Abbreviated Magnetic Resonance Imaging for HCC Surveillance

Abbreviations

AASLD

American Association for the Study of Liver Diseases

AFP

alpha‐fetoprotein

AMRI

abbreviated magnetic resonance imaging

CPT

Current Procedural Terminology

CT

computed tomography

DWI

diffusion‐weighted imaging

Dyn

dynamic

FP

false positive

GALAD

Gender, Age, AFP‐L3, AFP

Des‐carboxy‐prothrombin

(DCP)

HBP

hepatobiliary phase

HCC

hepatocellular carcinoma

LI‐RADS

Liver Imaging Reporting and Data System

MRI

magnetic resonance imaging

NA

not available

NC

noncontrast

NPV

negative predictive value

PPV

positive predictive value

T1wi

T₁‐weighted imaging

T2wi

T₂‐weighted imaging

TP

true positive

US

ultrasound

Hepatocellular carcinoma (HCC) has been the most rapidly increasing cause of cancer‐related death in the United States over the last two decades and is the second leading cause of cancer‐related death worldwide. To enable early detection and curative treatment options, the American Association for the Study of Liver Diseases (AASLD) clinical practice guidelines recommend HCC screening and surveillance with abdominal ultrasound (US) every 6 months with or without serum alpha‐fetoprotein (AFP) in patients at risk (patients with cirrhosis and/or chronic hepatitis B virus). ¹

The ability of US to optimally visualize the entire liver may be limited by several factors, such as patient obesity, hepatic steatosis, or advanced cirrhosis. ² When visualization of the entire liver is limited, the sensitivity of US for detecting HCC is reduced, particularly for early‐stage HCC. A recent meta‐analysis including international studies reported pooled sensitivity of 84% for detecting HCC of all stages, whereas the sensitivity declined to 45% for detecting early‐stage HCC (as defined by the Milan criteria: 1 nodule < 5 cm or 2 to 3 nodules each <3 cm in diameter without gross vascular invasion or extrahepatic metastasis ³ ). ⁴ Two prospective studies reported even lower sensitivity (27.9%‐31.7%) for early‐stage HCC detection. ⁵ , ⁶ The combination of US and serum AFP provided higher sensitivity (63%) with a decrease in specificity (92% for US versus 84% for US+AFP). ⁴ Although the addition of AFP modestly improves sensitivity, it has suboptimal performance for HCC screening and surveillance, being elevated in the setting of background liver disease and not expressed by all HCCs. As such, it is considered optional in the setting of screening/surveillance in the most updated AASLD guidelines ⁷ and is not recommended by the European Association for the Study of the Liver. ⁸ Other emerging biomarkers for HCC show promising results, including the GALAD score (a serum biomarker‐based model derived from Gender, Age, AFP‐L3, AFP, and Des‐carboxy‐prothrombin (DCP)) ⁹ and liquid biopsy, ¹⁰ , ¹¹ and may find a place in future surveillance guidelines.

The diagnosis of HCC is typically performed with contrast‐enhanced cross‐sectional imaging techniques (computed tomography [CT] or magnetic resonance imaging [MRI]), which provide reasonable sensitivity (depending on tumor size) and high specificity when stringent criteria such as the Liver Imaging Reporting and Data System (LI‐RADS) are used. ¹² Despite favorable accuracy and availability, diagnostic CT and MRI are not ideal for HCC surveillance due to radiation exposure for CT and long examination duration and cost for MRI (approximately 30‐40 minutes for a complete contrast‐enhanced MRI).

Building on the benefits of diagnostic MRI and attempting to mitigate the disadvantages, abbreviated MRI (AMRI) has been proposed as an alternative surveillance strategy. AMRI consists of select sequences designed to answer the targeted question of HCC detection. A reduced number of sequences, compared with a full diagnostic MRI, results in decreased acquisition and interpretation times, improved patient throughput, and possibly reduced cost, all while maintaining high sensitivity for HCC detection.

Different AMRI approaches have been proposed (Fig. 1). Noncontrast (NC)‐AMRI protocols have included T₂‐weighted imaging (T2wi) and diffusion‐weighted imaging (DWI), ¹³ , ¹⁴ , ¹⁵ whereas dynamic (Dyn)‐AMRI protocols include dynamic T₁‐weighted imaging (T1wi) using an extracellular contrast agent with or without T2wi. ¹⁵ , ¹⁶ Gadoxetate‐enhanced hepatobiliary phase (HBP)‐AMRI typically uses T1wi obtained in the HBP (approximately 20 minutes postinjection) with T2wi with or without DWI ¹³ , ¹⁵ , ¹⁷ (Fig. 2). NC‐AMRI has the advantage of being simpler and cheaper to perform compared with contrast‐enhanced AMRI protocols. In HBP‐AMRI, gadoxetate is injected outside the MRI room without need of a power injector or dynamic imaging, ¹⁸ which reduces complexity and potentially improves efficiency. HBP‐AMRI has shown high sensitivity for HCC detection. ¹⁹ , ²⁰ It is important to note that both NC‐AMRI and HBP‐AMRI require a recall study in case of positivity (Fig. 3A). By comparison, Dyn‐AMRI has the advantage of generally not requiring a recall CT or MRI after a positive examination (Fig. 3B), because the major LI‐RADS criteria can be applied on dynamic imaging, thereby potentially enabling simultaneous screening and diagnosis.

FIG 1.

A 63‐year‐old male patient with alcoholic cirrhosis and 12‐mm HCC in segment 8 (arrows) diagnosed with complete gadoxetate MRI performed for HCC screening. Simulated NC, Dyn, and HBP AMRI reconstructed from the complete MRI are shown. The tumor demonstrates hypointensity on axial T1wi without contrast (A), nonrim arterial phase hyperenhancement (B), washout and capsule appearance at portal venous and transitional phases (C and D), moderate hyperintensity on axial fat‐suppressed T2wi (E), hyperintensity on axial high b‐value DWI (F), and hypointensity on T1wi in the HBP (G). Lesion is characterized as LI‐RADS 5 (“definitely HCC”) on complete MRI and Dyn‐AMRI, and positive with the need for a recall study on NC‐AMRI and HBP‐AMRI.

FIG 2.

A 70‐year‐old woman with nonalcoholic steatohepatitis cirrhosis and biopsy‐proved inflammatory adenoma, detected with gadoxetate AMRI performed for HCC screening. Magnetic resonance images (A, T1wi at the HBP; B, T2wi) demonstrate a 3‐cm tumor in segment 5, which is T₁ hypointense and mildly T₂ hyperintense in a cirrhotic liver. The lesion was stable at 2‐year follow‐up MRI (not shown). Lesion is scored positive on HBP‐AMRI and represents a FP.

FIG 3.

Diagnostic algorithms for AMRI in the context of HCC screening and surveillance: (A) NC‐AMRI and HBP‐AMRI. A positive study requires recall full contrast‐enhanced MRI or CT for verification of the initial findings, which could be related to HCC (TP) or non‐HCC lesion (FP). Subthreshold: one or more lesions <10 mm that are not clearly benign; negative: no visible lesion or benign lesion(s): includes cysts or hemangiomas; positive: one or more lesions ≥10 mm that are not clearly benign (e.g., increased signal on diffusion not related to a cyst and/or hypointense nodule on T1wi HBP). (B) Dyn‐AMRI. A positive Dyn‐AMRI study does not generally require a recall verification study because major LI‐RADS criteria are used for diagnosis.

Most published AMRI studies have retrospectively evaluated the performance of simulated AMRI examinations by extracting the relevant imaging sets from complete MRI examinations. These studies have shown variable diagnostic performance for HCC detection (Table 1). Reported per‐patient sensitivity of NC‐AMRI ranges between 61.5% and 86.1%, with high specificity (92.7%‐97.9%). ¹³ , ¹⁴ , ²¹ Based on a retrospective simulation analysis from a prospective Korean screening trial, NC‐AMRI was superior to US (sensitivity of 79.1% versus 27.9%, specificity of 97.9% versus 94.5%). ¹⁴ Simulated HBP‐AMRI, the most published approach, provides excellent sensitivity (80.6%‐89.7%) and specificity (92.7%‐96.1%). ¹³ , ¹⁵ , ¹⁷ , ²¹ , ²² These numbers have been confirmed by the only published study to date of clinically implemented, nonsimulated HBP‐AMRI. ¹⁸ There are fewer available data on simulated Dyn‐AMRI, with two recent studies reporting per‐patient sensitivity of 84.6%‐92.1% and specificity of 88.6%‐99.8%. ¹⁵ , ¹⁶ Another simulated Dyn‐AMRI study showed high concordance with complete MRI for LI‐RADS 5 score assignment. ²³ A retrospective comparison of the three simulated AMRI protocols extracted from a complete gadoxetate MRI in a screening population showed superiority of contrast‐enhanced AMRI sets over NC‐AMRI (Table 1). ¹⁵ Notably, most retrospective AMRI studies have included convenience population samples with higher HCC prevalence rate (range: 13%‐100%) than the prevalence encountered in a screening population (0.25%‐8%). Study enrichment with patients with high prevalence of cancer may inflate the observed performance of AMRI and highlights the need for prospective evaluation of AMRI in HCC screening/surveillance populations.

TABLE 1.

Summary of Study Characteristics and Per‐Patient Diagnostic Performance of Published AMRI Studies

Study	Screening Population	Study Design	n	HCC Prevalence Rate	Sensitivity	Specificity	PPV	NPV
NC‐AMRI
Park et al. 14	Yes	Prospective	382	7%	79.1%	97.9%	61.8%	99.1%
					(US 27.9%)	(US 94.5%)	(US 17.7%)	(US 96.9%)
Vietti Violi et al. 15	Yes	Retrospective	237	5.5%	61.5%	95.5%	44.4%	97.7%
Whang et al. 21	No	Retrospective	263	53.2%	86.1%	92.7%	93.1%	85.7%
Dyn‐AMRI
Lee et al. 23	No	Retrospective	156	NA	High (95%) LI‐RADS concordance with full MRI
Khatri et al. 16	No	Retrospective	86	32.6%	92.1%	88.6%	NA	NA
Vietti Violi et al. 15	Yes	Retrospective	237	5.5%	84.6%	99.8%	95.7%	99.1%
HBP‐AMRI
Marks et al. 17	No	Retrospective	298	16.4%	83.7%	93.2%	71.4%	96.7%
Besa et al. 13	No	Retrospective	174	35.6%	80.6%	96.1%	92%	90%
Tillman et al. 22	No	Retrospective	79	16.5%	85.2%	NA	78%	95%
Vietti Violi et al. 15	Yes	Retrospective	237	5.5%	80.8%	94.9%	47.7%	98.8%
Whang et al. 21	No	Retrospective	263	53.2%	89.7%	92.7%	93.3%	89.1%
Brunsing et al. 18	Yes	Retrospective	141	8.5%	92%	91%	48%	99%

Further studies are needed to define the most accurate and practical AMRI approach for HCC screening and surveillance. In addition to diagnostic performance, clinical feasibility and cost‐effectiveness are important factors to consider. With regard to feasibility in comparison with US, AMRI availability may be lower given fewer MRI scanners, lack of dedicated CPT (Current Procedural Terminology) codes, and more contraindications to MRI. However, preliminary data suggest that AMRI may be a feasible option at least in select patients in academic centers. ¹⁸

A cost‐effectiveness study comparing HCC surveillance with US, CT, complete MRI, and HBP‐AMRI concluded that HCC screening performed with HBP‐AMRI was cost‐effective in a conservative scenario that assumed suboptimal (~50%) surveillance compliance. ²⁴ Goossens et al. ²⁵ showed that an individualized surveillance strategy is cost‐effective, whereby HBP‐AMRI is performed in patients at intermediate/high HCC risk, whereas US is reserved for low‐risk patients. An initial cost‐effectiveness analysis comparing each AMRI approach with US demonstrated that all approaches are cost‐effective compared with US, with superiority of contrast‐enhanced AMRI strategies. ¹⁵

In summary, AMRI is emerging as a promising alternative to US for HCC screening and surveillance. The most accurate and feasible protocol, as well as the target population, need to be defined. Prospective studies, preferably in multicenter HCC screening populations, are needed to rigorously compare AMRI with US. The potential of emerging tumor biomarkers such as GALAD or liquid biopsy should also be assessed in comparison/combination with imaging. ¹⁰ , ¹¹ Additional cost‐effectiveness analysis, with modeling of population characteristics (HCC prevalence, type of chronic liver disease, body habitus) and radiological examination costs, are required to inform the optimal integration of AMRI in clinical practice. Finally, practical steps, such as generation of dedicated CPT codes and standardized reporting templates, will further promote AMRI as a useful clinical tool.