When Cancer Clots: An Extensive Radiologic Analysis of Cancer-Associated Thromboembolism.

1. Introduction

1.1. Epidemiology and Burden of CAT

Cancer-associated thrombosis (CAT) is a major source of morbidity and mortality in patients with malignancy and remains a critical challenge across oncologic disciplines, including radiology and interventional radiology. Cancer patients have a markedly increased risk of venous thromboembolism (VTE), with reported rates approximately four- to seven-fold higher than those observed in the general population [ 1 ]. Even among hospitalized patients, the incidence of pulmonary embolism (PE) is nearly twice as high in those with cancer compared to those without [ 2 ]. Importantly, cancer patients face a threefold increased risk of recurrent VTE overall, with the highest risk occurring within the first 6 months following the initial cancer diagnosis [ 3 ].

The burden of CAT has continued to rise over recent decades. It is driven by prolonged patient survival, increasing use of thrombogenic systemic therapies, expanded reliance on central venous catheters, and heightened clinical awareness leading to increased detection [ 4 ]. Approximately 15% of cancer patients will experience VTE during the course of their disease. Also, up to 20% of unprovoked VTE events may represent the first manifestation of an underlying malignancy [ 2 ]. This bidirectional relationship underscores the pivotal role of imaging in both the detection of thromboembolic disease and the identification of occult malignancy.

Beyond venous events, arterial thromboembolism (ATE) is increasingly recognized as a clinically significant complication of cancer. In a large population-based analysis, the six-month ATE incidence among cancer patients was more than twice that of matched controls [ 5 ]. ATE in this population encompasses a clinically heterogeneous spectrum. It is ranging from cerebrovascular accident and myocardial infarction to peripheral arterial occlusion and visceral infarction, arising through mechanisms that differ fundamentally from conventional atherosclerotic disease.

1.2. The Radiologist’s Role in CAT

Although many of these events present with overt clinical symptoms, radiologists contribute meaningfully to their evaluation by identifying predisposing vascular and cardiac pathology, characterizing the extent, and recognizing cancer-related mechanisms that may not be apparent from clinical presentation alone [ 5 ].

Thromboembolism in cancer is not confined to conventional presentations such as deep vein thrombosis (DVT) and PE. Unusual-site thromboses, including upper extremity, cerebral venous, and splanchnic vein thrombosis, are increasingly encountered, often as incidental findings on routine staging or surveillance imaging [ 6 ]. Upper extremity thrombosis is significantly more common in patients with active malignancy, largely attributable to indwelling central venous catheters [ 2 ], while visceral vein thrombosis is particularly associated with gastrointestinal malignancies, including pancreatic cancer. Although many of these events are incidentally detected, both symptomatic and incidental thromboembolism have been linked to increased mortality, highlighting their prognostic relevance [ 7 ].

From a prognostic standpoint, VTE is the second leading cause of death in ambulatory cancer patients receiving chemotherapy, after disease progression [ 8 ]. It is independently predicts poor survival irrespective of cancer stage, age, or race [ 9 , 10 ]. PE carries particular lethality in this population; fatal PE occurs at rates up to four times higher than fatal bleeding from anticoagulation, with approximately 88% of those deaths occurring within the first month of anticoagulation [ 11 ].

Given the prognostic weight of these events, timely and accurate diagnosis is essential. However, diagnosis remains complicated by the limitations of conventional clinical assessment, as symptoms of VTE are frequently nonspecific. While structured diagnostic algorithms incorporating clinical probability scores and D-dimer testing are well validated in the general population, their performance is significantly diminished in oncology patients due to baseline biomarker elevation and overlapping symptomatology [ 12 ]. Consequently, imaging, primarily compression ultrasonography (CUS) and computed tomography pulmonary angiography (CTPA), plays a central role in the diagnostic pathway for suspected VTE in cancer patients, and many guidelines recommend proceeding directly to imaging without preliminary D-dimer testing in this population [ 12 ]. Complementing these modalities, 18F-FDG PET/CT (18 Fluorodeoxyglucose positron emission tomography/computed tomography) has demonstrated important value in identifying occult malignancy in patients presenting with unprovoked VTE, a role discussed in detail later in this review [ 13 , 14 ].

Radiologists are uniquely positioned at the intersection of thrombosis detection, cancer staging, and longitudinal disease assessment, often serving as the first to identify thromboembolic events, characterize their extent, and flag findings that alter clinical management. This review provides a comprehensive examination of the imaging spectrum of cancer-associated venous and arterial thromboembolism, with emphasis on imaging manifestations, staging implications, and the evolving role of radiology in clinical decision-making.

1.3. The Scope of the Review

This narrative review was conducted using a systematic approach to identify relevant peer-reviewed literature. PubMed/MEDLINE, Embase, and Scopus databases were searched from January 2000 through December 2025 using combinations of the following keywords: “cancer-associated thrombosis,” “venous thromboembolism,” “arterial thrombosis,” “FDG-PET,” “CT angiography,” “MR venography,” “Doppler ultrasound,” “tumor thrombus,” “paraneoplastic vasculitis,” and “immune checkpoint inhibitor vasculitis.” Additional articles were identified by manual review of reference lists from included studies and relevant review articles. Only English-language articles published in peer-reviewed journals were included. Priority was given to meta-analyses, systematic reviews, prospective studies, and established clinical guidelines. Case reports and preclinical studies were included only when they provided unique mechanistic insights not available from higher-level evidence. The final reference list reflects the authors’ judgment of the most relevant and highest-quality evidence supporting each section of this review.

2. Etiology

CAT arises from a complex interplay between tumor-related, treatment-related, and patient-specific factors ( Table 1 ).

2.1. Tumor-Intrinsic Mechanisms

Malignancy induces a systemic prothrombotic state through tumor cell expression of tissue factor and release of tissue factor-bearing microparticles. Cancer also causes activation of platelets and leukocytes, formation of neutrophil extracellular traps, endothelial dysfunction, and direct vascular invasion. Oncogenic alterations, including JAK2V617F [Janus kinase 2 mutation (valine to phenylalanine at codon 617)] mutations in myeloproliferative neoplasms and MET or KRAS mutations in solid tumors, have also been associated with increased thrombotic risk, further highlighting the biological link between oncogenic signaling and coagulation activation. Certain tumors has increased rate of CAT such as pancreatic, hepatobiliary and gastric tumors [ 12 , 15 ].

2.2. Treatment-Related Risk

Cancer-directed therapies substantially amplify thrombotic risk. Surgical interventions, particularly abdominal and pelvic oncologic procedures, are associated with a 2-fold increased risk of DVT and a 3-fold increased risk of PE compared with patients without cancer undergoing similar procedures. Chemotherapy increases VTE risk by approximately 2-6-fold, and its expanding use likely contributes to the rising incidence of CAT [ 3 ].

Certain agents carry particularly high thrombotic potential. Cisplatin-based regimens are associated with higher thromboembolic rates compared with oxaliplatin-based therapies. Immunomodulatory drugs used in multiple myeloma, including thalidomide, lenalidomide, and pomalidomide, are linked to increased venous and arterial thromboembolic events. Antiangiogenic therapies such as bevacizumab and multitargeted tyrosine kinase inhibitors (e.g., ponatinib) increase the risk of ATE. More recently, immune checkpoint inhibitors have been associated with elevated risks of both VTE and ATE, potentially through immune-mediated endothelial injury and inflammatory cytokine release. Across these diverse drug classes, the common thread is disruption of endothelial integrity or coagulation balance, reinforcing that thrombotic risk is an inherent consideration in the selection and monitoring of cancer-directed therapy.

2.3. Supportive Care and Hospitalization-Related Risk

Beyond cancer-directed treatments, a range of additional exposures inherent to comprehensive oncologic care independently contribute to thrombotic burden. These include supportive pharmacologic agents such as erythropoiesis-stimulating agents, procedural factors such as central venous catheter placement, and situational factors such as hospitalization and immobility [ 2 ]. Blood transfusions are of particular note; frequently administered when erythropoiesis-stimulating agents are contraindicated or insufficient, they have been independently associated with both VTE and ATE as well as increased in-hospital mortality in large multicenter analyses of cancer hospitalizations [ 16 ]. Taken together, these exposures create a layered thrombotic risk that compounds across the trajectory of cancer treatment.

2.4. Patient-Level Risk Factors

Patient-level factors further modulate thrombotic risk in ways that are not always captured by tumor type or treatment regimen alone, and their assessment is central to individualized risk stratification in oncology patients. Increasing age, particularly beyond 70 years, is associated with up to a 2-fold higher VTE risk during chemotherapy, while poor functional status, and medical comorbidities including anemia, infection, obesity, and pulmonary or renal disease each contribute independently to overall thrombotic burden [ 8 ]. These factors are clinically important. They are often present at baseline and may compound the thrombotic risk already conferred by the malignancy and its treatment, creating a cumulative vulnerability that persists throughout the care continuum.

Inherited thrombophilias such as antithrombin deficiency, protein C/S deficiency, and factor V Leiden confer additional risk, though they are encountered infrequently in the oncology population and rarely drive clinical decision-making in this context. Their presence may, however, assume greater relevance in patients with unprovoked thrombosis or a strong personal or family history of thromboembolic disease, where thrombophilia testing may inform the duration and intensity of anticoagulation. A prior history of VTE is among the strongest predictors of future thromboembolism in this population [ 8 ]. This observation has direct practical implications for the radiologist: prior VTE should be actively sought when reviewing oncologic imaging, as its presence elevates the pre-test probability of new thromboembolic findings and should inform the threshold for further evaluation when borderline or equivocal imaging features are encountered.

3. Biomarkers

Having established the multifactorial nature of thrombotic risk in oncology patients, the question of how that risk is identified clinically brings biomarker testing into focus. In patients with low or intermediate clinical probability, D-dimer testing is recommended as an initial diagnostic step. D-dimer is a fibrin degradation product released into the circulation following clot formation and lysis, and a negative result in this context effectively excludes PE or DVT, whereas a positive result warrants imaging evaluation [ 17 , 18 ].

In oncology patients, however, D-dimer levels are frequently elevated due to malignancy-associated inflammation and hypercoagulability, reducing test specificity and substantially limiting clinical utility [ 19 , 20 ]. As a result, the conventional stepwise diagnostic approach is less reliable in this population, and clinicians often proceed directly to imaging without preliminary biomarker testing.

Emerging biomarkers such as soluble P-selectin and tissue factor-bearing microparticles have demonstrated potential for predicting first VTE events. Soluble P-selectin has been incorporated into certain risk assessment models, whereas data on tissue factor-bearing microparticles remain inconsistent and appear strongest in pancreatic cancer [ 12 ]. Collectively, these limitations underscore why imaging remains the cornerstone of thromboembolism evaluation in cancer patients, a principle that informs the diagnostic framework addressed throughout the remainder of this review.

4. Venous Thromboembolism—Deep Vein Thrombosis

When clinical assessment and biomarker evaluation raise concern for DVT, imaging becomes the definitive next step. Compression ultrasonography (CUS) is the established first-line modality, valued for its high diagnostic accuracy, noninvasiveness, and bedside availability.

4.1. Diagnostic Principles and Sonographic Findings

The fundamental diagnostic principle of CUS is the assessment of venous compressibility. Under normal conditions, deep veins are fully compressible with gentle transducer pressure. The hallmark sonographic finding of acute DVT is loss of complete vein compressibility, which remains the primary and most reliable diagnostic criterion. An acute thrombus typically appears as hypoechoic or mildly echogenic intraluminal material within a dilated vein. During probe compression, the affected vein fails to collapse, often demonstrating a measurable residual luminal diameter ( Figure 1 ). Distinguishing acute from chronic thrombosis carries important clinical implications, particularly when evaluating suspected recurrent DVT. Chronic or post-thrombotic changes tend to demonstrate only partial compressibility, increased echogenicity from fibrosis, venous wall thickening, collateral vessel formation, and, in some cases, the characteristic dual-rail appearance of the vein wall [ 21 ].

4.2. Color Doppler Assessment and Adjunct Modalities

Color Doppler imaging provides complementary information. Acute occlusive thrombosis is associated with absent or markedly reduced venous flow signals. Additional indirect findings include loss of normal respiratory phasicity and an abnormal response to distal augmentation maneuvers [ 21 ]. These Doppler features are particularly useful when evaluating proximal venous segments, such as the iliac veins, where direct compression may be technically limited. In these anatomically challenging regions, CT venography (CTV) or MR venography (MRV) serves as an important adjunct, offering reliable depiction of iliac and pelvic vein thrombosis that may be inaccessible or inadequately visualized on ultrasound [ 20 ].

4.3. Examination Protocols and Diagnostic Performance

Standard examination protocols for lower extremity DVT typically include evaluation of the common femoral vein, superficial and deep femoral veins, popliteal vein, and, when performing whole-leg ultrasonography, the tibial and peroneal veins. While limited two- or three-point compression protocols may be used in selected clinical contexts, whole-leg assessment reduces the likelihood of missing distal thrombosis. Both approaches are considered equivalent in terms of safety, with large management studies demonstrating false-negative rates below 1% [ 19 ]. Accuracy is somewhat lower for isolated distal (calf) thrombosis and in technically challenging scenarios such as obesity, pregnancy, or pelvic vein involvement. Proximal DVT warrants prompt and accurate detection, as incomplete or delayed treatment carries the risk of post-thrombotic syndrome. This chronic condition is characterized by persistent limb swelling, pain, and skin changes resulting from valvular incompetence and venous hypertension and is of particular consequence in oncology patients who may face prolonged immobility and repeated venous instrumentation [ 21 ].

4.4. Upper Extremity DVT

Upper extremity DVT warrants specific attention in oncology patients, given that it occurs up to 18 times more frequently in patients with active malignancy, largely attributable to indwelling central venous catheters [ 2 ]. Ultrasound is the primary imaging modality for evaluation of the internal jugular, subclavian, and axillary veins, with compressibility assessment and color Doppler interrogation forming the diagnostic core. However, the central thoracic veins, including the subclavian vein medial to the clavicle, the brachiocephalic veins, and the superior vena cava, are not accessible to direct compression and often cannot be adequately evaluated sonographically. In these territories, CTV or MRV is required for definitive assessment and should be pursued when central venous obstruction is clinically suspected despite a non-diagnostic or technically limited ultrasound examination ( Figure 2 ) [ 19 ].

5. Venous Thromboembolism—Pulmonary Embolism

Computed tomography pulmonary angiography (CTPA) is the first-line imaging modality for suspected PE in oncology patients, offering rapid, high-resolution evaluation of the pulmonary vasculature. The primary diagnostic finding is a filling defect within the pulmonary arterial tree, which may be central, segmental, or subsegmental in distribution. Radiologists also assess for indirect signs of hemodynamic significance, including right ventricular enlargement, interventricular septal flattening, and reflux of contrast into the hepatic veins, as these findings carry important prognostic implications and should be explicitly reported ( Table 2 ) [ 17 ].

Beyond direct thrombus detection, CTPA frequently identifies incidental findings relevant to the oncologic patient, including mediastinal lymphadenopathy, pleural effusion, parenchymal consolidation, and previously unrecognized or progressive malignancy. This dual diagnostic yield underscores the value of systematic image review extending beyond the pulmonary arteries in this population [ 19 ].

Patient Selection and Alternative Modalities

Despite its central role, CTPA remains frequently overutilized in oncology patients due to the nonspecific nature of PE symptoms in this population. In a recent study, acute PE was confirmed in only 21.4% of patients undergoing CTPA, with nearly 80% of studies negative. Clinical features such as unilateral limb edema, adenocarcinoma histology, inpatient status, and signs of right ventricular strain may assist in refining pre-test probability and improving patient selection for imaging.

When CTPA is contraindicated, such as in patients with severe contrast allergy or significant renal insufficiency, ventilation–perfusion (V/Q) scintigraphy represents the recommended alternative. V/Q imaging retains meaningful diagnostic utility in oncology patients, though interpretation may be complicated by underlying structural abnormalities or post-treatment pulmonary changes that can produce ventilation–perfusion mismatch unrelated to thromboembolism [ 23 ].

6. Incidental Thromboembolism

6.1. Prevalence and Risk Factors

While efforts to improve patient selection for CTPA address the problem of suspected but unconfirmed PE, a clinically distinct and increasingly prevalent scenario arises when thromboembolism is identified entirely in the absence of clinical suspicion ( Figure 3 ). This phenomenon is encountered with growing frequency as multidetector CT has become the standard of care for cancer staging and restaging. These incidentally detected events, variably termed asymptomatic, unsuspected, or unexpected VTE in the literature, are encountered across tumor types and represent a diagnostically and prognostically significant subset of cancer-associated thromboembolism. Prevalence varies by tumor type and patient population. In a meta-analysis of more than 10,000 cancer patients, the weighted mean prevalence of incidental PE was 2.6%. Incidental abdominal DVT and lower extremity DVT ( Figure 4 ) occur at comparable rates of approximately 1.1% and 1.3%, respectively [ 24 ]. These figures are substantially higher in specific high-risk subgroups. Among patients with gastrointestinal cancers undergoing routine staging imaging, unsuspected venous or visceral thrombosis was identified in 7.3% overall, with incidental PE accounting for 2.7% of cases [ 25 ], and incidental abdominal DVT has been reported in up to 2.5% of selected oncology populations in dedicated CT studies ( Figure 5 ) [ 26 ].

6.2. Prognostic Significance

The prognostic weight of incidental VTE is substantial and should inform how radiologists communicate these findings. Current guidelines recommend anticoagulant treatment equivalent to that for symptomatic events, based on evidence that rates of recurrent VTE, major bleeding, and mortality are clinically significant in both groups [ 12 , 27 ]. In pancreatic cancer specifically, incidental DVT, PE, and visceral thrombosis each independently predicted increased mortality [ 7 ]. The anatomic distribution of incidental PE further supports this equivalence; nearly half involve the central pulmonary arteries [ 28 ], a distribution indistinguishable from that of symptomatic PE. This reinforces that the absence of symptoms does not imply a lesser thromboembolic burden or the potential for hemodynamic compromise.

6.3. Communication and Reporting Standards

Incidental PE and DVT on routine staging CT require prompt clinical action. Communication urgency should be tiered: immediate (same-day, direct physician contact) for central or extensive PE, bilateral PE with right heart strain, or extensive caval/iliocaval DVT; routine (24–48 h) for subsegmental PE or incidental extremity DVT without proximal extension; and not only for isolated muscular vein thrombosis or chronic non-occlusive thrombus. The radiology report should include six key elements: (1) precise location; (2) acuity (acute vs. chronic); (3) burden; (4) hemodynamic relevance; (5) comparison to prior studies; and (6) relationship to indwelling devices. Standardized reporting templates can improve communication consistency for incidental VTE.

Collectively, these findings make clear that incidentally detected PE and DVT are not passive radiologic observations. They carry meaningful clinical and prognostic implications, and their accurate detection and reporting are central to the radiologist’s role in oncologic care [ 21 ].

7. Cancer-Associated Cerebral Venous Thrombosis

Beyond the conventional venous territories addressed above, cancer-associated thromboembolism may also involve the cerebral venous system ( Figure 6 ). This represents an uncommon but serious manifestation occurring in approximately 6% of hospitalized CVT patients, with hematopoietic malignancies representing the most frequently associated tumor type [ 29 ]. Given the potential for rapid clinical deterioration, timely and accurate imaging diagnosis is essential.

MRI with MR venography (MRV) is the preferred modality for diagnosis and characterization of CVT. On conventional sequences, the thrombosed sinus or cortical vein demonstrates loss of the normal flow void, with signal characteristics that evolve predictably over days with thrombus age. Susceptibility-weighted imaging (SWI) is particularly sensitive for detecting cortical vein thrombosis, where blooming artifact from deoxyhemoglobin provides early diagnostic information. The superior sagittal and transverse sinuses are the most affected and demonstrate absence or irregularity of flow signal on MRV imaging [ 30 ].

CT and CT venography (CTV) serve as important alternatives, particularly in acute settings or when MRI is contraindicated. Non-contrast CT may demonstrate the hyperdense cord or dense triangle sign reflecting acute thrombus within a dural sinus, though sensitivity is limited. CTV offers a rapid and reliable depiction of sinus occlusion and associated parenchymal complications, including hemorrhagic venous infarction and cerebral edema, findings that carry direct implications for management and prognosis [ 30 ].

In cancer patients, CA-CVT tends to present with greater severity than in the general population, with higher rates of cerebral edema, hemorrhagic transformation, and herniation. Radiologists should maintain a low threshold for evaluating the cerebral venous system on routine brain imaging in oncology patients, as these findings may otherwise be overlooked [ 29 ].

8. Portal Vein Thrombosis

Portal vein thrombosis (PVT) represents the most frequent type of splanchnic venous thrombosis in oncology patients and is particularly prevalent in hepatobiliary and pancreatic malignancies. Its presence can complicate staging, obfuscate imaging interpretation, and adversely affect treatment planning. The presence of PVT has direct implications for surgical candidacy, including hepatic resection and transplant eligibility, as well as suitability for trans arterial chemoembolization and other locoregional procedures [ 31 ]. PVT in cancer patients is associated with increased risks of portal hypertension, hepatic dysfunction, and reduced overall survival, underscoring the importance of early detection [ 7 ].

The central diagnostic challenge in cancer-associated PVT is distinguishing bland thrombus from tumor thrombus, a distinction with direct consequences for staging and management. Please see relevant subsequent section.

9. Splanchnic and Visceral Vein Thrombosis

Beyond the portal vein, thrombosis of the broader splanchnic and visceral venous system represents an important and diagnostically distinct subset of cancer-associated thrombosis (CAT). Mesenteric vein thrombosis is most commonly encountered in the setting of pancreatic and other gastrointestinal malignancies, where tumor-related vascular compression, local inflammation, and systemic hypercoagulability converge to promote thrombosis. On contrast-enhanced CT, mesenteric vein thrombosis appears as a hypodense intraluminal filling defect within the superior mesenteric vein and is often accompanied by bowel wall thickening or mesenteric edema reflecting venous congestion. Splenic vein thrombosis is similarly associated with pancreatic malignancy due to its anatomic proximity and is a recognized complication of both pancreatic ductal adenocarcinoma and malignancy-associated pancreatitis [ 31 ].

Hepatic vein thrombosis, or Budd–Chiari syndrome, carries a particularly strong association with myeloproliferative neoplasms [ 32 ]. On imaging, acute Budd–Chiari syndrome demonstrates hepatic vein or IVC occlusion, caudate lobe hypertrophy, heterogeneous hepatic parenchymal enhancement reflecting altered perfusion, and variable degrees of ascites. Hepatic vein obstruction may also arise from direct tumor invasion, most notably in hepatocellular carcinoma, and should be distinguished from primary Budd–Chiari syndrome as it carries distinct staging and management implications. [ 22 ].

10. Bland vs. Tumor Thrombus: Imaging Differentiation

As referenced in the preceding sections on portal and visceral vein thrombosis, the distinction between bland and tumor thrombus represents one of the most consequential diagnostic determinations in oncologic imaging, with implications spanning TNM staging, surgical planning, and systemic treatment selection. While the clinical relevance of this distinction arises across multiple vascular territories, it is most extensively characterized in renal cell carcinoma involving the renal vein and inferior vena cava and in hepatocellular carcinoma involving the portal and hepatic venous systems. Bland thrombus does not contain viable tumor cells and is managed with anticoagulation, whereas tumor thrombus reflects direct intravascular propagation of malignancy and requires oncologic rather than anticoagulant-directed treatment. On contrast-enhanced cross-sectional imaging, the key discriminating features of tumor thrombus include arterial-phase enhancement, expansion of the involved vessel beyond its native caliber, and direct anatomic continuity with the primary tumor. Bland thrombus characteristically lacks enhancement and does not distend the vessel ( Figure 7 ). When cross-sectional imaging is equivocal, diffusion-weighted MRI demonstrating restricted diffusion within the thrombus and contrast-enhanced ultrasound demonstrating internal vascularity both favor tumor thrombus [ 22 ]. FDG-PET/CT can provide complementary information, as discussed in a later section, with uptake in the thrombus suggesting tumoral rather than bland composition. Accurate thrombus characterization has direct and consequential implications for staging, surgical planning, and systemic treatment selection that are determined in large part by the radiologist’s interpretation. In renal cell carcinoma, the level of IVC tumor thrombus extension directly determines T stage and dictates surgical approach, while in hepatocellular carcinoma, portal vein tumor thrombus affects transplant eligibility and locoregional treatment candidacy. Misclassification in either direction, treating tumor thrombus as bland or failing to identify venous invasion, can result in inadequate oncologic therapy or unnecessary anticoagulation in a population already at elevated bleeding risk. Systematic multiparametric assessment is therefore not merely a diagnostic refinement but a clinical imperative ( Figure 8 ) [ 22 ].

11. Cancer-Associated Arterial Thromboembolism

While the preceding sections have addressed the full spectrum of cancer-associated venous thromboembolism, arterial thromboembolic events represent a distinct and increasingly recognized complication of malignancy that warrants separate consideration. Arterial thromboembolism (ATE) in cancer arises through mechanisms that differ fundamentally from conventional atherosclerotic disease, including tumor-related hypercoagulability, therapy-induced endothelial injury, and nonbacterial thrombotic endocarditis, and produces imaging patterns that may not conform to expected vascular distributions or risk profiles. The risk of ATE is highest in the first six months following cancer diagnosis and is associated with substantial morbidity and mortality [ 5 ]. Radiologists encountering these findings on routine oncologic imaging or in the acute setting should recognize their potential cancer-related etiology, as this recognition carries direct implications for both thrombotic management and oncologic workup. The following subsections address the cerebrovascular, coronary, and peripheral arterial territories in turn, each of which presents distinct imaging manifestations and diagnostic considerations in the oncologic context.

11.1. Cerebrovascular Imaging

Among the arterial territories affected by cancer-associated thromboembolism, the cerebrovascular system is particularly important to recognize, as the imaging signature of cancer-related stroke is often distinctive enough to suggest malignancy-associated hypercoagulability before the underlying diagnosis is confirmed. Ischemic stroke in cancer patients frequently presents on MRI with multifocal areas of restricted diffusion on diffusion-weighted imaging (DWI) scattered across multiple vascular territories, a pattern that strongly suggests an embolic mechanism rather than large-vessel atherosclerosis or small-vessel disease. The three-territory sign is defined as simultaneous involvement of the bilateral anterior and posterior circulations. Up to approximately one in five patients presenting with this DWI pattern have an underlying malignancy. Therefore, this finding should prompt consideration of cancer-related hypercoagulability, particularly when no conventional embolic source is identified. In one series, cancer-associated hypercoagulability accounted for approximately 75% of three-territory infarcts without an identifiable source. On contrast enhanced imaging, these infarcts tend to be non-enhancing, peripherally located, and range from 0.5 to 2 cm in size, without the cortical ribbon or deep gray nuclei involvement seen in hypoxic–ischemic injury [ 33 ].

One of the principal mechanisms underlying this multifocal embolic pattern is nonbacterial thrombotic endocarditis (NBTE). This entity represents an important and underdiagnosed source of cardio-embolism in cancer patients, particularly in pancreatic and other mucin-secreting adenocarcinomas [ 34 ]. When NBTE is the embolic source, DWI characteristically shows numerous small to large lesions distributed across multiple territories. This pattern should prompt echocardiographic evaluation for valvular vegetations, given its direct implications for diagnosis and anticoagulation management. When echocardiography is inconclusive, cardiac MRI may provide additional characterization of valvular lesions, while CT angiography can delineate downstream embolic consequences, including territorial infarction and end-organ involvement.

Beyond thromboembolic mechanisms, certain cancer-directed therapies produce cerebrovascular toxicity through a distinct vascular pathway that can present with acute neurological symptoms overlapping with stroke, requiring MRI to distinguish vasogenic edema from ischemic injury. VEGF pathway inhibitors, including bevacizumab, sunitinib, and lenvatinib, are associated with posterior reversible encephalopathy syndrome (PRES) in addition to arterial thrombotic events. On MRI, PRES manifests as bilateral, predominantly subcortical vasogenic edema on FLAIR and T2-weighted sequences, most commonly involving the parieto-occipital regions, though other territories may be affected [ 35 ]. Recognition of PRES on imaging is critical, as prompt withdrawal of the offending agent typically leads to resolution.

11.2. Coronary and Cardiac Imaging

Turning from the cerebrovascular territory, the coronary and cardiac system represents a second important domain of cancer-associated arterial disease, encompassing a range of ischemic and structural complications that may be encountered on dedicated cardiac studies or identified incidentally on routine oncologic imaging. Cancer-associated acute coronary syndromes may result from accelerated atherosclerosis, therapy-induced coronary vasospasm, or direct prothrombotic effects [ 34 ]. Coronary CT angiography is particularly useful in cancer patients because it can noninvasively distinguish obstructive coronary artery disease from conditions that mimic acute coronary syndrome, such as fluoropyrimidine-associated vasospasm, immune checkpoint inhibitor-related myocarditis, and takotsubo cardiomyopathy, all of which may present with troponin elevation and chest pain [ 36 ].

Radiation therapy causes progressive macrovascular and microvascular injury within the radiation field. In the coronary territory, this manifests as accelerated atherosclerosis with a characteristic predilection for ostial lesions following mediastinal radiation. On CTA, these changes typically appear as long, smooth stenoses and may develop years to decades after treatment. Routine oncologic staging CT also provides an opportunity for incidental detection of coronary artery calcification. Even on non-ECG-gated scans, coronary calcification can be identified and provides prognostic information that may prompt cardiovascular risk factor modification in patients with favorable oncologic prognoses [ 34 , 36 ]. Radiologists should note the presence and extent of coronary calcification on staging CT reports when it is evident.

11.3. Peripheral Arterial Imaging

Beyond the coronary circulation, cancer-associated ATE also manifests in the peripheral arterial system, often through treatment-related mechanisms or in atypical distributions that may not conform to conventional atherosclerotic patterns. BCR-ABL tyrosine kinase inhibitors, particularly nilotinib and ponatinib, are associated with accelerated peripheral atherosclerosis and arterial occlusive events. Arterial thrombosis may occur in atypical locations or in the absence of traditional atherosclerotic risk factors, and CTA may reveal occlusions in otherwise healthy-appearing vessels. Similarly, radiation therapy directed at the head and neck, or axilla, can produce carotid or subclavian artery stenosis, respectively, with characteristic long-segment smooth narrowing on CTA or MRA that may present years after treatment [ 34 , 36 ].

Routine oncologic staging CT also serves as an opportunistic window for detecting incidental arterial findings that may represent the first imaging manifestation of cancer-associated arterial disease. Multifocal or bilateral splenic or renal infarcts, for instance, should raise suspicion for a systemic embolic process and warrant further cardiac evaluation. Similarly, aortic mural thrombus identified in the absence of significant underlying atherosclerosis may reflect the prothrombotic state of malignancy rather than conventional aortic disease. Clinically significant cardiovascular findings, including arterial thrombosis, arterial aneurysms, and intracardiac thrombus, are detected on staging CT in a meaningful proportion of cancer patients. As with incidental venous thromboembolism, these findings may have immediate management implications and may represent the first manifestation of cancer-associated arterial disease [ 5 , 36 ]. Radiologists should systematically evaluate the cardiovascular structures on all oncologic CT examinations, integrating arterial assessment into the same vigilant approach applied to venous thromboembolism throughout this review.

12. Role of FDG-PET

The preceding sections have addressed the imaging spectrum of cancer-associated venous and arterial thromboembolism, focusing on direct thrombus detection and characterization across conventional and unusual vascular territories. While cross-sectional imaging with CT and MRI remains the foundation of this diagnostic framework, 18F-FDG PET/CT occupies a complementary role that extends beyond anatomic thrombus identification. Rather than depicting thrombus directly, FDG-PET contributes to the CAT framework through three distinct applications: detection of occult malignancy in patients presenting with unprovoked VTE, metabolic characterization of thrombus to distinguish bland from tumor thrombus ( Figure 9 and Figure 10 ), and identification of active vasculitis as an independent thrombotic risk modifier. Each of these applications reflects a dimension of cancer-associated thromboembolism that conventional cross-sectional imaging alone cannot fully address, and together they define the unique diagnostic niche that FDG-PET occupies in the oncologic thrombosis workup.

12.1. FDG-PET in Occult Malignancy Detection

In patients who present with unprovoked VTE, excluding the presence of underlying malignancy may constitute a critical diagnostic priority, as undetected cancer may be present in a meaningful proportion of this population. FDG-PET imaging has demonstrated utility in this setting, achieving a sensitivity of 87.3% and a negative predictive value of 98.9% for occult cancer detection [ 13 , 14 ]. However, it should be noted that FDG-PET demonstrates poor sensitivity for detecting incidental VTE itself, with only 3% sensitivity for pulmonary embolism and 31% for deep vein thrombosis, despite high specificities of 99% and 88%, respectively, making it inadequate for routine VTE diagnosis ( Figure 11 ) [ 37 ].

12.2. Thrombus Characterization and Acute Versus Chronic DVT

Beyond malignancy detection, FDG-PET provides metabolic information that can aid in thrombus characterization, a capability with direct relevance to both DVT recurrence assessment and tumor versus bland thrombus differentiation. The FDG uptake that can sometimes be detected in thrombi is driven primarily by neutrophil-mediated inflammation rather than the thrombus itself, with metabolic activity reflecting glucose consumption by inflammatory cells infiltrating the clot. FDG-PET can potentially distinguish acute from chronic thrombosis, as metabolic activity peaks early after thrombus formation and declines progressively, returning to baseline at approximately 3 months after acute DVT. Recurrent same-site DVT demonstrates significantly higher FDG uptake than organized chronic DVT, reflecting the neutrophil-rich composition of recently formed thrombus [ 38 , 39 ].

Despite poor sensitivity for detecting incidental VTE itself, when avid thrombus is detected, FDG-PET can play a role in distinguishing malignant tumor thrombus from bland venous thrombosis. Tumor thrombi generally demonstrate higher SUVmax values than benign thrombi (SUVmax values ranging from 6.0 to 17.61 in tumor thrombi vs. 1.37 to 11.09 for benign thrombi), though considerable overlap exists and no universally accepted threshold has been established, reflecting the inherent limitations of FDG-PET for this purpose [ 40 , 41 , 42 , 43 ]. When tumor thrombus is identified, it carries substantial therapeutic implications, as it often requires oncologic intervention beyond anticoagulation alone, including surgical resection and systemic therapies, which are addressed in the Bland vs. Tumor Thrombus section of this review [ 44 ].

13. Vasculitis and Thrombotic Risk

FDG-PET plays an important diagnostic role in cancer-associated vasculitis, a condition directly relevant to this review because active vasculitis independently increases VTE risk and may be encountered on oncologic imaging in several distinct clinical contexts ( Figure 12 ).

13.1. Paraneoplastic and Treatment-Related Vasculitis

Paraneoplastic vasculitis occurs in a small but clinically significant proportion of cancer patients, with cutaneous leukocytoclastic vasculitis representing the most frequent histologic subtype [ 45 , 46 ]. The temporal relationship between paraneoplastic vasculitis and the underlying malignancy is variable. In some series, vasculitis antedates cancer diagnosis in nearly half of cases and flares have heralded tumor recurrence [ 46 ]; in others, the two diagnoses fall within a median of only 17 days of one another [ 47 ]. This unpredictable timing underscores why new vasculitides identified on FDG-PET should prompt consideration of occult or recurrent malignancy. A separate and increasingly recognized form of treatment-related vasculitis has emerged with the widespread adoption of immune checkpoint inhibitor therapy. This typically develops within the first several months of treatment initiation and generally responds to glucocorticoid therapy [ 48 , 49 ].

13.2. FDG-PET and CT Imaging of Vasculitis

Regardless of etiology, whether paraneoplastic, treatment-related, or arising independently of malignancy, FDG-PET has emerged as a valuable tool for detecting and characterizing vasculitides across vessel sizes. In large-vessel vasculitis, FDG-PET reveals circumferential segmental uptake in the vessel wall, with a sensitivity of 85% and specificity of 83% for differentiating vasculitis from other conditions [ 50 , 51 , 52 ]. Medium-vessel vasculitis demonstrates linear or focal muscular artery uptake patterns, while small-vessel disease presents with variable patterns depending on the organs involved [ 50 , 53 , 54 ]. On CT, complementary findings include vessel wall thickening, luminal irregularity, and abnormal perivascular enhancement. Beyond initial diagnosis, FDG-PET serves as an imaging biomarker for monitoring treatment response through the PET Vascular Activity Score, a semiquantitative grading system measuring the degree of FDG uptake within vessel walls. High PET Vascular Activity Scores during clinical remission predict future relapse in 55% of patients compared to 11% of those with low scores, underscoring its prognostic value [ 52 ].

13.3. Thrombotic Implications

The thrombotic implications of active vasculitis are substantial and represent the primary reason this entity warrants inclusion in a review of cancer-associated thromboembolism. Active vasculitis increases VTE risk through endothelial injury, inflammatory activation of coagulation pathways, and neutrophil extracellular trap formation expressing tissue factor [ 55 , 56 , 57 ]. Among the vasculitic syndromes most relevant to this population, ANCA-associated vasculitis carries a pooled VTE incidence of 12.4%, with the highest risk occurring in the first three months following diagnosis [ 58 , 59 ]. For the radiologist, recognizing the imaging features of active vasculitis on FDG-PET and CT is therefore important not only for the vasculitis diagnosis itself but also for identifying patients at elevated thrombotic risk who may warrant further clinical evaluation and thromboprophylaxis consideration.

14. Role of MRI

While CT and ultrasound form the foundation of most thromboembolism imaging protocols in oncology, MRI occupies a distinct and complementary role in clinical scenarios where anatomic detail, tissue characterization, or avoidance of ionizing radiation and iodinated contrast become priorities. Its applications in this setting span thrombus composition characterization, central venous evaluation, and pulmonary embolism diagnosis in patients for whom conventional first-line modalities are contraindicated or technically limited. As detailed in the Bland vs. Tumor Thrombus section of this review, MRI plays a central role in distinguishing tumor thrombus from bland thrombus through multiparametric assessment integrating arterial phase enhancement, diffusion-weighted imaging, and morphologic features including vessel expansion and tumor contiguity [ 60 , 61 ]. This capability is particularly consequential in hepatocellular carcinoma, renal cell carcinoma [ 62 , 63 ], and other malignancies with a recognized propensity for direct venous invasion, where accurate thrombus characterization directly influences staging and surgical planning.

MR Venography and MR Pulmonary Angiography

Among MRI’s specific applications in cancer-associated thromboembolism, MR venography merits particular attention for its ability to evaluate vascular territories that are inaccessible to conventional ultrasound. MRV is particularly useful for the pelvic veins and IVC, where bowel gas and body habitus limit sonographic evaluation, and in cancer patients, thrombosis in these territories may result from direct tumor compression, surgical changes, or the hypercoagulable state of malignancy. Accurate delineation of thrombus extent in these territories is essential for treatment planning, and MRV’s additional advantage of avoiding iodinated contrast and ionizing radiation makes it a practical alternative in patients with renal insufficiency or contrast allergy, populations that are disproportionately represented in oncology. MR pulmonary angiography (MRPA) is an alternative to CTPA for diagnosing pulmonary embolism in patients who cannot receive iodinated contrast or in whom cumulative radiation exposure is a concern. MRPA offers high specificity but lower sensitivity than CTPA, particularly for subsegmental emboli, and technically inadequate studies remain a recognized limitation [ 64 ]. As with MRV, its role is best understood as complementary rather than competitive with established first-line modalities, extending diagnostic access to patients for whom conventional approaches are not feasible.

15. Catheter-Related Thrombosis (CRT)

The preceding sections have addressed the imaging evaluation of thromboembolism arising from the malignancy itself or its systemic effects. However, a distinct and clinically important subset of cancer-associated thrombosis is directly attributable to the devices used to deliver treatment. Central venous access devices are essential for chemotherapy delivery in oncology patients, and radiologists are directly involved in their placement, tip positioning, and surveillance. Catheters are, however, inherently prothrombotic, promoting thrombus formation through all three components of Virchow’s triad: endothelial injury from mechanical trauma during insertion and ongoing vessel wall contact, venous stasis from partial luminal obstruction, and a foreign body response that triggers platelet adhesion and fibrin deposition, compounded by the underlying hypercoagulable state of malignancy. Catheter-related thrombosis (CRT) is accordingly a common complication in this population, with symptomatic upper extremity deep venous thrombosis (UEDVT) reported in approximately 3–14% of cancer patients with peripherally inserted central catheters (PICCs), though screening studies suggest asymptomatic thrombosis occurs at substantially higher rates [ 65 ]. Radiologists are therefore frequently the first to detect CRT, whether on dedicated diagnostic studies or incidentally on staging imaging.

15.1. Imaging Diagnosis

Compression ultrasonography with duplex Doppler is the recommended first-line imaging modality for suspected CRT [ 66 ]. Ultrasound performs well for the jugular, axillary, basilic, cephalic, and brachial veins, with reported sensitivity and specificity above 80% for symptomatic UEDVT [ 67 ]. However, the proximal subclavian and brachiocephalic veins are difficult to visualize directly because of overlying bony structures, and compressibility cannot be assessed in these locations. When ultrasound of the distal veins appears patent, but proximal thrombosis is suspected, an abnormal Doppler pattern, particularly reduced or absent respiratory variation in flow velocity, may serve as an indirect sign of more central obstruction [ 64 , 67 ].

When ultrasound is negative or indeterminate and clinical suspicion remains high, CT venography (CTV) or MR venography (MRV) can evaluate the central veins, including the brachiocephalic vein, superior vena cava, and right atrium, which are beyond the reliable reach of ultrasound [ 68 ]. CTV and MRV are also useful when CRT is suspected in the setting of complex anatomy, such as after prior thrombosis, surgery, or radiation to the chest. Contrast venography, while rarely used as a primary diagnostic tool, remains the reference standard and may be performed when catheter-directed intervention is planned [ 66 , 69 ].

15.2. Imaging-Guided Catheter Management

A key clinical question in CRT is whether the catheter should be removed or retained. Current guidelines support catheter retention during anticoagulation when imaging confirms that the catheter is functional, appropriately positioned, and free of infection [ 68 , 70 ]. Imaging plays a central role in this determination: radiologists should assess catheter tip position, confirm the absence of malpositioning or kinking, and evaluate for thrombus progression on follow-up studies. When imaging demonstrates persistent or progressive thrombus despite anticoagulation, or when the catheter is no longer functional, removal or escalation of therapy should be considered [ 68 ].

In select patients, particularly those with severe or rapidly progressive thrombosis or central thrombus propagation despite anticoagulation, catheter-directed therapy including pharmacochemical thrombolysis or mechanical thrombectomy may be considered in conjunction with interventional radiology. Imaging surveillance with serial ultrasound is also indicated in patients who have contraindications to anticoagulation and in whom the catheter has been retained, to monitor for thrombus progression [ 68 , 71 ].

16. Endovascular Management of Cancer-Associated Venous Thromboembolism

The radiologist’s role in cancer-associated thromboembolism extends beyond diagnosis and characterization into active therapeutic participation. Interventional radiology offers a range of adjunctive and definitive treatment strategies for acute, chronic, and refractory thrombotic disease in oncology patients, and the imaging knowledge that informs diagnosis is equally central to procedural planning, patient selection, and post-intervention surveillance. The following sections address the principal endovascular approaches available in this setting, with emphasis on the indications, limitations, and imaging considerations most relevant to the oncologic context [ 72 ].

16.1. Inferior Vena Cava Filters

Among the endovascular strategies available for cancer-associated VTE, IVC filtration, catheter-directed thrombolysis, and mechanical thrombectomy each occupy distinct roles determined by the acuity of disease, the patient’s bleeding risk, and the feasibility of anticoagulation, considerations that are particularly complex in the oncology population. IVC filters serve primarily as a preventive modality in cancer patients with absolute or relative contraindications to anticoagulation, or in those with recurrent VTE despite adequate therapy. Absolute indications include active bleeding and inability to maintain therapeutic anticoagulation, while relative indications encompass massive PE, limited cardiopulmonary reserve, free-floating IVC thrombus, and anticipated treatment noncompliance [ 73 ]. Although permanent IVC filters have demonstrated long-term reductions in PE incidence, as shown in the PREPIC trial, they are associated with increased DVT rates, carry the additional risk of filter-related complications, including fracture, migration, caval perforation, and IVC thrombosis, and demonstrate no mortality benefit [ 72 , 73 ]. Retrievable filters are preferred in oncology patients given the expectation that anticoagulation may eventually be feasible, though retrieval rates in this population remain suboptimal in practice [ 73 ]. Filter placement should therefore be individualized, with regular reassessment of retrieval candidacy as the clinical course evolves.

16.2. Catheter-Directed Thrombolysis

For patients in whom anticoagulation is feasible and more aggressive thrombus clearance is warranted, catheter-directed thrombolysis (CDT) offers an alternative to systemic anticoagulation alone. CDT involves the localized infusion of thrombolytic agents directly into thrombus via an endovascularly placed catheter, offering more rapid clot lysis than systemic anticoagulation alone and with a lower hemorrhagic risk than systemic thrombolysis. Its primary indication in the venous system is acute iliofemoral DVT, where early thrombus clearance may preserve valvular function and reduce the severity of post-thrombotic syndrome, a complication of particular consequence in oncology patients who may face prolonged periods of reduced mobility and repeated venous instrumentation. However, its application in this population is substantially limited both by the absence of direct trial evidence and by the high prevalence of contraindications that are disproportionately common in oncology patients. Relative contraindications, including recent surgery, organ biopsy, coagulopathy, and brain metastases, are frequently present in this population and require careful weighing against the potential benefit of thrombus clearance [ 70 , 74 ]. Absolute contraindications, including active bleeding, recent intracranial events, and intracranial tumors, preclude its use entirely in a meaningful subset of patients. The ATTRACT trial, the largest randomized evaluation of CDT in iliofemoral DVT, excluded patients with active cancer entirely, and its findings therefore cannot be directly extrapolated to this population [ 75 ]. The combination of CDT and systemic anticoagulation has also been associated with increased intracranial hemorrhage risk compared to anticoagulation alone [ 70 ], further narrowing the eligible population. CDT is therefore used selectively and with multidisciplinary input in oncology patients [ 72 , 74 ].

16.3. Percutaneous Mechanical Thrombectomy

Given the frequency of thrombolytic contraindications in the oncology population, percutaneous mechanical thrombectomy (PMT), a catheter-based technique that physically disrupts and aspirates thrombus without the use of thrombolytic agents, has emerged as a favored endovascular strategy in many cancer patients, relying on mechanical clot removal with systemic anticoagulation alone. Device selection is guided by an individualized assessment of clot burden, vessel anatomy, and the presence of intravascular hardware such as IVC filters or stents, with aspiration systems, basket retrievers, and stent-retriever platforms each offering distinct mechanical advantages depending on these variables [ 72 ]. Where residual stenosis or extrinsic tumor compression limits flow restoration, adjunctive angioplasty and venous stenting may be performed. Complication rates in institutional series have been low, with meaningful reductions in post-thrombotic syndrome severity reported in the majority of treated patients. Notably, pathologic evaluation of retrieved clot has identified tumor thrombus in a meaningful proportion of cases, underscoring the potential diagnostic value of PMT beyond its therapeutic purpose and reinforcing the importance of submitting retrieved material for histologic analysis [ 76 ].

16.4. Endovascular Management of Pulmonary Embolism

Thrombolytic-sparing principles are particularly relevant in the management of high-risk PE, where mechanical thrombectomy offers an alternative when systemic thrombolysis is contraindicated or has failed. In hemodynamically stable intermediate-risk PE, intervention is generally reserved for patients who deteriorate despite anticoagulation. In cancer patients presenting with high-risk or deteriorating intermediate-risk PE, management is increasingly coordinated through multidisciplinary Pulmonary Embolism Response Teams (PERT), integrating interventional radiology, cardiology, and critical care expertise to guide timely treatment decisions, though institutional availability varies [ 72 ]. The PEERLESS trial demonstrated a composite clinical endpoint favoring large-bore mechanical thrombectomy over catheter-directed thrombolysis in intermediate-risk PE, with benefits driven primarily by reductions in clinical deterioration and more rapid recovery of right ventricular function. While cancer patients were not the primary focus of this trial, they represent an important subgroup for whom thrombolytic-sparing approaches may be particularly attractive given their elevated baseline bleeding risk, though prospective data specifically addressing this subgroup remain limited [ 77 ].

16.5. Portal Vein Thrombosis Interventions

Portal vein thrombosis in cancer patients carries significant implications for surgical candidacy, liver transplant eligibility, and locoregional treatment planning, as addressed in the venous thromboembolism section of this review. In patients with minimally symptomatic PVT, anticoagulation alone remains the standard approach. However, endovascular intervention becomes necessary in patients with progressive portal hypertension, deteriorating hepatic function, variceal hemorrhage risk, or time-sensitive surgical or transplant considerations that cannot await the natural history of anticoagulation alone. The choice of technique is guided by the clinical urgency, thrombus burden, and underlying oncologic context. In patients requiring rapid portal decompression, trans-jugular intrahepatic portosystemic shunt (TIPS) placement, which creates a direct intrahepatic communication between the portal and hepatic venous systems to reduce portal pressure, addresses both the thrombosis and the resulting portal hypertension simultaneously. Where recanalization of the portal vein itself is the primary goal, transhepatic portal access with mechanical thrombectomy or sharp recanalization followed by venous stenting may restore flow and preserve treatment options in patients approaching transplant or major hepatic resection. As with other endovascular interventions in oncology patients, the decision to intervene should be made within a multidisciplinary framework that accounts for the patient’s overall disease trajectory, performance status, and treatment goals [ 72 ].

Collectively, the endovascular strategies described in this section extend the radiologist’s role in cancer-associated thromboembolism beyond diagnosis and characterization into active therapeutic participation. Their appropriate application requires not only procedural expertise but a detailed understanding of thrombus morphology, vessel anatomy, and the oncologic context that distinguishes this population from general interventional practice.

17. Cancer-Associated Thromboembolism in the Era of Artificial Intelligence

Artificial intelligence (AI) and machine learning (ML) are under active investigation to outline their role in the diagnosis and risk stratification of cancer-associated thromboembolism (CAT). Potential applications include triaging, image acquisition and analysis, predictive modeling, and automated case detection. However, no AI tools have yet achieved routine clinical implementation for CAT, and current evidence is limited by retrospective design, single-institution datasets, and a lack of prospective validation. This section reviews the most relevant applications focused on imaging interpretation and predictive modeling with explicit attention to promise and current limitations.

18. Incidental Pulmonary Embolism Detection on Oncology Imaging

Incidental pulmonary embolism (PE) on routine staging and surveillance CT represents a distinct diagnostic context from suspected PE evaluated with dedicated CTPA. Miss rates for incidental PE on non-dedicated CT have been reported as high as 50–79% in retrospective cohorts.

Emerging evidence suggests that AI-based algorithms may reduce this diagnostic gap, though most studies remain preliminary. A prospective single-center study reported that implementation of an AI triage system was associated with increased radiologist sensitivity for incidental PE detection from 80% to 96% on contrast-enhanced chest and abdominal CT. A retrospective analysis of over 1000 oncology patients found that an AI algorithm identified 91% of incidental PEs, including cases not initially reported and those involving subsegmental arteries. Following clinical implementation of an FDA-cleared algorithm at one institution, missed incidental PE rates decreased from 50% to 7%. Collectively, these findings suggest that AI may serve as a useful second-reader tool in routine oncology imaging. However, important limitations must be acknowledged: all cited studies were single-center, most were retrospective, and none have demonstrated improved patient outcomes (e.g., reduced VTE-related mortality or morbidity). Prospective multi-center validation is needed before widespread adoption [ 78 , 79 , 80 ].

19. Pulmonary Embolism Detection on Dedicated CTPA

For dedicated CTPA performed in the setting of suspected PE, deep learning models have demonstrated pooled sensitivity of 88% and specificity of 86% in meta-analysis, although prospective validation in oncology-specific populations remains limited. In one prospective study, AI triage reduced time to identification of positive PE studies from approximately 21 to 11 min, without a significant change in radiologist accuracy or miss rates. Taken together, these results suggest that AI may offer workflow benefits in the acute diagnostic setting, although its added diagnostic value beyond expert interpretation in dedicated CTPA remains uncertain. The difference in performance between surveillance imaging and symptom-driven CTPA likely reflects differences in clinical context, including prevalence, pre-test probability, and reader focus [ 81 , 82 ].

20. Risk Prediction Models

Machine learning models trained on electronic health record data have demonstrated improved discrimination for cancer-associated thrombosis (CAT) risk compared with the Khorana score, which is a static baseline-only risk model. A systematic review found that ML approaches, particularly gradient boosting algorithms, generally outperform the Khorana score across datasets, though the magnitude of improvement varies substantially by cohort. More recent approaches enable longitudinal risk prediction throughout cancer therapy rather than only at treatment initiation. One model achieved an AUROC of 0.71 for predicting CAT within 90 days of treatment cycles, compared with 0.57 for the Khorana score. A liquid biopsy-based ML model incorporating circulating tumor DNA achieved c-indices of 0.73–0.74 for venous thromboembolism prediction, outperforming conventional clinical risk models in a single retrospective study. These approaches could potentially complement imaging-based surveillance by helping identify high-risk patients who might benefit from closer radiologic follow-up. However, several critical limitations remain: (1) all cited models lack prospective validation, (2) most were developed and tested on single-institution data, (3) generalizability across different cancer types and treatment protocols is unproven, and (4) no study has demonstrated that model-guided surveillance improves clinical outcomes. These methods remain investigational and should not yet be used to guide clinical decision-making [ 83 , 84 , 85 ].

21. Large Language Models

Large language models (LLMs), appear to be a versatile tool in radiology, with applications extending substantially beyond image interpretation and radiomics into integration of clinical history, knowledge retrieval and consequent decision support [ 86 , 87 ]. LLMs also can translate complex findings into patient-friendly summaries, thereby improving communication, shared decision-making. However, to date, no published studies have specifically evaluated LLMs for CAT detection, risk prediction, or management. Current evidence is limited to proof-of-concept studies in general radiology, and significant concerns remain regarding hallucination (generation of false information), lack of domain-specific training, and absence of regulatory approval for clinical use in thrombotic disease [ 88 ].

22. Challenges and Limitations

Despite substantial progress and increasing investment in artificial intelligence (AI) for medical imaging, the integration of AI into routine radiology practice for cancer-associated thromboembolism (CAT) detection and risk stratification remains limited [ 89 ]. Multiple interrelated challenges constrain widespread clinical adoption of AI tools in this domain.

Human factors: Limited AI illiteracy among practicing radiologists represents a significant barrier. Few radiologists have received formal training in AI development and validation, and the rapid pace of technological advancement has outpaced continuing medical education offerings. Additionally, resistance to AI tool adoption exists, driven in part by concerns about false-positive findings and perceived threats to professional autonomy. The steep learning curve associated with new AI interfaces requires time and effort that many radiologists find difficult to accommodate within clinical workflows [ 90 , 91 , 92 , 93 ].

Technical challenges: AI models are vulnerable to several technical limitations that are particularly relevant to CAT. Algorithmic bias—stemming from training data that may underrepresent certain cancer types, demographic groups, or imaging protocols—can lead to systematically inaccurate predictions [ 90 , 94 , 95 ].

Poor reasoning transparency (the “black box” problem) limits clinicians’ ability to understand why a particular finding was flagged or dismissed. Lack of standardization across AI tools, imaging platforms, and reconstruction algorithms complicates multi-institutional deployment. For CAT specifically, few AI tools have undergone rigorous prospective validation in oncology populations, limiting confidence in real-world performance [ 90 ].

Ethical, legal, and Financial Barriers: Data security and patient privacy concerns remain incompletely addressed, with no clear oversight mechanism to ensure HIPAA compliance for cloud-based AI processing. Cost barriers, including purchase, integration, and maintenance expenses, may exacerbate existing health care access disparities between well-resourced and under-resourced institutions. Strict human oversight—with radiologists remaining the final authority on all outputs—is a logical necessity, but governance frameworks for this oversight are not yet standardized [ 93 , 96 , 97 , 98 ].

In summary, while AI holds theoretical promise for CAT detection and risk stratification, substantial barriers to clinical implementation remain. Prospective multi-center validation, standardized reporting guidelines, and clear regulatory frameworks are needed before AI tools can be responsibly integrated into routine oncologic imaging practice. To date, no AI tool can make an autonomous decision in CAT diagnosis and treatment, and human oversight remains essential with radiologists/physicians are the final authority in all outputs ( Table 3 ).

Key advices for Radiologists and physicians:

Always assess for incidental thromboembolism on routine staging CT—it carries prognostic weight equivalent to symptomatic events.

Distinguish bland from tumor thrombus using multiparametric imaging (enhancement, vessel expansion, DWI, FDG-PET).

Report incidental thromboembolism with structured elements (location, acuity, burden, hemodynamic relevance, comparison, device relationship).

Evaluate cardiovascular structures systematically on all oncologic CT—arterial events are underrecognized.

Know when to recommend advanced imaging (MRI for pelvic/IVC thrombosis, PET/CT for occult malignancy in unprovoked VTE).

Recognize catheter-related thrombosis and guide catheter retention vs. removal decisions.

Be aware of AI limitations—no current AI tool is clinically routine for CAT; human oversight remains essential. Please refer to Table 4 .

23. Conclusions

Cancer-associated thrombosis (CAT) extends beyond DVT and PE. The radiologist’s role spans venous and arterial thromboembolism, catheter-related complications, and endovascular management. Key themes include incidental thromboembolism (prognostically equivalent to symptomatic events), distinction between bland and tumor thrombus (affecting staging and treatment), and a multimodal imaging toolkit (ultrasound, CT, MRI, FDG-PET/CT). AI shows promise but requires validation; human oversight remains essential. Interventional radiology offers thrombolytic-sparing mechanical thrombectomy for high-bleeding-risk patients.

Realizing this potential requires a synthesized approach incorporating the patient’s treatment history, biomarker status, and thrombotic risk profile. This review supports oncology-informed radiologic practice, positioning the radiologist as a central figure in oncologic care.

Abbreviations

Author Contributions

J.B. and O.A.E.M.: The first and co-first author contributed equally to the work and share first authorship: Their work includes data curation, visualization, writing the original daft and review and editing. J.H.M.: Preparation of images, data curation and writing—original draft. H.M.: Writing—original draft and review and editing. A.H.: Conceptualization, investigation, visualization, writing—original draft and final review and editing. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

No new data were created or analyzed in this review. All referenced data sources are publicly available in the peer-reviewed literature as cited in the reference list.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This research received no external funding.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

References

1. Mukai M., Oka T. Mechanism and management of cancer-associated thrombosis. J. Cardiol. 2018;72:89–93. doi: 10.1016/j.jjcc.2018.02.011.

2. Eichinger S. Cancer associated thrombosis: Risk factors and outcomes. Thromb. Res. 2016;140:S12–S17. doi: 10.1016/S0049-3848(16)30092-5.

3. Sorensen H.T., Pedersen L., van Es N., Buller H.R., Horvath-Puho E. Impact of venous thromboembolism on the mortality in patients with cancer: A population-based cohort study. Lancet Reg. Health Eur. 2023;34:100739. doi: 10.1016/j.lanepe.2023.100739.

4. Patel H.K., Khorana A.A. Anticoagulation in Cancer Patients: A Summary of Pitfalls to Avoid. Curr. Oncol. Rep. 2019;21:18. doi: 10.1007/s11912-019-0767-5.

5. Navi B.B., Reiner A.S., Kamel H., Iadecola C., Okin P.M., Elkind M.S.V., Panageas K.S., DeAngelis L.M. Risk of Arterial Thromboembolism in Patients With Cancer. J. Am. Coll. Cardiol. 2017;70:926–938. doi: 10.1016/j.jacc.2017.06.047.

6. Donadini M.P., Ageno W. Unusual site thrombosis. Semin. Hematol. 2011;48:264–270. doi: 10.1053/j.seminhematol.2011.08.005.

7. Menapace L.A., Peterson D.R., Berry A., Sousou T., Khorana A.A. Symptomatic and incidental thromboembolism are both associated with mortality in pancreatic cancer. Thromb. Haemost. 2011;106:371–378. doi: 10.1160/TH10-12-0789.

8. Gervaso L., Dave H., Khorana A.A. Venous and Arterial Thromboembolism in Patients With Cancer: JACC: CardioOncology State-of-the-Art Review. JACC CardioOncol. 2021;3:173–190. doi: 10.1016/j.jaccao.2021.03.001.

9. Chew H.K., Wun T., Harvey D., Zhou H., White R.H. Incidence of venous thromboembolism and its effect on survival among patients with common cancers. Arch. Intern. Med. 2006;166:458–464. doi: 10.1001/archinte.166.4.458.

10. Sorensen H.T., Mellemkjaer L., Olsen J.H., Baron J.A. Prognosis of cancers associated with venous thromboembolism. N. Engl. J. Med. 2000;343:1846–1850. doi: 10.1056/NEJM200012213432504.

11. Trujillo-Santos J., Martos F.M., Font C., Farge-Bancel D., Rosa V., Lorenzo A., Barron M., Lorente M.A., Pedrajas J.M., Monreal M. Analysis of clinical factors affecting the rates of fatal pulmonary embolism and bleeding in cancer patients with venous thromboembolism. Heliyon. 2017;3:e00229. doi: 10.1016/j.heliyon.2016.e00229.

12. Falanga A., Marchetti M. Cancer-associated thrombosis: Enhanced awareness and pathophysiologic complexity. J. Thromb. Haemost. 2023;21:1397–1408. doi: 10.1016/j.jtha.2023.02.029.

13. Robin P., van Es N., Le Roux P.Y., Rondina M., Lecumberri R., Beckers M., Le Gal G., Salaun P.Y. Performance of 18F-fluorodesoxyglucose positron-emission tomography/computed tomography for cancer screening in patients with unprovoked venous thromboembolism: Results from an individual patient data meta-analysis. Thromb. Res. 2020;194:153–157. doi: 10.1016/j.thromres.2020.06.034.

14. Robin P., Le Roux P.-Y., Planquette B., Accassat S., Roy P.-M., Couturaud F., Ghazzar N., Prevot-Bitot N., Couturier O., Delluc A., et al. Limited screening with versus without 18F-fluorodeoxyglucose PET/CT for occult malignancy in unprovoked venous thromboembolism: An open-label randomised controlled trial. Lancet Oncol. 2016;17:193–199. doi: 10.1016/S1470-2045(15)00480-5.

15. Wan T., Song J., Zhu D. Cancer-associated venous thromboembolism: A comprehensive review. Thromb. J. 2025;23:35. doi: 10.1186/s12959-025-00719-7.

16. Khorana A.A., Francis C.W., Blumberg N., Culakova E., Refaai M.A., Lyman G.H. Blood transfusions, thrombosis, and mortality in hospitalized patients with cancer. Arch. Intern. Med. 2008;168:2377–2381. doi: 10.1001/archinte.168.21.2377.

17. Konstantinides S.V., Meyer G., Becattini C., Bueno H., Geersing G.J., Harjola V.P., Huisman M.V., Humbert M., Jennings C.S., Jimenez D., et al. 2019 ESC Guidelines for the diagnosis and management of acute pulmonary embolism developed in collaboration with the European Respiratory Society (ERS) Eur. Heart J. 2020;41:543–603. doi: 10.1093/eurheartj/ehz405.

18. Mazzolai L., Aboyans V., Ageno W., Agnelli G., Alatri A., Bauersachs R., Brekelmans M.P.A., Buller H.R., Elias A., Farge D., et al. Diagnosis and management of acute deep vein thrombosis: A joint consensus document from the European Society of Cardiology working groups of aorta and peripheral vascular diseases and pulmonary circulation and right ventricular function. Eur. Heart J. 2018;39:4208–4218. doi: 10.1093/eurheartj/ehx003.

19. Di Nisio M., van Es N., Buller H.R. Deep vein thrombosis and pulmonary embolism. Lancet. 2016;388:3060–3073. doi: 10.1016/S0140-6736(16)30514-1.

20. Righini M., Le Gal G., Bounameaux H. Venous thromboembolism diagnosis: Unresolved issues. Thromb. Haemost. 2015;113:1184–1192. doi: 10.1160/th14-06-0530.

21. D’Oria M., Girardi L., Amgad A., Sherif M., Piffaretti G., Ruaro B., Calvagna C., Dueppers P., Lepidi S., Donadini M.P. Expert-Based Narrative Review on Compression UltraSonography (CUS) for Diagnosis and Follow-Up of Deep Venous Thrombosis (DVT) Diagnostics. 2025;15:82. doi: 10.3390/diagnostics15010082.

22. LeGout J.D., Bailey R.E., Bolan C.W., Bowman A.W., Chen F., Cernigliaro J.G., Alexander L.F. Multimodality Imaging of Abdominopelvic Tumors with Venous Invasion. Radiographics. 2020;40:2098–2116. doi: 10.1148/rg.2020200047.

23. Moore A.J.E., Wachsmann J., Chamarthy M.R., Panjikaran L., Tanabe Y., Rajiah P. Imaging of acute pulmonary embolism: An update. Cardiovasc. Diagn. Ther. 2018;8:225–243. doi: 10.21037/cdt.2017.12.01.

24. Douma R.A., Kok M.G., Verberne L.M., Kamphuisen P.W., Buller H.R. Incidental venous thromboembolism in cancer patients: Prevalence and consequence. Thromb. Res. 2010;125:e306–e309. doi: 10.1016/j.thromres.2010.02.010.

25. Singh R., Sousou T., Mohile S., Khorana A.A. High rates of symptomatic and incidental thromboembolic events in gastrointestinal cancer patients. J. Thromb. Haemost. 2010;8:1879–1881. doi: 10.1111/j.1538-7836.2010.03929.x.

26. Ageno W., Squizzato A., Togna A., Magistrali F., Mangini M., Fugazzola C., Dentali F. Incidental diagnosis of a deep vein thrombosis in consecutive patients undergoing a computed tomography scan of the abdomen: A retrospective cohort study. J. Thromb. Haemost. 2012;10:158–160. doi: 10.1111/j.1538-7836.2011.04565.x.

27. den Exter P.L., Hooijer J., Dekkers O.M., Huisman M.V. Risk of recurrent venous thromboembolism and mortality in patients with cancer incidentally diagnosed with pulmonary embolism: A comparison with symptomatic patients. J. Clin. Oncol. 2011;29:2405–2409. doi: 10.1200/JCO.2010.34.0984.

28. Shinagare A.B., Guo M., Hatabu H., Krajewski K.M., Andriole K., Van den Abbeele A.D., DiPiro P.J., Nishino M. Incidence of pulmonary embolism in oncologic outpatients at a tertiary cancer center. Cancer. 2011;117:3860–3866. doi: 10.1002/cncr.25941.

29. Vazqez S., Das A., Spirollari E., Brabant P., Nolan B., Clare K., Dominguez J.F., Dangayach N., Amuluru K., Yaghi S., et al. Inpatient Outcomes of Cerebral Venous Thrombosis in Patients With Malignancy Throughout the United States. J. Stroke. 2024;26:425–433. doi: 10.5853/jos.2023.04098.

30. Saposnik G., Bushnell C., Coutinho J.M., Field T.S., Furie K.L., Galadanci N., Kam W., Kirkham F.C., McNair N.D., Singhal A.B., et al. Diagnosis and Management of Cerebral Venous Thrombosis: A Scientific Statement From the American Heart Association. Stroke. 2024;55:e77–e90. doi: 10.1161/STR.0000000000000456.

31. Cohen O., Caiano L.M., Tufano A., Ageno W. Cancer-Associated Splanchnic Vein Thrombosis. Semin. Thromb. Hemost. 2021;47:931–941. doi: 10.1055/s-0040-1722607.

32. Garcia-Pagan J.C., Valla D.C. Primary Budd-Chiari Syndrome. N. Engl. J. Med. 2023;388:1307–1316. doi: 10.1056/NEJMra2207738.

33. Finelli P.F., Nouh A. Three-Territory DWI Acute Infarcts: Diagnostic Value in Cancer-Associated Hypercoagulation Stroke (Trousseau Syndrome) AJNR Am. J. Neuroradiol. 2016;37:2033–2036. doi: 10.3174/ajnr.A4846.

34. Cheong M.A., Leader A. Cancer and arterial thrombosis: Therapeutic options. Res. Pract. Thromb. Haemost. 2024;8:102393. doi: 10.1016/j.rpth.2024.102393.

35. Fugate J.E., Hawkes M.A., Rabinstein A.A. Posterior reversible encephalopathy syndrome: Evolving insights in diagnosis, management, and outcomes. Lancet Neurol. 2025;24:789–800. doi: 10.1016/S1474-4422(25)00232-7.

36. Baldassarre L.A., Ganatra S., Lopez-Mattei J., Yang E.H., Zaha V.G., Wong T.C., Ayoub C., DeCara J.M., Dent S., Deswal A., et al. Advances in Multimodality Imaging in Cardio-Oncology: JACC State-of-the-Art Review. J. Am. Coll. Cardiol. 2022;80:1560–1578. doi: 10.1016/j.jacc.2022.08.743.

37. Le Roux P.Y., Robin P., Delluc A., Tardy B., Abgral R., Couturaud F., Reffad A., Le Gal G., Salaun P.Y. Performance of 18F fluoro-2-desoxy-D-glucose positron emission tomography/computed tomography for the diagnosis of venous thromboembolism. Thromb. Res. 2015;135:31–35. doi: 10.1016/j.thromres.2014.10.008.

38. Hara T., Truelove J., Tawakol A., Wojtkiewicz G.R., Hucker W.J., MacNabb M.H., Brownell A.L., Jokivarsi K., Kessinger C.W., Jaff M.R., et al. 18F-fluorodeoxyglucose positron emission tomography/computed tomography enables the detection of recurrent same-site deep vein thrombosis by illuminating recently formed, neutrophil-rich thrombus. Circulation. 2014;130:1044–1052. doi: 10.1161/CIRCULATIONAHA.114.008902.

39. Rondina M.T., Lam U.T., Pendleton R.C., Kraiss L.W., Wanner N., Zimmerman G.A., Hoffman J.M., Hanrahan C., Boucher K., Christian P.E., et al. 18F-FDG PET in the evaluation of acuity of deep vein thrombosis. Clin. Nucl. Med. 2012;37:1139–1145. doi: 10.1097/RLU.0b013e3182638934.

40. Sharma P., Kumar R., Jeph S., Karunanithi S., Naswa N., Gupta A., Malhotra A. 18F-FDG PET-CT in the diagnosis of tumor thrombus: Can it be differentiated from benign thrombus? Nucl. Med. Commun. 2011;32:782–788. doi: 10.1097/MNM.0b013e32834774c8.

41. Lee E.Y., Khong P.L. The value of 18F-FDG PET/contrast-enhanced CT in detection of tumor thrombus. Clin. Nucl. Med. 2013;38:e60–e65. doi: 10.1097/RLU.0b013e318266d53e.

42. Sarswat S., Khan D., Sagar S., Kumar R. Role of 18F-FDG PET/CT in the diagnosis of benign vs. malignant tumor thrombus. Nucl. Med. Commun. 2023;44:726–731. doi: 10.1097/MNM.0000000000001708.

43. Zhu A.H., Hou X.Y., Tian S., Zhang W.F. Diagnostic value of 18F-FDG PET/CT versus contrast-enhanced MRI for venous tumour thrombus and venous bland thrombus in renal cell carcinoma. Sci. Rep. 2022;12:587. doi: 10.1038/s41598-021-04541-9.

44. Kumar S.A., Jain S., Goyal H., Halanaik D. Imaging tumor thrombus across malignancies: Current concepts and the expanding role of PET/computed tomography. Nucl. Med. Commun. 2026;47:267–279. doi: 10.1097/MNM.0000000000002090.

45. Loricera J., Calvo-Rio V., Ortiz-Sanjuan F., Gonzalez-Lopez M.A., Fernandez-Llaca H., Rueda-Gotor J., Gonzalez-Vela M.C., Alvarez L., Mata C., Gonzalez-Lamuno D., et al. The spectrum of paraneoplastic cutaneous vasculitis in a defined population: Incidence and clinical features. Medicine. 2013;92:331–343. doi: 10.1097/MD.0000000000000009.

46. Solans-Laque R., Bosch-Gil J.A., Perez-Bocanegra C., Selva-O’Callaghan A., Simeon-Aznar C.P., Vilardell-Tarres M. Paraneoplastic vasculitis in patients with solid tumors: Report of 15 cases. J. Rheumatol. 2008;35:294–304.

47. Do H., Pyo J.Y., Ahn S.S., Song J.J., Park Y.B., Lee S.W. Cancer-associated vasculitides: A single-centre case series. Clin. Rheumatol. 2023;42:151–158. doi: 10.1007/s10067-022-06363-0.

48. Chanson N., Galvagni A., Ramos-Casals M., Ruiz J.I., Suijkerbuijk K.P.M., Gente K., Kerschen P., Karam J.D., Belkhir R., Outh R., et al. Immune checkpoint inhibitors-associated vasculitis: A heterogeneous condition with possible severe disease course. Rheumatology. 2025;64:3685–3690. doi: 10.1093/rheumatology/keae711.

49. Lee C.M., Wang M., Rajkumar A., Calabrese C., Calabrese L. A scoping review of vasculitis as an immune-related adverse event from checkpoint inhibitor therapy of cancer: Unraveling the complexities at the intersection of immunology and vascular pathology. Semin. Arthritis Rheum. 2024;66:152440. doi: 10.1016/j.semarthrit.2024.152440.

50. Cacoub P., Vieira M., Langford C.A., Tazi Mezalek Z., Saadoun D. Large-vessel vasculitis. Lancet. 2025;406:2017–2032. doi: 10.1016/S0140-6736(25)01436-9.

51. Slart R., Nienhuis P.H., Glaudemans A., Brouwer E., Gheysens O., van der Geest K.S.M. Role of 18F-FDG PET/CT in Large Vessel Vasculitis and Polymyalgia Rheumatica. J. Nucl. Med. 2023;64:515–521. doi: 10.2967/jnumed.122.265016.

52. Grayson P.C., Alehashemi S., Bagheri A.A., Civelek A.C., Cupps T.R., Kaplan M.J., Malayeri A.A., Merkel P.A., Novakovich E., Bluemke D.A., et al. 18F-Fluorodeoxyglucose-Positron Emission Tomography As an Imaging Biomarker in a Prospective, Longitudinal Cohort of Patients With Large Vessel Vasculitis. Arthritis Rheumatol. 2018;70:439–449. doi: 10.1002/art.40379.

53. Fagart A., Machet T., Collet G., Quemeneur T., Ben Ticha R., Verstraete M., Le Gouellec N., Demailly F., Rousselin C. Fluorodeoxyglucose positron emission tomography-computed tomography findings in a first series of 10 patients with polyarteritis nodosa. Rheumatology. 2022;61:1663–1668. doi: 10.1093/rheumatology/keab591.

54. Chen Z., Zhao Y., Wang Q., Li Y., Li H., Zhou Y. Imaging features of 18F-FDG PET/CT in different types of systemic vasculitis. Clin. Rheumatol. 2022;41:1499–1509. doi: 10.1007/s10067-021-06039-1.

55. Bello F., Fagni F., Bagni G., Hill C.L., Mohammad A.J., Moiseev S., Olivotto I., Seyahi E., Emmi G. Arterial and venous thrombosis in systemic and monogenic vasculitis. Nat. Rev. Rheumatol. 2025;21:355–369. doi: 10.1038/s41584-025-01252-7.

56. Zhu K., Lv F., Hou X., Wang F., Pang L., Zhong M. Thrombosis in vasculitis: An updated review of etiology, pathophysiology, and treatment. Heliyon. 2024;10:e30615. doi: 10.1016/j.heliyon.2024.e30615.

57. Misra D.P., Thomas K.N., Gasparyan A.Y., Zimba O. Mechanisms of thrombosis in ANCA-associated vasculitis. Clin. Rheumatol. 2021;40:4807–4815. doi: 10.1007/s10067-021-05790-9.

58. Hansrivijit P., Trongtorsak A., Gadhiya K.P., Lnu K., Dimech C.T., Thongprayoon C., Cheungpasitporn W., Lecker S.H. Incidence and risk factors of venous thromboembolism in ANCA-associated vasculitis: A metaanalysis and metaregression. Clin. Rheumatol. 2021;40:2843–2853. doi: 10.1007/s10067-021-05589-8.

59. Liapi M., Jayne D., Merkel P.A., Segelmark M., Mohammad A.J. Venous thromboembolism in ANCA-associated vasculitis: A population-based cohort study. Rheumatology. 2021;60:4616–4623. doi: 10.1093/rheumatology/keab057.

60. Catalano O.A., Choy G., Zhu A., Hahn P.F., Sahani D.V. Differentiation of malignant thrombus from bland thrombus of the portal vein in patients with hepatocellular carcinoma: Application of diffusion-weighted MR imaging. Radiology. 2010;254:154–162. doi: 10.1148/radiol.09090304.

61. Sherman C.B., Behr S., Dodge J.L., Roberts J.P., Yao F.Y., Mehta N. Distinguishing Tumor From Bland Portal Vein Thrombus in Liver Transplant Candidates With Hepatocellular Carcinoma: The A-VENA Criteria. Liver Transplant. 2019;25:207–216. doi: 10.1002/lt.25345.

62. Rose T.L., Kim W.Y. Renal Cell Carcinoma: A Review. JAMA. 2024;332:1001–1010. doi: 10.1001/jama.2024.12848.

63. Raj V., Alpendurada F., Christmas T., Moat N.E., Mohiaddin R.H. Cardiovascular magnetic resonance imaging in assessment of intracaval and intracardiac extension of renal cell carcinoma. J. Thorac. Cardiovasc. Surg. 2012;144:845–851. doi: 10.1016/j.jtcvs.2011.11.035.

64. Expert Panel on Cardiac Imaging. Renapurkar R.D., Rajiah P.S., Bartel T.B., El-Sherief A.H., Francois C.J., Hanneman K., Hsu J.Y., Jackson C.D., Lenge de Rosen V., et al. ACR Appropriateness Criteria® Imaging for Pulmonary Embolism, Known Clot. J. Am. Coll. Radiol. 2025;22:S261–S273. doi: 10.1016/j.jacr.2025.02.028.

65. Fallouh N., McGuirk H.M., Flanders S.A., Chopra V. Peripherally Inserted Central Catheter-associated Deep Vein Thrombosis: A Narrative Review. Am. J. Med. 2015;128:722–738. doi: 10.1016/j.amjmed.2015.01.027.

66. Lopez-Rubio M., Lago-Rodriguez M.O., Ordieres-Ortega L., Oblitas C.M., Moragon-Ledesma S., Alonso-Beato R., Alvarez-Sala-Walther L.A., Galeano-Valle F. A Comprehensive Review of Catheter-Related Thrombosis. J. Clin. Med. 2024;13:7818. doi: 10.3390/jcm13247818.

67. Kucher N. Deep-vein thrombosis of the upper extremities. N. Engl. J. Med. 2011;364:861–869. doi: 10.1056/NEJMcp1008740.

68. Streiff M.B., Holmstrom B., Angelini D., Ashrani A., Buckner T., Diep R., Fertrin K.Y., Fogerty A.E., Crestani N.G., Gangaraju R., et al. Cancer-Associated Venous Thromboembolic Disease, Version 2.2024, NCCN Clinical Practice Guidelines in Oncology. J. Natl. Compr. Cancer Netw. 2024;22:483–506. doi: 10.6004/jnccn.2024.0046.

69. Baskin J.L., Pui C.H., Reiss U., Wilimas J.A., Metzger M.L., Ribeiro R.C., Howard S.C. Management of occlusion and thrombosis associated with long-term indwelling central venous catheters. Lancet. 2009;374:159–169. doi: 10.1016/S0140-6736(09)60220-8.

70. Farge D., Frere C., Connors J.M., Khorana A.A., Kakkar A., Ay C., Munoz A., Brenner B., Prata P.H., Brilhante D., et al. 2022 international clinical practice guidelines for the treatment and prophylaxis of venous thromboembolism in patients with cancer, including patients with COVID-19. Lancet Oncol. 2022;23:e334–e347. doi: 10.1016/S1470-2045(22)00160-7.

71. Rajasekhar A., Streiff M.B. How I treat central venous access device-related upper extremity deep vein thrombosis. Blood. 2017;129:2727–2736. doi: 10.1182/blood-2016-08-693671.

72. Dongmo G., Pal K., Sheth R.A. Endovascular Management of Deep Venous Thrombosis in Cancer Patients. Semin. Interv. Radiol. 2025;42:533–538. doi: 10.1055/s-0045-1812083.

73. Pandhi M.B., Desai K.R., Ryu R.K., Lewandowski R.J. The Role of Inferior Vena Cava Filters in Cancer Patients. Semin. Interv. Radiol. 2016;33:71–74. doi: 10.1055/s-0036-1581090.

74. Goldhaber S.Z., Magnuson E.A., Chinnakondepalli K.M., Cohen D.J., Vedantham S. Catheter-directed thrombolysis for deep vein thrombosis: 2021 update. Vasc. Med. 2021;26:662–669. doi: 10.1177/1358863X211042930.

75. Vedantham S., Goldhaber S.Z., Julian J.A., Kahn S.R., Jaff M.R., Cohen D.J., Magnuson E., Razavi M.K., Comerota A.J., Gornik H.L., et al. Pharmacomechanical Catheter-Directed Thrombolysis for Deep-Vein Thrombosis. N. Engl. J. Med. 2017;377:2240–2252. doi: 10.1056/NEJMoa1615066.

76. Patel R.M., Pal K., Ahmed S.H., Kuban J.D., Patel M., Shah K., Habibollahi P., Metwalli Z., Gurusamy V., Gupta S., et al. Mechanical Venous Thrombectomy for Deep Venous Thrombosis in Cancer Patients: A Single-Center Retrospective Study. Cardiovasc. Interv. Radiol. 2024;47:556–566. doi: 10.1007/s00270-024-03691-3.

77. Jaber W.A., Gonsalves C.F., Stortecky S., Horr S., Pappas O., Gandhi R.T., Pereira K., Giri J., Khandhar S.J., Ammar K.A., et al. Large-Bore Mechanical Thrombectomy Versus Catheter-Directed Thrombolysis in the Management of Intermediate-Risk Pulmonary Embolism: Primary Results of the PEERLESS Randomized Controlled Trial. Circulation. 2025;151:260–273. doi: 10.1161/CIRCULATIONAHA.124.072364.

78. Wiklund P., Medson K., Elf J. Incidental pulmonary embolism in patients with cancer: Prevalence, underdiagnosis and evaluation of an AI algorithm for automatic detection of pulmonary embolism. Eur. Radiol. 2023;33:1185–1193. doi: 10.1007/s00330-022-09071-0.

79. Graeve V.I.J., Laures S., Spirig A., Zaytoun H., Gregoriano C., Schuetz P., Burn F., Schindera S., Schnitzler T. Implementation of an AI Algorithm in Clinical Practice to Reduce Missed Incidental Pulmonary Embolisms on Chest CT and Its Impact on Short-Term Survival. Investig. Radiol. 2025;60:260–266. doi: 10.1097/RLI.0000000000001122.

80. Savage C.H., Elkassem A.A., Hamki O., Sturdivant A., Benson D., Grumley S., Tzabari J., Junck K., Li Y., Li M., et al. Prospective Evaluation of Artificial Intelligence Triage of Incidental Pulmonary Emboli on Contrast-Enhanced CT Examinations of the Chest or Abdomen. AJR Am. J. Roentgenol. 2024;223:e2431067. doi: 10.2214/AJR.24.31067.

81. Soffer S., Klang E., Shimon O., Barash Y., Cahan N., Greenspana H., Konen E. Deep learning for pulmonary embolism detection on computed tomography pulmonary angiogram: A systematic review and meta-analysis. Sci. Rep. 2021;11:15814. doi: 10.1038/s41598-021-95249-3.

82. Rothenberg S.A., Savage C.H., Abou Elkassem A., Singh S., Abozeed M., Hamki O., Junck K., Tridandapani S., Li M., Li Y., et al. Prospective Evaluation of AI Triage of Pulmonary Emboli on CT Pulmonary Angiograms. Radiology. 2023;309:e230702. doi: 10.1148/radiol.230702.

83. El-Sherbini A.H., Coroneos S., Zidan A., Othman M. Machine Learning as a Diagnostic and Prognostic Tool for Predicting Thrombosis in Cancer Patients: A Systematic Review. Semin. Thromb. Hemost. 2024;50:809–816. doi: 10.1055/s-0044-1785482.

84. Chen K., Zhang Y., Zhang L., Zhang W., Chen Y. Machine learning models for risk prediction of cancer-associated thrombosis: A systematic review and meta-analysis. J. Thromb. Haemost. 2025;23:610–626. doi: 10.1016/j.jtha.2024.11.001.

85. Jee J., Brannon A.R., Singh R., Derkach A., Fong C., Lee A., Gray L., Pichotta K., Luthra A., Diosdado M., et al. DNA liquid biopsy-based prediction of cancer-associated venous thromboembolism. Nat. Med. 2024;30:2499–2507. doi: 10.1038/s41591-024-03195-0.

86. Zhang L., Liu M., Wang L., Zhang Y., Xu X., Pan Z., Feng Y., Zhao J., Zhang L., Yao G., et al. Constructing a Large Language Model to Generate Impressions from Findings in Radiology Reports. Radiology. 2024;312:e240885. doi: 10.1148/radiol.240885.

87. Huang J., Wittbrodt M.T., Teague C.N., Karl E., Galal G., Thompson M., Chapa A., Chiu M.L., Herynk B., Linchangco R., et al. Efficiency and Quality of Generative AI-Assisted Radiograph Reporting. JAMA Netw. Open. 2025;8:e2513921. doi: 10.1001/jamanetworkopen.2025.13921.

88. Chen D., Chauhan K., Parsa R., Liu Z.A., Liu F.F., Mak E., Eng L., Hannon B.L., Croke J., Hope A., et al. Patient perceptions of empathy in physician and artificial intelligence chatbot responses to patient questions about cancer. npj Digit. Med. 2025;8:275. doi: 10.1038/s41746-025-01671-6.

89. Montasser M., Masoud H., Hamimi A. Beyond the Algorithm: Quick review of Real-World Hurdles for Radiology AI. Cancer Ther. Oncol. Int. J. 2025;30:1–3.

90. Salehi S., Keishing V., Singh Y., Wei D., Khosravi A., Habibi P., Jagtap J., Erickson B.J. Systematic Review: Agentic AI in Neuroradiology: Technical Promise with Limited Clinical Evidence. J. Imaging Inform. Med. 2026 doi: 10.1007/s10278-025-01839-2. ahead of print .

91. Schlemmer H.P. Navigating the AI revolution: Will radiology sink or soar? Jpn. J. Radiol. 2025;43:1628–1633. doi: 10.1007/s11604-025-01810-9.

92. Diaz-Lorenzo I., Alonso-Burgos A., Friera Reyes A., Pacios Blanco R.E., de Benavides Bernaldo de Quiros M.D.C., Gallardo Madueno G. Current Role of CT Pulmonary Angiography in Pulmonary Embolism: A State-of-the-Art Review. J. Imaging. 2024;10:323. doi: 10.3390/jimaging10120323.

93. Stogiannos N., Cuocolo R., Akinci D’Antonoli T., Pinto Dos Santos D., Harvey H., Huisman M., Kocak B., Kotter E., Lekadir K., Shelmerdine S.C., et al. Recognising errors in AI implementation in radiology: A narrative review. Eur. J. Radiol. 2025;191:112311. doi: 10.1016/j.ejrad.2025.112311.

94. McKinney S.M., Sieniek M., Godbole V., Godwin J., Antropova N., Ashrafian H., Back T., Chesus M., Corrado G.S., Darzi A., et al. International evaluation of an AI system for breast cancer screening. Nature. 2020;577:89–94. doi: 10.1038/s41586-019-1799-6.

95. Liang Y., Chao H., Zhang J., Wang G., Yan P. Unbiasing Fairness Evaluation of Radiology AI Model. Meta Radiol. 2024;2:100084. doi: 10.1016/j.metrad.2024.100084.

96. Lee R.W., Jun T.J., Lee J.M., Cho S.I., Park H.J., Suh J. Vulnerability of Large Language Models to Prompt Injection When Providing Medical Advice. JAMA Netw. Open. 2025;8:e2549963. doi: 10.1001/jamanetworkopen.2025.49963.

97. Goisauf M., Cano Abadia M. Ethics of AI in Radiology: A Review of Ethical and Societal Implications. Front. Big Data. 2022;5:850383. doi: 10.3389/fdata.2022.850383.

98. Santra S., Kukreja P., Saxena K., Gandhi S., Singh O.V. Navigating regulatory and policy challenges for AI enabled combination devices. Front. Med. Technol. 2024;6:1473350. doi: 10.3389/fmedt.2024.1473350.

Associated Data

Data Availability Statement

No new data were created or analyzed in this review. All referenced data sources are publicly available in the peer-reviewed literature as cited in the reference list.