Modern multimodality diagnosis of portal vein infiltration in hepatocellular carcinoma and expected changes during current therapies

Introduction

Hepatocellular carcinoma (HCC) represents the fifth most common cancer worldwide and the second most common cause of cancer-related mortality making out 90% of primary malignant neoplasm of the liver (1–3). Tumor invasion of the portal vein as well as benign portal vein thrombosis (PVT) are both associated with HCC occurring in the setting of liver cirrhosis and portal hypertension with a prevalence of 44% and 42%, respectively (4,5). Moreover, these complications do often co-exist. Differentiation between malignant infiltration of the portal vein and PVT has great impact on prognosis of HCC according to the Barcelona Clinic Liver Cancer (BCLC) staging which is significantly poorer (advanced stage-C) in patients with malignant PVT with a mean survival time of 2.7–4.0 months (6,7). Venous invasion is associated with higher grade and larger tumors and represents an independent predictor of survival (6). Moreover, malignant infiltration of the portal vein in HCC is usually an exclusion criterion for aggressive treatments like transarterial chemoembolization (TACE) and liver surgery or orthotopic liver transplantation due to high recurrence and complication rates (5). In this clinical setting, recent data showed that TACE can be safely performed in advanced HCC patients with vascular invasion improving overall survival (8,9). Even repeated TACE sessions seem to have no negative impact on the liver function (8). Positive results have also been reported for other local therapy techniques like radiofrequency ablation (RFA), transarterial radioembolization (TARE), hepatic artery infusional chemotherapy with 5-fluorouracil, TACE and portal vein and stent combined with ¹²⁵I seed, three-dimensional (3D) conformal radiotherapy combined with TACE, as well as surgical resection and systemic treatment with sorafenib combined with TACE (9–16).

Two different pathways of malignant portal vein infiltration are recognized: tumor invasion into the portal vein by direct venous extension or by hematogenous seeding. In the latter, there is no apparent anatomic relation between the intraluminal portal vein mass and the primary hepatic tumor making diagnosis particularly challenging. Malignant infiltrates of the portal vein have their own arterial supply stemming from the vasa vasorum of the portal vein (17). The gold standard for characterization of a portal venous infiltration is histopathological examination. However, due to the invasive nature of biopsy and associated bleeding risk in liver cirrhosis, differentiation in clinical routine is usually based on imaging findings (18). Furthermore, imaging is crucial not only for diagnosis, but also for evaluation of response to therapy. According to Response Evaluation Criteria in Solid Tumors (RECIST) which are based on size measurements of tumor target lesions, even focal malignant infiltration of the portal vein is considered non-target. This is due to difficulties in determining its nature, and the poor measurability because of its amorphous configuration (19). New therapeutic agents exhibiting antiangiogenic properties often show little influence on tumor size, but rather affect other tumor parameters like vascularization or water diffusivity. Hence, the combined use of morphologic and functional imaging techniques becomes imperative. The most frequently used diagnostic imaging tools are ultrasound (US) with Doppler mode, contrast-enhanced ultrasound (CEUS), contrast-enhanced computed tomography (CECT), contrast-enhanced magnetic resonance imaging (MRI), and fluorodeoxyglucose positron emission tomography (FDG-PET).

Ultrasound, color Doppler, power Doppler, and contrast-enhanced ultrasound

The echogenicity of a thrombus is a poor discriminator between malignant infiltration of the portal vein and bland PVT (Figs. 1–3). Thrombus echogenicity is increased in both disorders, comparable to primary hepatic tumors or adjacent parenchyma, but variable echogenicity can also be encountered (Figs. 1a, 2a, and 3a). With time, echogenicity decreases in bland PVT paralleling volume loss, contrary to malignant infiltration of the portal vein which is expected to farther expand and demonstrate neovascularization if treated inefficiently (20). Flow visualization of the portal vein using CD and PD may confirm lack of patency with either missing flow signal or presence of pulsatile flow due to a-v shunts between hepatic artery (HA) and portal vein (VP), as well as collateralization. Collateralization usually recruits and dilates the preformed peribiliary venous plexus (Figs. 2b and 3b). Moreover, some HCC-induced infiltrates of the portal vein are highly vascularized and therefore simulate normal flow in the portal vein on CD or PD if no spectral analysis is performed. Similarly, post-thrombotic syndromes showing partial recanalization of the vessel lumen sometimes mimic either marked vascularization of intravascular tumor or normal vessel patency. In such cases, the use of pulsed Doppler with spectral analysis can disclose arterial flow characteristics of the tumor infiltrating the portal vein. With the advent of second generation US contrast agents, the clinical practice of HCC evaluation has been revolutionized and reports attest higher sensitivity and accuracy for the diagnosis of hepatic malignancies with CEUS compared to standardized CECT (Fig. 1c and d) (21). Early enhancement of a mass within the portal vein lumen on CEUS is an indicator of malignant infiltration (Fig. 2c and d) (20). In conformity with established criteria used for non-invasive characterization of HCC through other imaging techniques, the presence of contrast washout of the thrombus in the venous phase on CEUS is suggestive for malignant PVT (22). Fresh bland PVT shows no intraluminal flow on CEUS (Fig. 3b). In contrast to CT or MRI contrast agents, microbubbles used in CEUS remain in the intravascular space, thus displaying tumor vascularization rather than extravascular diffusion of the contrast agent. For CEUS a specificity of 100% and a sensitivity of 83% are reported for detection and discrimination of malignant infiltration of the portal vein. The lower sensitivity of CEUS due to limitations like patient’s body habitus and operator experience may limit its use in some cases. During antiangiogenic therapy, CEUS reveals decreasing vascularization and potential recanalization of the portal vein, paralleling changes in the primary tumor (Fig. 1e). If the tumor completely regresses, monophasic venous flow can be detected on CD reflecting partially recovered portal vein patency. OPEN IN VIEWER

OPEN IN VIEWER

Fig. 2.

A 76-year-old male patient with known HCC and segmental distension of portal vein lumen of unknown origin. On B-mode US image (a), papillary projections into the portal vein could be detected (arrows). Corresponding color Doppler (CD) demonstrated intravascular flow through the patent vessel lumen (arrowhead) but, no vascularization (arrow) inside the suspected mural nodules (b). After intravenous administration of US contrast agent, marked, early (arterial) accumulation of microbubbles is seen in the suspected area demonstrating tumor invasion of the portal vein (arrow) (c). Note similar contralateral tumor infiltration (arrowhead). Fifteen seconds later, intraluminal flow is seen (long slim arrow) through the patent portal vein in addition to the arterial flow in the mural tumor infiltrate (arrow and contralateral arrowhead) (d).

OPEN IN VIEWER

Fig. 3.

A 68-year-old female patient with HCC and PVT of unknown origin. Marked luminal expansion and increased intraluminal echogenicity are seen on B-mode US (a); but they are both unspecific. With CEUS, lack of intraluminal flow (arrow) as well as incipient collateralization of the venous peribiliary plexus (arrowhead) are noticed (b). At follow-up, spontaneous shrinkage of the bland thrombus (arrow) was documented confirming the diagnosis of benign PVT (c).

Computed tomography, volume perfusion CT, and PET-CT

In unenhanced CT, malignant and fresh benign intravascular portal vein thrombosis produces a hypo- to isoattenuating filling defect expanding the vascular lumen (Figs. 4–6). Most HCCs are slightly hypoattenuating on precontrast images, except for rare spontaneous hemorrhage which may cause hyperattenuation. On CECT, enhancement of an intraluminal mass usually discriminates between malignant infiltration of the portal vein and bland PVT. Rarely, enhancement may be encountered even in bland PVT reflecting collateral channels in or along a blood clot or slow flow in an organized thrombus with intrinsic vascularization, challenging the radiologist (Figs. 4a, 5a and 6a) (23). Moreover, subtle changes in CT attenuation of malignant infiltration of the portal vein reflecting neovascularization can be missed by double-phase CT and differentiation from partial volume averaging due to enhancing adjacent liver parenchyma, or partially patent portal vein lumen, may hamper such measurements. Multi-phase CT with image subtraction might represent a solution for detection of neovascularization in malignant infiltration of the portal vein, similar to that measured in the primary tumor (HCC). Multi-phase CECT using iodine contrast agent, holds a sensitivity of 86 % and a specificity of 100% for detecting HCC, comparable to that of CEUS (23,24). Current international guidelines define a typical HCC vascular pattern (assessed by non-invasive imaging techniques) as a homogeneous hyperenhancement (wash-in) in the arterial phase followed by washout in the venous or late phase (25). Therefore, multi-phase CECT has gained great popularity on this diagnostic field. Nonetheless, some HCCs do not fulfil these criteria (wash-in/washout). Besides, detection of wash-in is strongly dependent on the chosen delay time, circulation time, viscosity of contrast media, etc. For this purpose, a multi-arterial (volume perfusion CT [VPCT]) protocol is more likely to detect these tumor characteristics. Furthermore, the main challenge in cirrhotic patients is differentiation between regenerative liver nodules and less vascularized HCCs (e.g. well-differentiated subtypes). Whereas regenerative hepatic nodules are supposed to be entirely supplied by the portal vein, HCCs become more and more arterialized with progressive dedifferentiation. OPEN IN VIEWER

OPEN IN VIEWER

Fig. 5.

A 78-year-old male patient with known HCC and newly occurred malignant PVT of the right portal vein branch (arrow) of unknown origin documented on arterial (a) and portal–venous postcontrast phase (b). Note high perfusion (arrows) of occluded portal vein on BF color-coded map (c), as well as some patchy infiltration of the liver parenchyma by diffuse HCC that was not obvious on CECT.

OPEN IN VIEWER

Fig. 6.

A 68-year-old male patient with HCC and PVT of unknown origin. CECT shows slightly hypoattenuating thrombus of the left portal vein branch (arrows) in the arterial (a) and portal–venous (b) postcontrast phases. VPCT shows lack of perfusion on BF (c) and BV (d) color-coded maps (arrows) despite large HCC next to it in the liver segment 3. At follow-up (e), T1W MRI sequence illustrates typical methemoglobin-related hyperintense signal and slight shrinkage of the PVT despite adjacent HCC confirming the diagnosis of benign PVT.

VPCT as a novel technique measures temporal changes in tissue density after intravenous contrast injection using repeated CT scans of the volume being analyzed. Different kinetic modeling techniques are used by software providers. A one-compartment analysis of short-duration scans during the initial phase after contrast injection allows the calculation of perfusion (blood flow [BF]) and time to peak (TTP), whereas a two-compartment (Patlak) analysis quantifies exchanges between intra- and extravascular spaces, enabling calculation of blood volume ([BV], excluding stagnant blood) and permeability (PMB) of vessels. The latter is also known as transit constant ([K_trans], defined as the sum of the flow within the microvasculature and capillary permeability) (26,27). Additionally, separate calculation of arterial and portal–venous liver perfusion (ALP and PVP) by appropriate placement of ROIs in the portal vein and the spleen is possible due to the simultaneous perfusion of both organs via hepatic and splenic arteries. This is analogous to former CT hepatic arteriography (CTHA) and CT arterioportography (CTAP) techniques, but can be performed in a non-invasive way. This separate illustration of hepatic arterial and portal–venous blood supply via VPCT is further increasing sensitivity and specificity disclosing exclusive arterialization of otherwise poorly vascularized tumors. VPCT has become an important biomarker for monitoring treatment response, especially to antiangiogenic drugs (28). Previous studies focused on the role of perfusion CT for discrimination between well-differentiated and moderately or poorly differentiated HCC (29). It was shown that perfusion CT techniques can quantify tumor perfusion parameters like BF, BV, or K_trans (Figs. 4b, c, 5b, 6b, c, and 7a, b) (30,31). Of note, non-invasive diagnostic criteria for HCC are fully transferable to the diagnosis of malignant PVT. Hence, highly arterialized processes infiltrating the portal vein (high hepatic arterial perfusion index) lacking portal–venous perfusion are diagnostic for tumor infiltration. These characteristics allow better detection of additional inconspicuous hepatic HCC sites and improve monitoring response to systemic or local therapy. Here, declining perfusion values reflect either tumor necrosis or a temporary, drug-induced ‘switch off’ in tumor angiogenesis (Figs. 4d, 6c). Sophisticated motion correction algorithms enable accurate recognition of vessel invasion and quantification of perfusion parameters. Therefore, this method should be used either in primary diagnosis of HCC when portal vein invasion is suspected or in equivocal cases (e.g. if CEUS and/or CECT have failed). The only disadvantage of this modality is radiation exposure, which is slightly higher for VPCT. Nevertheless, this aspect pales in comparison to the disease course and its mean survival time, especially in case of palliative treatment. OPEN IN VIEWER

In contrast to the other imaging techniques described showing morphological changes in tissues, 18F-FDG-PET is able to detect metabolic changes in tumors. Few data exist on the use of PET-CT when trying to differentiate malignant infiltration of the portal vein from benign PVT, but the current experiences with FDG-PET for HCC diagnosis argue against its routine application in this clinical setting (32–34).

Magnetic resonance imaging

Malignant portal vein infiltration can exhibit variable signal intensities, but usually presents with hypo- to isointense signal on T1-weighted (T1W) images and slightly hyper- to isointense signal on T2-weighted (T2W) images (Fig. 8 a and b). In contrast, bland PVT often shows low signal intensity on T2W images due to its hemosiderin content. Li et al. reported comparable results of T2*W-weighted (T2*W) imaging as compared to postcontrast studies for characterization of benign PVT (35). However, differentiation of malignant from benign intravascular portal vein processes on the basis of MRI signal intensity alone remains difficult. Similar to CT, MRI diagnosis of HCC is mainly based on the evidence of contrast enhancement in tumor (36). On postcontrast studies of malignant infiltration of the portal vein, an early increase in signal intensity (arterialization) on fat-saturated T1W scans followed by progressive washout on later dynamic phases is typical (Fig. 8c). Beside gadopentat dimeglumin (GD-DPTA), hepatocellular contrast agents (e.g. gadoxetate disodium) have gained in popularity for the diagnosis of HCC and related pathologies (e.g. malignant infiltration of the portal vein) due to a better tumor to tissue contrast (36). Their use enhances the sensitivity to non-invasively diagnose small hepatocellular carcinoma nodules in cirrhotic patients. Double hypointensity in the portal/venous and hepatobiliary phases has been proposed as a new MR pattern, highly suggestive of hypovascular hepatocellular carcinoma (37). By some authors the signal intensity increase in the dynamic postcontrast T1W images is still considered superior for extracellular contrast media when compared to hepatocellular-specific agents. In equivocal cases, image subtraction of the contrast-enhanced dynamic series from the unenhanced scan has proved beneficial, as long as not limited by motion or breathing artifacts (Fig. 8d). Sandrasegaran et al. showed that the presence of at least two of the three following MRI findings had a sensitivity of 100% and a specificity of 90% for the diagnosis of malignant PVT: distance from tumor to PVT of less than 2 cm; HCC size of greater than 5 cm; and arterial enhancement of PVT (38). Similar to intrahepatic HCC, intravascular tumor infiltration may often but not always present restriction of water diffusivity with correspondingly low apparent diffusion coefficient (ADC) which may aid in the discrimination of benign from malignant PVT (18) (Fig. 8e). Successful therapy increases water diffusivity and results in a higher ADC value, which reflects derangement of macromolecular tumor architecture (e.g. due to necrosis). Interestingly, Maksimovic et al. demonstrated that early changes in ADC values in HCC patients treated with sorafenib show a contrary trend (39). Hence, acute PVT may also exhibit a decrease in ADC caused by shrinkage of the extracellular space secondary to resorption of plasma and changes in the conformation of the hemoglobin molecule. Other studies failed to demonstrate a benefit when using DWI as a discriminator between benign and malignant PVT (38). OPEN IN VIEWER

In any case, the application of DWI for response monitoring of HCC is limited to the subgroup of tumors presenting with restricted water diffusion in the baseline setting. During local or systemic therapy, signal intensity changes such as an increase on both T1W and T2W images as well as a decrease in contrast enhancement, reflect either hemorrhage or necrosis and need to be considered for correct evaluation of response to treatment (Fig. 8f–i) (40). The use of higher magnetic fields (7/9.4 Tesla) and the implementation of susceptibility-weighted imaging (SWI) are expected to further increase the sensitivity of MRI in this respect (41,42).

Conclusion

Accurate diagnosis of benign or malignant portal vein infiltration is accomplished by most contrast-enhanced imaging techniques. However, cognizance of their strengths and limitations is mandatory. CEUS should represent the method of choice in the primary diagnosis of PVT in compliant patients. VPCT is less affected by motion artifacts and (just like dynamic contrast-enhanced MRI) yields perfusion parameters that can be quantified and are therefore more suited for characterization and response monitoring studies.