):35–41. NR, Leizorovicz A. A systematic review of the ac- 37. De Castro S, Cartoni D, Fiorelli M, et al. Morpho- curacy of ultrasound in the diagnosis of deep venous logical and functional characteristics of patent fora- thrombosis in asymptomatic patients. Thromb Hae- men ovale and their embolic implications. Stroke most 2004;91(4):655–666. 2000;31(10):2407–2413. 22. Cham MD, Yankelevitz DF, Shaham D, et al. Deep 38. Lechat P, Mas JL, Lascault G, et al. Prevalence of venous thrombosis: detection by using indirect CT patent foramen ovale in patients with stroke. N Engl venography. The Pulmonary Angiography-Indirect J Med 1988;318(18):1148–1152. CT Venography Cooperative Group. Radiology 39. Webster MW, Chancellor AM, Smith HJ, et al. Pat- 2000;216(3):744–751. ent foramen ovale in young stroke patients. Lancet 23. Laissy JP, Cinqualbre A, Loshkajian A, et al. As- 1988;2(8601):11–12. sessment of deep venous thrombosis in the lower 40. Wessels T, Röttger C, Jauss M, Kaps M, Traupe H, limbs and pelvis: MR venography versus duplex Stolz E. Identification of embolic stroke patterns by RG • Volume 34 Number 6 Saremi et al 1591

diffusion-weighted MRI in clinically defined lacunar 57. Konstantinides S, Geibel A, Kasper W, Olschewski stroke syndromes. Stroke 2005;36(4):757–761. M, Blümel L, Just H. Patent foramen ovale is an 41. Kang DW, Chalela JA, Ezzeddine MA, Warach S. important predictor of adverse outcome in patients Association of ischemic lesion patterns on early with major pulmonary embolism. Circulation 1998; diffusion-weighted imaging with TOAST stroke 97(19):1946–1951. subtypes. Arch Neurol 2003;60(12):1730–1734. 58. Kasper W, Geibel A, Tiede N, Just H. Patent fo- 42. Steiner MM, Di Tullio MR, Rundek T, et al. Patent ramen ovale in patients with haemodynamically foramen ovale size and embolic brain imaging find- significant pulmonary embolism. Lancet 1992;340 ings among patients with ischemic stroke. Stroke (8819):561–564. 1998;29(5):944–948. 59. Homma S, Di Tullio MR, Sacco RL, Mihalatos D, 43. Venketasubramanian N, Sacco RL, Di Tullio M, Li Mandri G, Mohr JP. Characteristics of patent Sherman D, Homma S, Mohr JP. Vascular distribu- foramen ovale associated with cryptogenic stroke: tion of paradoxical emboli by transcranial Doppler. a biplane transesophageal echocardiographic study. Neurology 1993;43(8):1533–1535. Stroke 1994;25(3):582–586. 44. Jauss M, Wessels T, Trittmacher S, Allendörfer J, 60. Saremi F, Channual S, Raney A, et al. Imaging of Kaps M. Embolic lesion pattern in stroke patients patent foramen ovale with 64-section multidetector with patent foramen ovale compared with patients CT. Radiology 2008;249(2):483–492. lacking an embolic source. Stroke 2006;37(8): 61. Ho SY, McCarthy KP, Rigby ML. Morphologi- 2159–2161. cal features pertinent to interventional closure of 45. Santamarina E, González-Alujas MT, Muñoz V, patent oval foramen. J Interv Cardiol 2003;16(1): et al. Stroke patients with cardiac atrial septal ab- 33–38. normalities: differential infarct patterns on DWI. J 62. Natanzon A, Goldman ME. Patent foramen ovale: Neuroimaging 2006;16(4):334–340. anatomy versus pathophysiology—which determines 46. Hagen PT, Scholz DG, Edwards WD. Incidence and stroke risk? J Am Soc Echocardiogr 2003;16(1): size of patent foramen ovale during the first 10 de- 71–76. cades of life: an autopsy study of 965 normal hearts. 63. Bhatnagar KP, Nettleton GS, Campbell FR, Wagner Mayo Clin Proc 1984;59(1):17–20. CE, Kuwabara N, Muresian H. Chiari anomalies 47. Dörr M, Hummel A. Images in clinical medicine: in the human right atrium. Clin Anat 2006;19(6): paradoxical embolism—thrombus in a patent fora- 510–516. men ovale. N Engl J Med 2007;357(22):2285. 64. Schneider B, Hofmann T, Justen MH, Meinertz T. 48. Thaler DE, Di Angelantonio E, Di Tullio MR, et al. Chiari’s network: normal anatomic variant or risk The Risk of Paradoxical Embolism (RoPE) Study: factor for arterial embolic events? J Am Coll Cardiol initial description of the completed database. Int J 1995;26(1):203–210. Stroke 2012 Aug 9. [Epub ahead of print] 65. Hanley PC, Tajik AJ, Hynes JK, et al. Diagnosis 49. Lock JE. Patent foramen ovale is indicted, but the and classification of atrial septal aneurysm by case hasn’t gone to trial. Circulation 2000;101(8): two-dimensional echocardiography: report of 80 838. consecutive cases. J Am Coll Cardiol 1985;6(6): 50. Lynch JJ, Schuchard GH, Gross CM, Wann LS. 1370–1382. Prevalence of right-to-left atrial shunting in a 66. Pearson AC, Nagelhout D, Castello R, Gomez CR, healthy population: detection by Valsalva maneuver Labovitz AJ. Atrial septal aneurysm and stroke: a contrast echocardiography. Am J Cardiol 1984;53 transesophageal echocardiographic study. J Am Coll (10):1478–1480. Cardiol 1991;18(5):1223–1229. 51. Zanchetta M, Rigatelli G, Ho SY. A mystery featur- 67. Mattioli AV, Aquilina M, Oldani A, Longhini C, ing right-to-left shunting despite normal intracar- Mattioli G. Atrial septal aneurysm as a cardioem- diac pressure. Chest 2005;128(2):998–1002. bolic source in adult patients with stroke and nor- 52. Langholz D, Louie EK, Konstadt SN, Rao TL, mal carotid arteries: a multicentre study. Eur Heart Scanlon PJ. Transesophageal echocardiographic J 2001;22(3):261–268. demonstration of distinct mechanisms for right to 68. Mügge A, Daniel WG, Angermann C, et al. Atrial left shunting across a patent foramen ovale in the septal aneurysm in adult patients: a multicenter absence of pulmonary hypertension. J Am Coll Car- study using transthoracic and transesophageal diol 1991;18(4):1112–1117. echocardiography. Circulation 1995;91(11): 53. Gallaher ME, Sperling DR, Gwinn JL, Meyer BW, 2785–2792. Fyler DC. Functional drainage of the inferior vena 69. Schuchlenz HW, Saurer G, Weihs W, Rehak P. Per- cava into the left atrium: three cases. Am J Cardiol sisting eustachian valve in adults: relation to patent 1963;12:561–566. foramen ovale and cerebrovascular events. J Am Soc 54. Laybourn KA, Martin ET, Cooper RA, Holman Echocardiogr 2004;17(3):231–233. WL. Platypnea and orthodeoxia: shunting associ- 70. Mohrs OK, Petersen SE, Erkapic D, et al. Diagnosis ated with an aortic aneurysm. J Thorac Cardiovasc of patent foramen ovale using contrast-enhanced Surg 1997;113(5):955–956. dynamic MRI: a pilot study. AJR Am J Roentgenol 55. Smeenk FW, Postmus PE. Interatrial right-to-left 2005;184(1):234–240. shunting developing after pulmonary resection 71. Nusser T, Höher M, Merkle N, et al. Cardiac mag- in the absence of elevated right-sided heart pres- netic resonance imaging and transesophageal echo- sures: review of the literature. Chest 1993;103(2): cardiography in patients with transcatheter closure 528–531. of patent foramen ovale. J Am Coll Cardiol 2006;48 56. Clergeau MR, Hamon M, Morello R, Saloux E, (2):322–329. Viader F, Hamon M. Silent cerebral infarcts in pa- 72. Kim YJ, Hur J, Shim CY, et al. Patent foramen tients with pulmonary embolism and a patent fora- ovale: diagnosis with multidetector CT—compari- men ovale: a prospective diffusion-weighted MRI son with transesophageal echocardiography. Radiol- study. Stroke 2009;40(12):3758–3762. ogy 2009;250(1):61–67. 1592 October Special Issue 2014 radiographics.rsna.org

73. Saremi F, Kang J, Rahmanuddin S, Shavelle D. As- aged 18 to 44 years in northern Sweden. Stroke sessment of post-atrial switch baffle integrity using a 1997;28(9):1702–1709. modified dual extremity injection cardiac computed 81. Gilkeson RC, Nyberg EM, Sachs PB, Wiant AM, tomography angiography technique. Int J Cardiol Zahka KG, Siwik ES. Systemic to pulmonary ve- 2013;162(2):e25–e27. nous shunt: imaging findings and clinical presenta- 74. Droste DW, Schmidt-Rimpler C, Wichter T, et al. tions. J Thorac Imaging 2008;23(3):170–177. Right-to-left-shunts detected by transesophageal 82. Kjeldsen AD, Oxhøj H, Andersen PE, Green A, Vase echocardiography and transcranial Doppler sonog- P. Prevalence of pulmonary arteriovenous malfor- raphy. Cerebrovasc Dis 2004;17(2-3):191–196. mations (PAVMs) and occurrence of neurological 75. Graham LN, Melton IC, MacDonald S, Crozier IG. symptoms in patients with hereditary haemorrhagic Value of CT localization of the fossa ovalis prior to telangiectasia (HHT). J Intern Med 2000;248(3): transseptal left heart catheterization for left atrial 255–262. ablation. Europace 2007;9(6):417–423. 83. Burke A, Jeudy J Jr, Virmani RA. Cardiac tumours: 76. Rigatelli G, Dell’Avvocata F, Cardaioli P, et al. an update. Heart 2008;94(1): 117–123. Permanent right-to-left shunt is the key factor in 84. Tenenbaum A, Garniek A, Shemesh J, et al. Dual- managing patent foramen ovale. J Am Coll Cardiol helical CT for detecting aortic atheromas as a 2011;58(21):2257–2261. source of stroke: comparison with transesophageal 77. Mohrs OK, Petersen SE, Erkapic D, et al. Dynamic echocardiography. Radiology 1998;208(1):153–158. contrast-enhanced MRI before and after transcath- 85. Feinberg WM, Blackshear JL, Laupacis A, Kronmal eter occlusion of patent foramen ovale. AJR Am J R, Hart RG. Prevalence, age distribution, and gen- Roentgenol 2007;188(3):844–849. der of patients with atrial fibrillation: analysis and 78. Lee T, Tsai IC, Fu YC, et al. MDCT evaluation af- implications. Arch Intern Med 1995;155(5):469–473. ter closure of atrial septal defect with an Amplatzer 86. Feuchtner GM, Dichtl W, Bonatti JO, et al. Diag- septal occluder. AJR Am J Roentgenol 2007;188(5): nostic accuracy of cardiac 64-slice computed to- W431–W439. mography in detecting atrial thrombi: comparative 79. Kuppahally SS, Litwin SE, Green LS, Ishihara SM, study with transesophageal echocardiography and Freedman RA, Michaels AD. Utility of intracardiac cardiac surgery. Invest Radiol 2008;43(11):794–801. echocardiography for atrial baffle leak closure in 87. Hur J, Kim YJ, Lee HJ, et al. Left atrial appendage repaired transposition of the great arteries. Echocar- thrombi in stroke patients: detection with two- diography 2010;27(8):E90–E93. phase cardiac CT angiography versus transesoph- 80. Kristensen B, Malm J, Carlberg B, et al. Epidemiol- ageal echocardiography. Radiology 2009;251(3): ogy and etiology of ischemic stroke in young adults 683–690.

This journal-based SA-CME activity has been approved for AMA PRA Category 1 CreditTM. See www.rsna.org/education/search/RG. Teaching Points October Special Issue 2014

Paradoxical Embolism: Role of Imaging in Diagnosis and Treatment Planning Farhood Saremi, MD • Neelmini Emmanuel, MD • Philip F. Wu, BS • Lauren Ihde, MD • David Shavelle, MD • John L. Go, MD • Damián Sánchez-Quintana, MD, PhD

RadioGraphics 2014; 34:1571–1592 • Published online 10.1148/rg.346135008 • Content Codes:

Page 1572 The diagnosis of PDE is considered definitive when it is based on a finding at autopsy or at imaging of a thrombus that crosses an intracardiac defect in the setting of an arterial embolus. A diagnosis of PDE in the absence of these findings is considered presumptive. The triad of systemic embolism, venous thrombosis, and intracardiac communication defines the clinical diagnosis of PDE and allows treatment with a high level of confidence. The diagnosis of PDE is termed “possible” if an arterial embolus and PFO are detected; many physicians treat patients on the basis of a diagnosis of “possible PDE.”

Page 1573 Overall, the results of the preceding studies show that (a) PDE from the lower extremity and possibly the pelvis is one mechanism that accounts for ischemia related to systemic embolization in a subset of patients and (b) pelvic CT or MR imaging may be useful for determining whether pelvic DVT is present in patients in whom findings are negative for DVT of the lower extremities.

Page 1574 The cause of cryptogenic stroke remains undetermined in most cases because the event is transitory or reversible, investigators cannot look for all possible causes, and some causes remain unknown. The detection of a PFO in a patient with a confirmed stroke does not necessarily mean that the cause of the stroke has been identified. Establishing a causal relationship between the presence of a PFO and the occurrence of a stroke remains the crucial point in the diagnosis of PDE. The four criteria described earlier for the diagnosis of PDE may not always be met. The presence of other contributing factors, such as the morphologic characteristics of the PFO and associated structures, may increase the probability that PDE is present.

Page 1579–1580 When the flap length is very short, a bidirectional shunt is more probable. Patients with an atrial septal aneurysm also have a very short PFO tunnel length. In a recent postmortem study, Ho et al described two types of PFO: valve competent and valve incompetent. PFOs with a short overlapping flap in the presence of an atrial septal aneurysm were classified as incompetent, with a high likelihood of bidirectional flow.

Page 1585 The criteria for PFO closure are not standardized. Some authors recommend PFO closure in the presence of a large PFO, a permanent right-to-left shunt (versus a Valsalva maneuver–induced shunt), or an atrial septal aneurysm.