Noninvasive Imaging Assessment of Portal Hypertension

Abstract

Portal hypertension (PH) is a spectrum of complications of chronic liver disease (CLD) and cirrhosis, with manifestations including ascites, gastroesophageal varices, splenomegaly, hypersplenism, hepatic hydrothorax, hepatorenal syndrome, hepatopulmonary syndrome and portopulmonary hypertension. PH can vary in severity and is diagnosed via invasive hepatic venous pressure gradient measurement (HVPG), which is considered the reference standard. Accurate diagnosis of PH and assessment of severity are highly relevant as patients with clinically significant portal hypertension (CSPH) are at higher risk for developing acute variceal bleeding and mortality. In this review we discuss current and upcoming noninvasive imaging methods for diagnosis and assessment of severity of PH.

Introduction/Background

Chronic liver disease (CLD) results from hepatocyte damage due to a variety of insults and causes changes in liver function, and if left untreated, may result in liver fibrosis and cirrhosis. CLD and cirrhosis account for a significant burden of disease in the United States, of which the epidemiology is highly variable, unevenly distributed, and rapidly evolving. Typical etiologies of CLD include viral hepatitis (HBV and HCV), alcohol misuse, non-alcoholic fatty liver disease (NAFLD)/nonalcoholic steatohepatitis (NASH), primary biliary cholangitis (PBC), primary sclerosing cholangitis (PSC), autoimmune hepatitis, drug/toxin-induced liver disease and hereditary/metabolic disorders. The estimated prevalence of cirrhosis in the United States ranges from 300-1000 per 100,000 patients [1–3]. Currently, the anticipated increase in CLD disease burden is driven by the NAFLD epidemic and a rise in liver disease arising from alcohol misuse [4,5]. Conversely, there has been a substantial decrease in the burden of viral liver disease, namely chronic hepatitis C infection (HCV), due to the success of direct acting antiviral (DAA) therapy [6] and chronic hepatitis B (HBV) due to successful vaccination campaigns and antiviral therapy [7–9]. CLD and cirrhosis are the 8^th leading cause of death in the United States, resulting in at least 34,000 deaths annually and there has been a recent 65% increase in death rate from 1996 to 2006, estimates of which are most likely conservative [10–12]. Furthermore, deaths due to cirrhosis are anticipated to triple by 2030, due to the increasing prevalence of NALFD [4].

The natural history of CLD is variable and can worsen silently, until progressive liver injury and fibrosis result in the development of complications such as portal hypertension (PH) and hepatic insufficiency, both of which can ultimately lead to hepatic decompensation and death. The vast majority of PH is attributed to CLD and cirrhosis, while the less common etiologies include vascular diseases such as extrahepatic vascular occlusion [segmental extrahepatic portal vein (PV) occlusion] or noncirrhotic PH [13]. The development of PH sets the stage for future events, including the development of variceal hemorrhage, ascites, hepatic encephalopathy and jaundice, all of which contribute significantly to liver related morbidity, mortality, diminished quality of life and substantial healthcare costs [14–16]. PH has been reported to develop at a rate of 5-7% annually [17]. Approximately 50-70% of patients with compensated cirrhosis demonstrate clinically significant portal hypertension (CSPH), which is defined as an increase in the hepatic venous pressure gradient (HVPG) to ≥ 10 mmHg and is an established independent predictor of clinical decompensation [18,19]. In patients with decompensated cirrhosis, life expectancy as low as 2 years and a 1-year mortality rate of 20.2% have been reported [20,21].

Pathophysiologically, architectural distortion caused by fibrosis, microvascular thrombosis, hepatic sinusoidal endothelial cell dysfunction and hepatic stellate cell activation in the liver lead to increased vascular resistance at the sinusoidal level resulting in PH development [22]. The increased vascular resistance in the liver results in downstream hemodynamic effects, such as splanchnic and systemic arterial vasodilatation, hyperdynamic circulation and development of portosystemic collaterals. These adaptive changes can result in further increases in portal pressures [23]. The complications of PH are potentially devastating and include hemorrhage from varices, hepatic encephalopathy from portosystemic shunts, ascites (and potential hepatic hydrothorax [24]), and renal failure due to hepatorenal syndrome [22]. Hepatopulmonary syndrome, a form of hypoxemia, is found in patients with cirrhosis and PH due to abnormal intrapulmonary vascular dilation [25], with poor outcome reported [26]. Portopulmonary hypertension, a form of pulmonary arterial hypertension, is also potentially present in patients with PH, however the degree of PH appears unrelated to the risk of development [27]. Hypersplenism is found in patients with splenomegaly, a common occurrence in PH [28].

A major challenge for patients with CLD and PH remains the ability to assess disease severity, risk of decompensation events and predict outcome. A variety of clinical measures and invasive tools have been developed. These include clinical and laboratory scoring systems that take into account either subjective variables, such as the Child-Pugh (CP) score, or numerical data, such as the Model for End-stage Liver Disease (MELD), which have been validated for cirrhosis. Invasive assessments, such as upper endoscopy, which allows direct visualization and possible management of portosystemic varices, and HVPG testing (indirectly via transjugular access) are also utilized for the assessment of PH, but are costly, require expertise, and are associated with potential risks and complications [29,30]. More recently, non-invasive qualitative and quantitative imaging methods predominantly using ultrasound (US), computed tomography (CT) and magnetic resonance imaging (MRI) have been investigated for this purpose, with promising results.

A conundrum in the clinical management of PH is the difficulty in differentiating advanced fibrosis from cirrhosis. Liver biopsy is the gold standard for cirrhosis diagnosis however well described limitations such as sampling error, inter-observer variability and risk of complications limit its utility. The diagnosis of cirrhosis has traditionally been a key marker in the clinical management of patients with CLD as it prompts screening for hepatocellular carcinoma (HCC) through imaging, esophageal varices through endoscopy and PH through HVPG and biopsy when available. The advancement of non-invasive methods of cirrhosis diagnosis, such as elastography, have led to cirrhosis being diagnosed at earlier stages and therefore rendering screening studies, particularly for esophageal varices, unnecessary. The Baveno VI consensus guidelines recognize this shortcoming in clinical protocol and the often-indistinguishable continuum between advanced liver fibrosis and cirrhosis without histological confirmation [31]. A new term has been proposed “compensated advanced chronic liver disease” (cACLD), which can be identified using non-invasive methods, specifically liver stiffness measurement (LS) obtained with transient elastography (TE).

Given the high prevalence, significant burden of disease, potentially devastating complications and poor outcome of PH, the ability to detect, quantify, and assess response to PH therapy and prognosis would be a major benefit to this patient population. In this review, we will review the imaging tools available for the non-invasive assessment of PH, with emphasis on quantitative MRI methods.

Reference Standard for Diagnosing PH

PH becomes a clinical concern upon diagnosis of liver cirrhosis, which could be compensated and asymptomatic. The reference standard for diagnosis of liver cirrhosis is still liver biopsy, however the emergence of non-invasive techniques such as elastography, with a high sensitivity for cirrhosis prediction [32], has significantly reduced the number of liver biopsies being performed. The Baveno VI consensus guidelines [31] recommend HVPG measurement as the reference standard to determine the presence of PH, and upper endoscopy as the reference standard for identifying the presence and grade of esophageal varices.

HVPG is a vital tool in diagnosis and prognosis of PH [33–35,19], with HVPG > 5 mmHg confirming PH and HVPG ≥ 10 mmHg indicating clinically significant (CS)PH and HVPG ≥12 mmHg considered severe portal hypertension. Patients with CSPH are at increased risk of developing esophageal varices and hepatic decompensation [19]. The HVPG procedure involves the insertion of a catheter via the jugular vein into a hepatic vein under fluoroscopy and subsequent measurement of free and wedged pressure, with HVPG being the difference between wedged and free hepatic vein pressure. HVPG measurements are typically repeated three times and an average HVPG recorded. In addition to serving as a diagnostic tool for PH, HVPG is also used as a marker of therapy response, with reduction in HVPG of 20% considered a positive response and associated with improved outcomes [36]. Liver biopsy can be performed safely at the same time and carries a lower bleeding risk as compared to a percutaneous approach [37]. In addition to confirming the presence of PH, HVPG may also be predictive of HCC development [38], hepatic decompensation [19], mortality [35] and is a key indicator of varices formation [18]. Finally, HVPG may be used to monitor response to transjugular intrahepatic portosystemic shunt (TIPS) placement [39]. Although an invasive measurement, complications of HVPG measurement such as transient cardiac arrythmia and local pain are rare [30]. A significant drawback however is the need for technical expertise and facilities typically only found in tertiary referral centers.

Upper endoscopy is the reference standard for establishing the presence and severity of esophageal varices, a serious complication of PH. The mortality rate of acute variceal bleeding has been reduced however it remains high [40]. Complications during upper endoscopy exams are infrequent and estimated risk does not exceed 1/500 per examination, with the risk of death at 1/10,000 [29]. Varices are graded based on size, with the American Association for the Study of Liver Disease (AASLD) suggesting varices be graded as either small or large, with 5mm diameter considered the threshold [41]. An earlier grading system from the Japanese Research Society for Portal Hypertension suggested varices be graded in 3 sizes; small, medium and large, however the description does not include threshold diameter measurements [42]. A recent study found no difference in observer agreement between the 2 grade and 3 grade systems [43].

Both HVPG measurement and endoscopy are key prognostic components in PH, however they are both relatively invasive, expensive and not without risk. With improvements in diagnostic performance in detecting cirrhosis, cirrhotic patients tend to be identified at an earlier stage and therefore often undergo unnecessary screening endoscopies. The newly proposed term cACLD seeks to reduce unnecessary screening endoscopies by stratifying patients based on non-invasive tests, specifically liver stiffness measurement (LS) obtained via transient elastography (TE) and platelet count. Non-invasive methods of PH assessment are discussed below.

Qualitative non-invasive imaging methods

There are several anatomic findings that occur in patients with PH that can be demonstrated by conventional imaging methods such as US, CT and MRI, each of which has advantages and disadvantages. Each modality offers the benefit of a comprehensive evaluation of the liver, the stigmata of PH, and portal vasculature in a single exam. Typical anatomic findings of PH manifestations include portosystemic varices, splenomegaly, ascites and abnormalities of the PV, which can be delineated on each of these modalities. Many of the imaging findings of PH are specific, however are limited by reduced sensitivity, especially in the setting of compensated liver disease. The presence of these findings enables confirmation of PH, but the absence of which does not exclude PH [44]. Other manifestations of PH, such as hepatic encephalopathy, hepatorenal syndrome and hepatopulmonary syndrome are typically diagnosed clinically.

Portosystemic varices:

The presence of portosystemic collateral vessels is pathognomonic of PH, are often numerous, widespread, variable and typically occur in the following anatomic sites: gastroesophageal, perigastric, perisplenic, gastrohepatic ligament, inferior mesenteric, periumbilical, omental, left gastric vein and splenorenal locations [45]. Recanalization of the paraumbilical vein can occur. Contrast-enhanced (CE)-CT and MRI allows for comprehensive mapping of portosystemic collateral vasculature is highly sensitive for the detection of varices. Accurate detection of esophageal and gastric varices is highly clinically relevant given their propensity to bleed, with CT and MRI studies demonstrating sensitivity of 58-96% and specificity of 45-82%, with lower sensitivity reported for varices < 3-5 mm in size [46–51].

Splenomegaly:

An enlarged spleen with a craniocaudal length >13 cm is a sensitive, yet non-specific finding in PH [44]. A specific MRI finding of PH is the presence of splenic Gamna-Gandy bodies, which is best delineated on MRI and appear as T1-W hypointense foci that demonstrate susceptibility artifact, and correspond to organizing hemorrhage [52]. A prior study found increasing spleen size is an independent predictor of gastroesophageal varices in compensated cirrhosis [53].

Ascites:

The presence of free fluid in the abdomen is readily identified at imaging, and typically appears as anechoic at US, fluid attenuation [<20 Hounsfield units (HU)] at CT and T2-W hyperintense/T1-W hypointense at MRI. Ascites can vary in severity and distribution as minimal perihepatic and perisplenic fluid, intraperitoneal fluid without abdominal wall distension, and intraperitoneal fluid causing significant abdominal wall distension [54].

PV abnormalities:

The presence of PV thrombosis/occlusion aids in the differential diagnosis of PH and is readily assessed with conventional imaging. Findings of PH on gray-scale and Doppler US include enlargement of the main PV diameter >13 mm, which is associated with risk of large varices, and reduced flow < 15 cm/s, which has shown sensitivity and specificity of 88% and 96%, respectively for the diagnosis of PH [55,44]. The finding of reversal of flow direction in the portal venous system is diagnostic for PH, and is best assessed by Doppler US [44].

Qualitative findings in combination:

These findings can be used in combination to assess the presence and severity of PH. A previous study investigated a scoring system integrating the number of variceal sites, severity of ascites, and craniocaudal spleen length, and found that there was high diagnostic performance for the diagnosis and severity of PH as defined by HVPG (AUCs 0.78–0.76 and 0.83–0.81 for ≥ HVPG 5 mmHg and HVPG ≥ 10 mmHg for 2 observers) and high inter-observer agreement (k=0.71) [54]. Other studies have assessed the following features using either multivariable or diagnostic modeling analyses: PV size, grade of varices, severity of ascites, and liver-to-spleen volume ratio [56,57].

Serum biomarkers

While there are several serum markers and scoring systems that have shown diagnostic value for cirrhosis, only a few have shown promise for PH. Thrombocytopenia is the most frequently reported laboratory abnormality, but platelet count is not reduced in the setting of all patients with CSPH and does not reflect the severity of PH. There is some evidence showing potential value for the Lok score (comprised of AST, ALT, platelets and INR), FIB-4 score (comprised of age, platelets, AST and ALT) and AST-to-platelet ratio index (APRI) for the diagnosis of PH [58–61]. The Lok and FIB-4 scores have been used for risk stratification in CLD to identify patients that have very low risk of varices requiring treatment and who may therefore safely avoid endoscopy [31]. Other serum markers including osteopontin, von Willebrand factor and VITRO score (Von Willebrand Factor Antigen/Thrombocyte Ratio) have also been studied [62] and found to have value in prognosticating hepatic decompensation and mortality [63,64].

Non-invasive Quantitative Imaging Methods

While many of the manifestations of PH can be readily identified at conventional imaging, they do not provide accurate measure of degree of severity of PH nor prognostic information. There is significant interest in the development and validation of quantitative imaging tools that can provide this information (Figure 1). In the setting of CLD, pathological changes in the liver, such as increased deposition of collagen, stiffen the liver parenchyma as the disease progresses. This stiffening of the liver also increases intrahepatic resistance which is a main component of increased portal pressure. The increasing stiffness of the liver provides a rich contrast for elastography techniques which can non-invasively estimate tissue stiffness. The hemodynamic and perfusion changes associated with cirrhosis and PH may also be quantified non-invasively through techniques such as 2D phase contrast MRI (2D PC-MRI), 4D flow and dynamic contrast-enhanced (DCE)-MRI. Changes in tissue composition as a result of CLD and PH may also be detected using MRI relaxometry approaches. These methods will be expanded upon in the following section.

Figure 1.

47-year-old female with autoimmune cholangiopathy and mild fibrosis (F1 at biopsy) without portal hypertension [Hepatic venous pressure gradient (HVPG)=1 mmHg] (top), and 22-year-old male with primary sclerosing cholangitis and cirrhosis with clinically significant portal hypertension (CSPH) (HVPG=10 mmHg) (bottom).

Magnetic resonance elastography (MRE) using 2D EPI sequence demonstrates markedly elevated liver and spleen stiffness in CSPH patient. T₁ relaxometry using Look-Locker technique with 32 inversion times shows higher spleen T₁ in CSPH patient, while liver T₁ is not different between patients. T_1ρ is markedly/mildly elevated in the spleen/liver, respectively in CSPH patient. Splenomegaly is visible on T2-weighted images in CSPH patient.

Elastography

Many disease states and pathologies are associated with a change in tissue mechanical properties e.g. liver fibrosis. Quantitative elastography methods enable the measurement of tissue mechanical properties through the imaging of shear wave propagation through the tissue of interest . Elastography has become a widely adopted tool for assessing liver fibrosis with excellent diagnostic performance reported over a range of etiologies [32]. Transient elastography (TE) is the most commonly applied method in the clinic owing to its relative ease of use, availability and low cost. Several meta-analyses have highlighted the excellent diagnostic accuracy of TE for diagnosing liver cirrhosis, with AUROC values >0.9 [65–72]. These meta-analyses indicate that TE is better at ruling out rather than ruling in liver cirrhosis with negative predictive value >90%. TE has also been shown to have excellent performance in predicting CSPH, with two meta-analyses reporting AUROCs ≥0.9 [73,74]. A selection of published reports evaluating the ability of TE to predict PH/CSPH are shown in Table 1. The high diagnostic performance of TE in detecting fibrosis/cirrhosis and the widespread use of the method in the clinic has allowed the diagnosis of liver cirrhosis at earlier stages than before, with CSPH and esophageal varices unlikely to be present. To reduce the number of unnecessary screening endoscopies and HVPG procedures, several studies have suggested a threshold of LS <20 kPa (using TE) and platelet count ≥150 x 10³ mm³ could reduce 40-45% of screening endoscopies with a 5% chance of missing varices requiring treatment [75,76]. This recommendation was incorporated in the Baveno VI guidelines [31]. Similarly, patients with LS ≥20-25 kPa can be considered to have CSPH and should undergo a screening endoscopy. Patients with LS <25 kPa with normal platelet count can be considered to not require screening endoscopy. These thresholds might allow 60% of unnecessary HVPG procedures to be avoided. Recently, the Baveno VI criteria have been expanded, with new thresholds of LS <25 kPa and platelet count >110 x 10³ mm³ increasing the number of unnecessary endoscopies avoided whilst maintaining an acceptable number of missed varices needing treatment [77].

Table 1.

Summary of studies in PH, with diagnostic performance against hepatic vein pressure gradient (HVPG) measurement.

Study	Sample size	Etiology	Technique	AUC
Vizzutti et al. [198]	N=61	HCV	TE	CSPH: 0.990
Bureau et al. [199]	N=144	Mixed	TE	CSPH: 0.945
Lemoine et al. [200]	N=92	ETOH/HCV	TE	CSPH: 0.840
Sanchez-Conde et al. [201]	N=38	HCV/HIV	TE	CSPH: 0.800
Colecchia et al. [79]	N=100	HCV	TE	CSPH - LS: 0.920 CSPH - SS: 0.966
Llop et al. [202]	N=79	Mixed	TE	CSPH: 0.840
Reiberger et al. [203]	N=502	Mixed	TE	PH: 0.794 CSPH:0.817
Berzigotti et al. [100]	N=173	Mixed	TE	CSPH: 0.883
Hong et al. [204]	N=59	Mixed	TE	CSPH: 0.851
Schwabl et al. [205]	N=188	Mixed	TE	CSPH: 0.962
Kitson et al. [206]	N=95	Mixed	TE	CSPH: 0.900
Cho et al. [207]	N=219	ETOH	TE	CSPH: 0.850
Zykus et al. [208]	N=107	Mixed	TE	CSPH - LS: 0.949 CSPH - SS: 0.846
Kumar et al. [209]	N=326	Mixed	TE	PH: 0.786 CSPH: 0.740
Salzl et al. [61]	N=88	Mixed	TE	CSPH: 0.870
			pSWE	CSPH: 0.855
Procopet et al. [210]	N=88	Mixed	TE	CSPH LS (compensated): 0.926
			2D-SWE	CSPH LS: 0.858 CSPH SS: 0.725
Elkrief et al. [85]	N=79	Mixed	TE	CSPH LS: 0.780 CSPH SS: 0.630
			2D-SWE	CSPH LS: 0.790 CSPH SS: 0.720
Kim et al. [211]	N=92	Mixed	2D-SWE	CSPH: 0.819
Lee et al. [212]	N=47	Mixed	2D-SWE	CSPH: 0.745
Jansen et al. [213]	N=155 – LS N=112 - SS	Mixed	2D-SWE	PH – LS: 0.890 PH – SS: 0.900 CSPH – LS: 0.860 CSPH – SS: 0.840
Zhu et al. [214]	N=137 - LS N=104 - SS	HBV	2D-SWE	CSPH LS: 0.720 CSPH SS: 0.810
Elkrief et al. [215]	N=191	Mixed	2D-SWE	CSPH LS: 0.800 CSPH SS: 0.610
Attia et al. [216]	N=78	Mixed	pSWE	CSPH LS: 0.929 CSPH SS: 0.968
Takuma et al. [88]	N=60	Mixed	pSWE	CSPH – LS: 0.833 CSPH – SS: 0.943
Ronot et al. [95]	N=36	Mixed	MRE	CSPH – Spleen loss modulus (56Hz): 0.810
Wagner et al. [150]	N=34	Mixed	MRE	PH: 0.809 CSPH: 0.742
			DCE-MRI	PH: 0.760–0.770 CSPH: 0.750–0.760
Gouya et al. [128]	N=69	Mixed	2D PC-MRI	PH: 0.960
Hectors et al. [168]	N=25	Mixed	T1ρ by multiple spin lock times	PH: Spleen T1ρ: 0.820 CSPH: Spleen T1ρ: 0.708

PH was diagnosed invasively as HVPG> 5 mmHg, and clinically significant PH was diagnosed invasively as HVPG ≥ 10 mmHg. Studies were excluded if: 1) No AUC for diagnosing PH or CSPH was reported 2) Sample size was <25 subjects 3) Subjects were post liver transplant.

2D PC-MRI=2D phase contrast MRI, CSPH=clinically significant portal hypertension, DCE-MRI=dynamic contrast enhanced MRI, ETOH=alcohol cirrhosis, HCV=hepatitis C virus, HIV=human immunodeficiency virus, HVPG=hepatic venous pressure gradient, LS=liver stiffness, MRE=magnetic resonance elastography, PH=portal hypertension, pSWE=point shear wave elastography, SS=spleen stiffness, 2D-SWE=2D shear wave elastography, TE=transient elastography.

The performance of TE in predicting the presence of esophageal varices is less well established, with two meta-analyses [73,78] reporting AUC values of 0.82-0.84. In one meta-analysis [78], subgroup analysis revealed cutoff values for the diagnosis of esophageal varices are dependent on etiology of disease and so larger etiology specific studies are required to establish etiology specific cutoffs for TE. Spleen stiffness has become a topic of some interest as a biomarker of esophageal varices with several studies reporting promising results [79–83]. TE is not well suited for spleen stiffness measurement owing to the lack of integrated B-mode US imaging in the device and so failure rate is quite high. Concomitant US guidance is generally required to ensure correct probe placement.

Overcoming this drawback, acoustic radiation force impulse (ARFI) elastography methods are alternative techniques which are gaining popularity owing to the integration with commercial US systems. ARFI also measures tissue stiffness with the added benefit of real-time imaging simplifying region of interest placement. ARFI is commonly available in two forms, point shear wave elastography (pSWE) and 2D shear wave elastography (2D-SWE). Both methods have been much less comprehensively studied than TE in PH and liver disease in general; however existing reports indicate ARFI methods exhibit excellent diagnostic performance in predicting CSPH and the presence of varices [84–88].

Aside from US elastography, magnetic resonance elastography (MRE) allows the measurement of tissue stiffness during clinical MR exams by encoding the propagation of shear waves into the MR phase signal. MRE has excellent ability to detect liver cirrhosis [89,90] and the cross sectional nature of MRI enables simultaneous acquisition of liver and spleen MRE data with additional equipment [91,92]. MRE as a predictor of PH and esophageal varices is an emerging field of research with promising results [93–97], specifically related to spleen stiffness [94,96]. A recent meta-analysis [98] found liver and spleen stiffness measured with MRE had encouraging AUC values for prediction of CSPH (AUC ≥0.88) with diagnostic performance reduced for LS when evaluating the presence of esophageal varices only (AUC=0.76 for liver stiffness, AUC=0.89 for spleen stiffness). A single MRI exam incorporating cirrhosis and PH screening would be a powerful clinical tool however further prospective studies are required to evaluate the utility of MRE for this purpose. Clinicians utilizing elastography methods for patient selection and risk assessment should be aware of the limitations of elastography methods, such as LS variation following meal intake [99–105], during ALT flares [106–108] and as a result of cholestasis [109–111]. Additionally, as liver fibrosis progresses and extrahepatic factors such as splanchnic vasodilation and hyperdynamic circulation begin to influence portal pressure independent of LS [112], the correlation between LS and HVPG is reduced [113]. Failure rate, and the ability to obtain reliable measures are also important considerations. TE has been found to be sensitive to obesity, with one study reporting uninterpretable readings in almost 20% of cases [114]. The introduction of the XL probe has improved this limitation somewhat [115]. ARFI methods appear to have higher reliability than TE [116] however failed and unreliable measures are still affected by BMI [117]. MRE is generally acquired with a gradient recalled echo sequence which is sensitive to high liver iron deposition [118] however spin-echo echo-planar imaging sequences are more robust in this setting [119]. Finally, care should be taken when comparing measurements between elastography techniques as differing parameters are reported. TE and ARFI methods generally report the Young’s modulus with units of kilopascals, however shear wave speed (m/s) is also sometimes reported. MRE typically displays the “shear stiffness”, also referred to as the magnitude of the complex shear modulus, which is also in units of kilopascals.

Phase-Contrast MRI

Like MRE, phase-contrast (PC)-MRI is another quantitative MRI technique in which mechanical measurements of tissue are encoded in the phase of the MRI signal. PC-MRI acquisitions use velocity-sensitive gradients to collect phase data over multiple cardiac cycles using ECG-gated imaging to measure pulsatile blood flow in time. PC image reconstructions generate time-resolved (CINE) sets of anatomical images (magnitude of the MR signal) and flow velocity (phase-difference) images that capture vessel morphology and allow measurement of blood flow (ml/min) and velocity (cm/s) during the cardiac cycle. The major advantages of PC-MRI are flow quantification and vessel visualization without the use of gadolinium contrast agent, as well as lower intra- and inter-observer variability compared to Doppler US [120–124]. However, as PC-MRI data is collected and averaged over multiple cardiac cycles, PC-MRI measurements do not capture velocity changes within the different cardiac cycles and short time fluctuations, and thus do not provide a real-time snapshot of the blood flow velocities like in Doppler US. The temporal averaging also causes PC-MRI to under-estimate Doppler US velocities [123].

2D PC-MRI

2D PC-MRI clinical protocols include single-direction velocity measurements orthogonal to a 2D imaging slice that is placed perpendicularly through the vessel of interest (through-plane encoding) [125,126]. While its first applications have been in the heart, 2D PC-MRI can be used for hemodynamic assessment of the liver vasculature [127–130], alone or in multiparametric MRI protocols [131,124]. 2D PC-MRI has been used to evaluate differential blood flow in the arterial, portal, splenic and splanchnic circulation of patients with liver cirrhosis and PH compared to controls [129], has been validated against HVPG [128] (Table 1), and has been used to assess response to beta-blockers [130]. Patients with cirrhosis and PH have increased resting blood flow in the thoracic aorta, hepatic artery, superior mesenteric artery [129], while total renal arterial flow is reduced [129], and PV flow is reduced or unchanged [128,129]. In several studies, supraceliac aorta [128], hepatic artery [127], PV flow [127,128] and total liver blood flow (TLBF) calculated by subtraction of infrahepatic inferior vena cava (IVC) flow from the suprahepatic IVC flow [127], were not directly correlated with HVPG. However, hepatic arterial flow fraction, calculated through caval subtraction, was strongly correlated (r=0.78, p=0.014) with HVPG [127]. The log of azygous flow was strongly correlated to HVPG (Pearson r=0.87, p<0.001), and identified CSPH with excellent diagnostic performance [AUC (95% CI)=0.96 (0.91-1.00); Table 1) [128]. This finding was reproduced in another study of 30 patients with liver disease and suspected PH, assessing the diagnostic utility of 2D PC-MRI in combination with other quantitative MRI methods (arterial spin labeling perfusion, liver T₁ mapping, LS measurement by MRE), which found moderate correlations of azygous vein flow (Pearson r=0.52, p=0.004) and velocity (Pearson r=0.67, p<0.001) with HVPG [131]. The same study identified a predictive linear regression model for HVPG, combining splenic artery velocity and liver T₁, which was validated in a separate cohort of 10 patients [131].

The main factor that affects the robustness of 2D PC-MRI measurements is the precise positioning of the measurement plane orthogonally to the vessel of interest [125,126,132]. To help with positioning, an angiography survey scan (with or without gadolinium) [129,130] or anatomical images in multiple orientations are usually acquired before PC-MRI. Other disadvantages of 2D-PC that affect image quality and quantitative measurements are the choice of the velocity encoding (VENC) parameter, which relates the phase difference to velocities [125,126,132], phase-offset errors [125,132], and faulty ECG gating due to arrhythmia. To obtain optimal image quality, the VENC should be as high as the typically observed peak velocities in the vessel to avoid aliasing, but as low as possible to reduce noise in velocity measurements. The choice of VENC is often vessel specific, with typical VENC setting of 40-60 cm/s for the portal vein (PV) [127,128,121,124], 60-80 cm/s for the IVC [127], and 150-250 cm/s for the supraceliac aorta, celiac trunk and hepatic artery [129]. Strategies to mitigate velocity aliasing, phase-offset errors (e.g. velocities of stationary tissues appear different from zero), and arrythmia rejection algorithms have been integrated in PC-MRI post-processing software packages or in the image reconstruction algorithms [125,132,133].

4D Flow MRI

4D flow is the extension of 2D PC-MRI in which volumetric (3D) time-resolved phase-contrast data is acquired with three-directional velocity encoding [125,126] (Figure 2). The three-directional velocity data allows visualization of flow streamlines, streak lines and particle traces outlining areas of turbulent flow along the vessel, and use of computational fluid dynamics to calculate wall shear stress [134] and pressure differences along the vessels [135]. 4D flow MRI has emerged as an important technique for non-invasive evaluation of hepatic hemodynamics in cirrhosis and PH [133,136–143,125,126]. Its major advantage compared to 2D-PC MRI is the ability to measure flow in any 2D cross-section of the measured vessels of interest, without the need for a separate acquisition for each measurement. The major disadvantage of 4D flow MRI and the main reason its use is restricted to research studies is its long acquisition time (10-20 min for a respiratory and cardiac-gated Cartesian acquisition [140]) and complex post-processing. However, improved respiratory gating techniques and novel data collection techniques such as non-Cartesian acquisition (acquiring k-space data in a radial [136,138,144,137,139] or spiral [133,145] trajectory) or accelerated Cartesian technique that exploits MR data sparsity have managed to accelerate data acquisition up to 5-fold [146] and decrease total acquisition time for an abdominal volume covering the liver vasculature to as short as a 20s breath-hold [145,133]. The second technical limitation of 4D flow is its use of a preset velocity encoding (VENC) parameter for measurements in all vessels captured within the acquisition volume. Accurate flow visualization and velocity quantification may be compromised with inappropriate VENC settings, as demonstrated in a patient with transjugular intrahepatic portosystemic shunt (TIPS) in Figure 3. The VENC for the 4D flow acquisition must thus be set in accordance to the clinical question investigated, and multiple acquisitions with different VENCs can be performed (e.g. one with a VENC= 50-60 cm/s for PV and hepatic veins, and one with VENC=100-150 cm/s for flow in the arteries and TIPS), albeit with the cost of further prolonged acquisition time.

Figure 2.

4D flow of the abdomen in a 30-year-old healthy male volunteer. Upper abdominal vessel conspicuity on a maximum intensity projection derived from 3D velocity measurements with spiral 4D flow scans with different VENC values. The acquisition with VENC=120 cm/s achieves the best depiction for the hepatic artery (HA, red arrow), with portal vein (white arrow) having poor conspicuity, while the acquisition with VENC=60 cm/s achieves intermediate conspicuity for the hepatic artery (red arrow) and the portal vein (white arrow). The acquisition with VENC=20 cm/s achieves the best conspicuity for the portal vein (PV, white arrow) and hepatic veins (HV, yellow arrows).

Figure 3.

4D flow of the abdomen in a 53-year-old male patient with hepatitis B virus cirrhosis who underwent transjugular intrahepatic portosystemic shunt (TIPS) placement for clinically significant portal hypertension (HVPG before TIPS: 27 mmHg; immediately after TIPS placement: 6 mmHg). Streamlines depiction superimposed on a maximum intensity projection from spiral 4D flow acquisitions of upper abdominal vessels: VENC=120 cm/s acquisition achieves better depiction of the shunt and high velocities (yellow arrow) than the acquisition with VENC=60 cm/s. Time averaged velocity, flow, and cross-sectional area in the portal vein (PV) and the TIPS for both acquisitions are given in table below.

The initial feasibility studies in patients with cirrhosis and PH found no correlation between visualization of 13 abdominal vessels on 4D flow images and MELD scores [144], and high variability in PV and supraceliac aorta flows in patients with PH compared to healthy controls [138]. An initial feasibility study found a different response of the splanchnic vasculature to a meal challenge in cirrhotic patients vs. controls [139]. The results suggest that patients have an impaired hepatic flow response to the meal challenge, which is partly compensated by the collateral circulation (azygous vein). Qualitative changes were also observed, with splenic vein flow direction changing from hepatopetal to hepatofugal in response to the meal challenge in two patients [139]. No studies to-date have correlated 4D flow measurements to HVPG, while one study [145] found positive correlations between flow in the splenic vein, velocities in the SMV, and peak velocities in the infrarenal IVC and an imaging-based PH composite score [54]. In the same study, 4D flow-derived area measurement in the splenic vein identified cirrhosis and CSPH (PH score ≥ 4) with good diagnostic performance. Diagnostic performance for CSPH was improved for a logistic regression model combining splenic vein area with alcohol misuse etiology of liver disease [145].

The findings of 2D PC-MRI and 4D flow studies confirm the known phenomena of hepatic arterial buffer response (HABR) [147] and hyperdynamic circulation in patients with liver cirrhosis and PH. By HABR, flow in the hepatic artery and the arterial fraction of total liver flow are increased, in an attempt to compensate portal venous resistance and maintain total liver flow constant. McAvoy et al. [129] speculate that the hyperdynamic circulation observed in patients with end-stage liver disease is due to “splanchnic steal”: an increase in splanchnic blood flow in an attempt to compensate for increased portal resistance and to maintain the PV flow constant, at the expense of renal blood flow [129], and possibly cerebral blood flow. Progressive splanchnic steal, observable by flow-sensitive MRI, may thus explain some of the clinical consequences of end-stage liver disease and PH, such as hepatorenal syndrome due to reduced renal blood flow, and hepatic encephalopathy, due to impaired auto-regulation of cerebral blood flow [129].

PC-MRI methods are valuable for qualitative and quantitative evaluation of abdominal hemodynamics in PH, with less operator-dependence than Doppler US. However, shortened acquisition time (in the case of 4D flow), greater vendor support of post-processing software, and more validation studies against the invasive reference standard are needed before clinical adoption of PC-MRI as a diagnostic and prognostic tool in PH.

Dynamic contrast-enhanced MRI (DCE-MRI)

DCE-MRI (using a gadolinium based contrast agent) uses the unique physiology of liver perfusion (dual input from the hepatic artery and the portal vein) to characterize diffuse liver fibrosis, which progresses to PH. DCE-MRI has been shown to capture the pathophysiological changes in the liver caused by fibrosis and PH by measuring prolonged blood transit time of contrast agents in the liver, and the increase in the relative contribution of the hepatic artery vs. the PV to total liver flow, described above as the HABR [148–150].

While 4D flow provides vascular flow values (ml/min), DCE-MRI allows an estimation of regional liver perfusion in ml/min/100g tissue. Regional perfusion can be assessed by following the uptake and washout of the contrast agent in the tissue, using either DCE-MRI or DCE-CT. Unlike CT, MRI does not use ionizing radiation, which makes it ideal to repeat in longitudinal studies to monitor the disease course or response to treatment. The gadolinium-based contrast agents used in MRI have a better renal and allergic safety profile compared to iodinated contrast agents for CT, although, given the concerns about nephrogenic systemic fibrosis (NSF) and gadolinium retention in the brain [151] and bones [152,153], gadolinium-based contrast agents should be prescribed more conservatively and avoided in patients with impaired renal function. Macrocyclic contrast agents, with reduced retention compared to linear agents [154] and virtually no documented cases of NSF and deposition, should be used when possible. Patients with CLD, cirrhosis and PH typically receive either extracellular or hepatobiliary gadolinium-based contrast agents as part of their clinical screening for hepatocellular carcinoma. Thus, DCE-MRI can be integrated in the clinical workflow, with no additional administration of gadolinium contrast agent beyond clinical necessity. DCE-MRI acquisitions have higher temporal resolution (2-4 s/volume vs. 15-20 s /volume) than the 3D T₁-weighted pre- and post-contrast acquisitions used for clinical interpretation, so the DCE-MRI frames corresponding to the arterial, portal venous, and hepatobiliary phases can be interpreted clinically, or the DCE-MRI acquisition can be interrupted for portal venous (1 min after injection) and hepatobiliary (10-20 min after injection) T₁-weighted acquisitions with higher spatial resolution.

Annet et al. were the first to correlate DCE-MRI parameters from their dual-input, single-compartment model of liver perfusion with severity of liver cirrhosis and HVPG [148]. All DCE-MRI parameters were significantly correlated with HVPG, with a substantial negative correlation for PV fraction (r=−0.77, p<0.001), and strong positive correlation for mean transit time (r=0.721, p<0.001) [148]. PV flow Fp also decreased, while arterial fraction ART% increased with increasing Child-Pugh score [148]. The DCE-MRI data from this study correlated to flow values in the hepatic artery and PV from Doppler US, and is illustrative of the HABR effect in cirrhosis and PH [148]. The subsequent study of Wagner et al. [150] used liver DCE-MRI parameters modeled by the dual-input, single-compartment model, spleen DCE-MRI parameters from the single-input, dual-compartment model, as well as model-free parameters for both organs (Table 1), in combination with MRE, for assessment of PH. Unlike in the Annet et al. study, the PV flow and ART modeled parameters did not correlate with HVPG. However, the model-free parameters of liver upslope, liver time-to-peak (TTP) and the modeled extracellular distribution volume (DV) correlated significantly with HVPG. Spleen flow did not correlate with HVPG, unlike in previous studies with DCE-CT [155,156]. The discrepancies in published results are likely due to differences in modalities (MRI and CT), and in MRI protocols (different acquisition parameters and contrast agents used in the MRI studies). The study of Wagner et al. highlights the diagnostic value of model-free DCE-MRI parameters, especially in combination with other quantitative MRI techniques such as MRE, however was limited in sample size.

Decreasing liver concentration-time curve upslope with severity of PH, as observed by Wagner et al. with an extracellular agent, is consistent with the observation by Asenbaum et al. of decreased relative enhancement of liver parenchyma using gadoxetic acid with severity of PH [157]. Moreover, the qualitative assessment of PV hyperintensity at 20 min. after administration of gadoxetic aid was strongly associated with severe PH, independently of splenic size, portosystemic collaterals, and ascites.

The studies above highlight the potential of both modeled and modeled-free quantitative DCE-MRI parameters to identify severity of PH and to achieve a physiological understanding of its effects on liver function. As obtaining DCE-MRI parameters is time and computation-intensive, the most clinical value can be derived by integrating DCE-MRI in the clinical evaluation, and streamlining data acquisition and post-processing.

Relaxometry (T₁, corrected T₁, T₂, T₂*, T₁ρ/magnetization transfer)

Topographic mapping of MR signal relaxometry parameters (T₁, corrected T₁, T₂, T₂*,T₁ρ) in the liver and spleen allows non-invasive assessment of tissue composition in CLD and PH. T₁, T₂, T₂* and T₁ρ of the liver and spleen have been shown to reflect fibrosis, edema, fat infiltration and iron deposition, pathophysiological processes that cause, or are exacerbated by, PH. Relaxometry parameters are seldom studied in isolation, and most often are incorporated in PH studies as an adjuvant to other techniques (e.g. pre-contrast T₁ is acquired as calibration parameter for DCE-MRI studies, as shown above), or in combination with other relaxometry parameters and with other quantitative MRI techniques (e.g. MRE, diffusion-weighted MRI, DCE-MRI).

T₁ mapping measures the proton spin-lattice (longitudinal) relaxation time for each image voxel. A T₁ map can be calculated from multiple raw images acquired with different time delays (inversion times) following an inversion recovery (IR; 180º) pulse, with different repetition times (TR) following a saturation recovery (SR; 90º) pulse, or with different flip angles while keeping the TR constant, as in the variable flip angle (VFA) method. In liver studies, T₁ has been typically measured with IR methods, such as Look-Locker or Modified Look-Locker Inversion Recovery (MOLLI), which have been shown to have good precision, reproducibility [158], and few artifacts [159], despite limited coverage of 1-2 slices acquired in each breath-hold. The VFA method can provide whole-liver coverage in one breath-hold [160], but is less repeatable and accurate than IR methods, especially at 3T [158].

As T₁ values reflect the MR relaxation behavior of both interstitial and intracellular protons, T₁ is sensitive to the presence of fat and iron, as well as to the T₁-shortening effect of gadolinium contrast agents present in the extravascular, extracellular space, or in the case of hepatobiliary contrast agents, in hepatocytes. T₁ mapping has been used for liver and spleen tissue characterization by measuring native (without gadolinium contrast) and post-gadolinium contrast T₁, as well as the extracellular volume fraction (ECV), the ratio of differences in relaxation rates (1/T₁) after and before contrast in the tissue and blood pool [161]. ECV is advantageous over pre and post-contrast T₁ measurements, as it is independent of field strength, minimizes systematic errors in image acquisition, and is less sensitive to the effects of fat and iron on T₁. Iron-content corrected native T₁ (cT₁), using T₂* measurements acquired in the same location as the T₁ map to estimate the effect of iron deposition, has been proposed to address the confounding effects of iron deposition on T₁, with good reproducibility [162], and diagnostic and prognostic value in CLD [163,164]. T₂ mapping techniques measure the proton spin-spin (transverse) relaxation time, by acquiring T₂-weighted images at different echo times (TE), with a long TR to minimize the effect of spin-lattice relaxation time [159]. T₂ is obtained by mono-exponential fit of signal values at the various TE. As T₂ reflects free water proton content, it has been used to quantify edema in liver and spleen tissues [165,166,161]. T₂ is usually measured by balanced steady-state free precision (bSSFP) or multi-echo gradient echo (mGRE) pulse sequences. Radial imaging has been proposed to obtain whole-liver T₂ maps [166]. T₂* mapping is a faster spin-spin decay constant, as it is affected by the signal dephasing effects of local field inhomogeneities from susceptibility differences (e.g. between water and fat, or iron-containing molecules and water) present in the voxel. mGRE sequences with multiple TE acquisitions are used to map T₂*. Because T₂* is sensitive to magnetic susceptibility differences, it is used to measure iron content in the liver with high accuracy [167]. Magnetization transfer imaging and T₁ρ, the longitudinal relaxation time in the rotating frame, are sensitive to the presence in tissues of macromolecules such as collagen [168,169]. In human studies, magnetization transfer was unable to distinguish cirrhotic from healthy livers [169], but T₁ρ has shown strong correlations to fibrosis and collagen content in the liver [170] and kidney [171].

Several studies showed that native [131], post-contrast T₁ and ECV positively correlated with HVPG [172], and predicted the presence of gastroesophageal varices on endoscopy [166,172]. The study of Luetkens et al. in two rat cohorts with liver fibrosis induced by bile duct ligation and CCl4 intoxication showed strong positive correlations of native T₁, T₂ and ECV with histopathological markers of fibrosis (Sirius red staining and collagen-1 content) and hepatic stellate cell activation (smooth muscle actin alpha staining, hydroproxyline content) which causes increased collagen deposition in the extracellular space [161]. This animal study offers a mechanistic explanation of increases in portal pressure due to increases in liver fibrosis by extracellular matrix deposition, a process which is marked by increasing ECV, and to a lesser extent by prolonged native T₁ and T₂.

However, the correlation of native liver T₁ with HVPG was not reproduced in human studies [168,150], and native T₁ and T₂ were weaker predictors than MRE for the presence of varices [166]. These results may be due to mixed etiology of PH, and to the unaccounted effect of iron on T₁. Liver cT₁, the iron-corrected alternative to native T₁, had good diagnostic performance (AUC=0.86, p=0.0003) in identifying radiological PH (e.g. splenomegaly, ascites and varices on MRI) in a study of pediatric patients with auto-immune liver disease and PH, and outperformed T₂ and DWI, but was outperformed by MRE [165]. In a proof-of-principle study in adults with PH of mixed etiologies, liver cT₁ was not correlated with HVPG, but spleen cT₁ was strongly correlated with HVPG, had excellent diagnostic performance for both PH and CSPH (AUC=0.92), outperforming LS by TE and liver function tests [173]. In another cohort of patients with PH of mixed etiology undergoing a randomized treatment study with two non-selective beta-blockers, liver cT₁ showed no changes in response to treatment [130]. The diagnostic performance of spleen cT₁ was superior to that of liver cT₁, stiffness and serum markers because changes in the spleen tissue reflect the consequences of PH (regardless of etiology), whereas the liver markers reflect liver fibrosis, a single element in the pathophysiology of PH that may vary according to liver disease etiology.

A recent study with native T₁ and T₁ρ assessment of both liver and spleen in adults with PH, in which spleen T₁ρ, but not liver T₁ρ or T₁, correlated with HVPG and had good diagnostic performance for PH and CSPH (AUC>0.8), shows parametric mapping in the spleen to be promising for the non-invasive assessment of PH [168]. This finding for liver T₁ρ is not surprising, given the mixed etiology of the patients; previous studies reported a strong correlation with fibrosis, the main liver tissue characteristic related to PH, in a cohort of patients with HCV [170], which could not be reproduced in another cohort of mixed etiologies [174]. Although changes brought about by PH in the spleen are increased vascular pressure and congestion, a study has also described hyperplasia of histiocytes and evolution of the reticuloendothelial fibers into diffuse fibrosis which T₁ρ is more likely to capture [175].

The recent studies of parametric mapping in PH show promise of liver and spleen quantitative MRI parameters for the diagnosis of PH. ECV calculated from T₁ measurements with an extracellular gadolinium based contrast agent is robust, field independent, and supported by evidence from mechanistic studies in animals [161]. Although native T₁, cT₁, T₂ and T₁ρ are field strength dependent, which limits their use in clinical trials on multiple platforms, their measurement can be performed with widely available sequences, does not require extensive acquisition time, specialized hardware as with MRE, or contrast administration as with DCE-MRI. cT₁ has had encouraging results in CLD, but mixed results in PH, as shown above. Although spleen cT1 had showed highest diagnostic performance for PH and CSPH (AUC=0.92), calculation of cT₁ requires acquisition of T₁ and T₂* data, and the use of proprietary software. As an emergent MR biomarker, spleen T₁ρ achieved substantial diagnostic performance (AUC=0.78-0.82) and is especially promising to quantify macromolecule deposition, in a single acquisition, without the need for proprietary software or contrast injection.

Evaluation of response to treatment in PH

Several studies have assessed the changes in LS following beta-blocker therapy for PH. One of the earliest papers found that following beta-blocker administration, the correlation between LS and HVPG in patients with HVPG >12 mmHg was improved, indicating beta-blocker therapy targeted extrahepatic factors increasing portal pressure [113]. Two other studies assessing change in LS after beta-blocker therapy (one with HVPG correlation [176] and one with heart-rate correlation [177] found LS to increase in some patients and decrease in others [176,177]. In the study by Piecha et al, HVPG was reduced in both the group with decreased LS and the group with increased LS. However, the group with increased LS was found to have an increased risk for death or transplantation [176]. Elastography methods have been utilized to determine changes in LS and SS following TIPS with multiple studies across TE [178,179], ARFI [179–181] and MRE [93,182] noting SS to be a superior indicator to LS in detecting modifications following TIPS placement. However, in a study of Budd-Chiari patients who underwent balloon angioplasty, LS was reported to decrease significantly following the procedure [183].

McDonald et al. combined 2D PC-MRI and iron-corrected cT₁ mapping in the liver and spleen to assess response to treatment of PH with non-selective beta-blockers, before and 4 weeks into treatment, and found no changes in vessel hemodynamics, except decrease in flow in the superior abdominal aorta, and no changes in tissue composition as measured by cT₁ [130]. Two studies examined the hemodynamic changes from 24 hours before, to 4 weeks [184], and 2 and 12 weeks [136] after TIPS insertion, respectively. At 4 weeks after TIPS insertion, 4D flow MRI measured increases in PV velocity and a near tripling of PV flow, as well as increases in the hepatic artery and superior mesenteric artery flows by about 50% [184]. In two patients with refractory ascites, 4D flow measurements were <50 cm/s, consistent with TIPS stenosis [185], which was then confirmed by Doppler US measurements [184]. In the longitudinal study with 4D flow at 2 and 12 weeks post TIPS of Bannas et al. [136], increases in flow in the portal, superior mesenteric and splenic vein were significant, but smaller than in the Stankovic et al. study at 4 weeks. The ratio of TIPS-to-PV flow at 2 and 12 weeks was six times larger in a patient who had refractory ascites because of arterio-portal-venous shunting, compared to patients whose ascites had been resolved by the TIPS procedure [136].

Future Directions

The non-invasive assessment of PH and esophageal varices has progressed significantly in recent years to the extent that TE is now an accepted tool in stratifying patient requirement for screening procedures. In future years it is expected that large prospective trials including TE and other elastography methods will be performed to establish solid cutoff values for prediction of esophageal varices over a range of etiologies. The predictive role of SS and potential liver/spleen combination scores are also yet to be fully elucidated. In addition to elastography, promising biomarkers such as T_1p need to be explored further in well controlled prospective trials. Quantification of collagen through the administration of collagen specific MR contrast agents has the benefit of direct measurement of collagen deposition and hence a potentially highly accurate method for establishing the degree of fibrosis or cirrhosis in patients [186,187]. Further investigation is required to establish safety, efficacy and cost in human subjects.

Flow quantification with 4D flow or DCE-MRI can benefit from novel pulse sequences with non-Cartesian acquisition and compressed sensing reconstruction algorithms, which can be used to achieve higher temporal and spatial resolution and reduce respiratory motion artifacts. Improved post-processing methods for 4D flow, using deep learning for automated segmentation [188] of abdominal vessels and flow quantification, would greatly streamline the currently time-consuming 4D flow post-processing. With compressed sensing, contrast-enhanced volumetric data can be reconstructed according to desired application (e.g. few time frames with high spatial resolution for clinical evaluation, or many time frames with high temporal resolution and lower spatial resolution, for quantitative DCE-MRI evaluation) [189–192]. Using MR fingerprinting techniques, the DCE-MRI acquisition sequence can integrate T₁, T₂ and T₂* mapping for signal-to-concentration time curve conversion and relaxometry parameters mapping in the same acquisition [193]. DCE-MRI models suitable for hepatobiliary contrast agents [194], as well as post-processing methods that automate arterial input function selection and liver segmentation would make the clinical adoption of DCE-MRI more likely.

Although administration of gadolinium based contrast agents in patients with CLD has a great clinical benefit that outweighs the risks of gadolinium retention, concerns about gadolinium deposition can be addressed by methods such as arterial spin labeling (ASL) MRI for quantification of perfusion without gadolinium administration. ASL has shown potential in a multiparametric MRI study in PH patients, in which liver perfusion measured by ASL showed a modest positive correlation with HVPG, while the tissue arrival time of labeled protons showed a negative correlation with HVPG [131].

The current trend towards radiomics and artificial intelligence has yet to be fully embraced for the non-invasive assessment of PH however several recent studies have shown encouraging results for both radiomics [195,196] and AI based methods [197] of PH assessment. The potential of big data driven diagnosis is exciting however significant development is required to provide non-expert clinicians with tools matching the ease of use and simplicity of interpretation of TE.

Finally, imaging studies on prediction of outcome are scarce. The contrast-enhanced MRI study of Asenbaum et al. found decreased relative liver enhancement and the portal venous hyperintensity sign on hepatobiliary phase post gadoxetate injection to be associated with lower three-year, transplantation-free survival [157]. Given the scarcity of imaging data on outcome, studies that would combine imaging with clinical and laboratory variables for the prediction of TIPS and transplantation-free survival are a worthwhile avenue of future investigation.

Conclusion

Non-invasive assessment of PH has become an important feature of clinical investigation in cases of advanced CLD and will become both more accurate and relied upon in future years. TE remains the most widely accepted non-invasive test however with the possible emergence of SS as a predictive marker of PH/CSPH and esophageal varices, ARFI methods and MRE may have a role to play. The utility of techniques such as 4D-fow and relaxometry as non-invasive measures of PH has yet to be determined.

Abbreviations

ALT

Alanine aminotransferase

APRI

AST-to-platelet ratio index

AST

Aspartate aminotransferase

cACLD

compensated advanced chronic liver disease

CLD

Chronic liver disease

CP

Child-Pugh

CSPH

Clinically significant portal hypertension

CT

Computed tomography

HABR

Hepatic arterial buffer response

HVPG

Hepatic venous pressure gradient

INR

International normalized ratio

LS

liver stiffness

MELD

Model for end stage liver disease

MRE

Magnetic resonance elastography

MRI

Magnetic resonance imaging

PH

Portal hypertension

PV

Portal vein

SS

spleen stiffness

TE

Transient elastography

US

Ultrasound