Dr. Dunphy is a third-year radiology resident at the
University of Florida in Gainesville. He received an MD from St.
Louis University School of Medicine in 1997. He is planning a
fellowship in magnetic resonance imaging/body imaging at
Northwestern University in 2002.
Accurate detection and characterization of liver lesions is
paramount for appropriate treatment in a wide variety of clinical
settings. Correct identification of benign lesions (e.g.,
hemangioma) will prevent unnecessary invasive procedures.
Detection, characterization, enumeration, and localization of
primary or metastatic hepatic neoplasms is critical for planning
appropriate therapy. During initial staging, after treatment, and
during follow up, the status of the liver helps to predict patient
outcome with many neoplasms, even those treated with only radiation
or chemotherapy. Specific interventions rely on imaging for
planning or guidance including cryoablation, laser
photocoagulation, radiofrequency/microwave ablation, percutaneous
ethanol injection, surgical resection, and transplantation.
In general, magnetic resonance imaging (MRI) has become the
modality of choice for the characterization of focal liver disease.
The radiologist must have a solid understanding of current MRI
techniques for appropriate management of focal liver lesions.
Researchers at the 86th Annual Meeting of the Radiologic Society
of North America (RSNA 2000) presented data demonstrating the
dynamic nature of liver MRI. Improved hardware, MR technique, and
contrast agents allow MRI to evaluate the liver completely and
noninvasively. MRI of the liver now encompasses parenchymal
imaging, magnetic resonance angiography (MRA), magnetic resonance
venography (MRV), and magnetic resonance cholangiopancreatography
(MRCP). Biochemical imaging with MR spectroscopy is available, but
largely experimental. The goal of a complete, noninvasive
evaluation of the liver has been realized and is widely available
with modern scanners and techniques.
The optimal imaging modality for the detection of focal liver
lesions has been robustly debated over the past two decades.
Imaging modalities currently available to specifically evaluate
focal liver disease include transabdominal and intraoperative
ul-trasound, triphasic computed tomography (CT), computed
tomographic arterial portography (CTAP), and MRI enhanced with one
or more types of contrast agent. These are often complementary and
various combinations may be appropriate in different clinical
At various times, delayed CT after high doses of iodinated
contrast, intraoperative ultrasound, and CTAP have each held sway
as the best method for the detection of focal liver lesions. Most
would now agree that MRI with one or more types of contrast has
surpassed all of these modalities in terms of characterization and
is at least equal to each one for detection. This is due to the
high intrinsic soft-tissue contrast, improved biochemical and
anatomic information, sensitivity to perfusion differences,
multiplanar capability, and lack of ionizing radiation.
Though it is clearly superior for the detection and
characterization of focal liver lesions, MRI has not yet replaced
CT or ultrasound for general liver imaging because of cost and
This review will present a practical guide for choosing
sequences and contrast enhancement strategies for MR imaging of
focal liver lesions, including data presented at RSNA 2000. The
typical MRI findings for the most commonly encountered types of
lesions will be described.
Quality MRI of the liver requires meticulous attention to
technical detail. A carefully validated standard sequence set is
critical for a successful examination. Clinical history should be
sought to direct the study and appropriate optional sequences added
if necessary. Moderate field strength magnets (1.5 T) are generally
available and provide a good balance between spatial resolution and
tissue differentiation. When possible, phased array multicoils
should be used instead of the standard body coil. Phased array
multicoils are surface coils that are placed around the abdomen
allowing the receiver coils to be as close to the generated signal
as possible, thus optimizing the signal-to-noise ratio. The
increased signal-to-noise ratio af-forded by phased array coils
allows one to obtain a smaller field of view (FOV), thinner slices,
and higher resolution images compared to the body coil. One
drawback to the phased array multicoil is that the subcutaneous fat
immediately subjacent to the coils tends to have high signal. This
can be suppressed with fat saturation techniques. The high signal
intensity fat can also generate marked ghost artifact due to
Significant improvements in convenience and efficacy of liver
MRI over the last several years have occurred in large part due to
improved gradients and fast imaging sequences. T1-weighted gradient
echo (T1 GRE) images can now be performed during a single
breath-hold. This rapid imaging capability is extremely important
for dynamic post contrast T1-weighted imaging. Table 1 lists
scanning parameters for a practical liver imaging protocol.
Scan orientation selection
--There are several schools of thought concerning the choices of
planes in which to scan. One paradigm suggests that a single
orientation should be used for all sequence types (e.g., axial)
both before and after contrast. This minimizes the number of
necessary sequences, thus minimizing scanning time and simplifying
image review. Lesions are compared across sequence types for
characterization. Additional planes are then imaged only with one
of the sequences (e.g., T1) for delineation of anatomic
Unenhanced T1-weighted im-ages
--Precontrast axial T1-gradient echo (GRE) breath-hold sequences
should be performed. Some institutions routinely acquire
precontrast T1-weighted images (T1WI) using "in phase" (4.2 ms at
1.5 T) and "opposed phase" (2.1 ms at 1.5 T) sequences.
Just as in adrenal imaging, this chemical shift imaging (CSI)
readily detects microscopic intracellular lipid. Such techniques
can diagnose diffuse fatty infiltration of the liver, focal fatty
infiltration, focal fatty sparing, and the intracellular lipid
present within hepatocellular neoplasms.
While unenhanced T1WI are the least sensitive imaging sequence for
lesion detection, they can improve confidence in lesion detection,
characterize blood and fat, and are important in the interpretation
of post contrast images.
Unenhanced T2 weighted images
--T2-weighted images (T2WI) should be performed using fast
sequences termed fast spin echo (FSE) or turbo spin echo (TSE).
These sequences are significantly faster than conventional spin
echo (SE) T2WI. Studies have shown that FSE sequences are at least
equivalent to conventional spin echo sequences for focal liver
lesion detection and characterization.
There are, however, differences in image contrast between FSE and
conventional SE images. Fat has higher signal intensity on FSE
images compared to conventional SE techniques. Also, magnetic
susceptibility effects of iron and other metals are minimized on
FSE images compared to conventional SE images.
Thus using FSE techniques is advantageous when imaging patients
with embolization coils, inferior vena cava filters, or spinal
and may be disadvantageous after injecting a superparamagnetic iron
Optional special purpose sequence
--Additional sequences and/ or imaging planes may be added to the
protocol based on clinical history, prior imaging, or findings seen
on initial images during the current examination. For suspected
hemangioma, the second and third echoes of T2-weighted sequences
are often helpful in confirming the diagnosis. Fat suppression
added to either T1- or T2-weighted sequences can often help to
distinguish focal fatty change or focal sparing with diffuse fatty
infiltration from true hepatic masses.
An extensive body of research has been published on
liver-specific contrast agents. Three classes of MR contrast agents
are now available for liver imaging: extracellular,
hepatocyte-specific, and reticuloendothelial agents (RES). All are
given by the intravenous (IV) route with manufacturer specified
doses, dilutions, and injection durations. Only the four agents
that are currently approved for clinical use in the United States
will be discussed below (Table 2).
--Three paramagnetic chelates of gadolinium have been approved for
imaging in the United States: gadopentetate dimeglumine (Magnevist,
Berlex Laboratories, Wayne, NJ), gadodiamide (Omniscan, Nycomed
Amersham, Wayne, PA), and Gadoteridol (ProHance, Bracco
Diagnostics, Princeton, NJ).
Gadolinium chelates function in a manner similar to iodinated
contrast media. They are nonspecific extracellular MR contrast
agents that rapidly equilibrate between intravascular and
The nonspecific distribution of gadolinium chelates into the
vascular and interstitial compartments of normal and pathologic
tissues requires rapid dynamic MR to maximize differential
enhancement of liver parenchyma and tumors.
Gadolinium increases tissue signal intensity on T1WI. Optimal
imaging parameters include a breath-hold, T1- weighted GRE during
arterial (20 to 30 seconds postinjection), portal venous (60 to 80
seconds postinjection), and equilibrium (3 minutes postinjection)
Specific lesions often demonstrate characteristic enhancement
--At present, only one product is available as a
hepatocyte-specific contrast agent: mangafodipir trisodium
(Teslascan, Nycomed Amersham, Wayne, PA). A chelate of paramagnetic
manganese, Teslascan increases normal liver signal intensity on
T1WI and is best seen on fat-saturated techniques.
Scans may be performed shortly after administration and up to 4
hours afterwards. The clinical utility of manganese chelates is
unclear at this time and at this point it does not appear to offer
more information than dynamic gadolinium imaging.
Reticuloendothelial specific agents
--The only clinically approved RES specific agent in the United
States is ferumoxides (Feridex-IV, Advanced Magnetics, Cambridge,
MA). Ferumoxides are taken up by the RES, which includes hepatic
Kupffer cells, the spleen, bone marrow, and lymph nodes.
Due to its magnetic susceptibility effects, the agent produces T2
shortening, thereby darkening the liver signal intensity on T2WI.
Scanning may be done just after injection and can be delayed for up
to 4 hours. Since most focal liver lesions and tumors have few
Kupffer cells, they maintain their native bright signal intensity
after ferumoxide injection.
One notable pitfall in using ferumoxides is that
well-differentiated hepatocellular carcinomas have been reported to
take up iron oxides.
The uptake of ferumoxides by well-differentiated hepatocellular
carcinoma is usually not uniform throughout the lesion and, in
general, should not provide a diagnostic dilemma, especially with
comparison of precontrast images.
Choice of contrast agents
--Over the last decade, there has been much interest in finding the
optimal contrast agent and scanning technique for the detection and
characterization of focal liver lesions. The debate has centered on
clinical indications for gadolinium chelates, ferumoxides,
manganese chelates, or some combination of these agents. A
significant body of clinically relevant data was presented at RSNA
2000. The general consensus of evidence suggests that dynamic
gadolinium-enhanced imaging alone is the best technique for
detection and characterization of focal liver lesions.
The perfusion information offered by dynamic gadolinium-enhanced
imaging is useful not only to detect liver lesions, but also in
characterizing them. Ferumoxide imaging is roughly equal to dynamic
post-gadolinium imaging for detecting, but somewhat inferior for
characterizing, focal liver lesions.
Ferumoxides may have some utility in characterizing focal nodular
hyperplasia by documenting Kupffer cells in the lesion.
However, ferumoxide uptake is also observed in well-differentiated
hepatocellular carcinomas, adenomas, and hemangiomas. The role of
manganese chelates at this time is unclear and may be of value when
other contrast agents are equivocal. Many trials are actively being
performed on new liver specific contrast agents.
Clinically relevant hepatic anatomy
--Standard nomenclature describes right and left lobes of the
liver, each divided into two segments (right anterior, right
posterior, left medial or quadrate, left lateral). The hepatic
arteries, portal veins, and bile ducts each have similarly named
branches. The caudate lobe is usually considered as a separate
segment. There are typically three hepatic venous branches (right,
middle, and left) and these define the borders of the lobes and
segments superiorly. Inferiorly, the gallbladder fossa separates
the right and left lobes and the fissure for the ligamentum teres
separates the medial and lateral segments of the left lobe.
--Modern liver surgery has evolved into hemostatically controlled
dissection along anatomic planes. Today, up to 75% of the liver can
be resected, provided the postsurgical liver remnant is not
Accurate localization of liver lesions provides critical
information for planning surgical and minimally invasive
intervention. In order to accurately localize liver lesions, the
radiologist must understand the refined functional anatomy of the
liver as defined by Couinaud (figures 1 and 2). The advantage of
this nomenclature is that it provides succinct information as to
the potential resection planes through the liver.
In this nomenclature, the liver is divided by vertical and oblique
planes defined by the three main hepatic veins and a transverse
plane following a line drawn through the right and left portal
branches. The eight segments are numbered clockwise in a frontal
view. Each segment is an independent functional unit supplied by an
individual portal triad.
--In order to have a systematic and practical approach to the
differential diagnosis of focal liver lesions, it is helpful to
consider that the liver is composed of several cellular elements:
hepatocytes, biliary tree (cholangiocellular epithelium), and
mesenchymal tissue. Liver pathology can then be classified as
hepatocellular, cholangiocellular, and mesenchymal.
Table 3 depicts this organizational scheme.
MRI characteristics of common benign lesions
--Hemangiomas (figure 3) are the most common benign lesion of the
liver, with an incidence ranging from 0.4% to 20% by autopsy
Hemangiomas are more frequently seen in middle-aged females, and
are unlikely to be seen in cirrhotic livers.
Histologically, they are composed of large blood-filled spaces.
Most lesions remain stable over time, but may enlarge during
pregnancy due to an estrogen effect.
The major hazard of hemangiomas is the potential for
misclassification as a malignant lesion.
Hemangiomas have characteristic MR features that allow confident
diagnosis in most cases. They are well-defined masses with high
signal on T2WI that become progressively brighter on second and
third echoes of T2WI. Dynamic gadolinium imaging demonstrates
progressive nodular, centripetal enhancement to eventual uniform
Most hemangiomas fill in completely with contrast within 5 to 30
minutes as observed on delayed post-contrast T1WI.
Focal nodular hyperplasia
--Focal nodular hyperplasia (FNH) is the second most common benign
liver lesion comprising 8% of all primary hepatic neoplasms.
FNH is not a tumor but a tumor-like condition which results from
hyperplasia of normal hepatocytes due to a vascular malformation.
Eighty percent of these lesions are seen in middle-aged woman, and
like hepatic adenomas, are associated with oral contraceptive use.
Malignant transformation has never been reported.
FNH is classically a solitary lesion with a central radiating
scar that is felt to be the arteriovascular malformation. MR
diagnosis of FNH requires certain classic imaging features. The
T1WI appearance is variable while the central scar is hyperintense
on T2WI. On dynamic gadolinium imaging, the lesion becomes
hyperintense on the arterial phase except the central scar, which
remains hypointense. During the equilibrium phase, the lesion
becomes isointense to hypointense and the central scar becomes
hyperintense. FNH lesions contain Kupffer cells and normal
hepatocytes, and therefore will take up ferumoxides and manganese
FNH can be confused with fibrolamellar hepatocellular carcinoma
(HCC). Fibrolamellar HCC is a rare form of HCC seen in young
patients without underlying liver disease. Fibrolamellar HCC, which
has a better prognosis than conventional HCC, is typically a large
mass with a central scar that is low signal intensity on T2WI.
After gadolinium administration, there is intense heterogeneous
enhancement on the arterial phase and the central scar may enhance
on the delayed images.
If there is any question that a lesion may be fibrolamellar HCC,
biopsy is mandatory.
--Hepatic cysts occur in 2.5% of patients referred for ultrasound.
The prevalence of cysts increases with age and in women, and unlike
hemangiomas, is seen in cirrhotic livers. Ultrasound is typically
diagnostic, but MR may be of value when complicated cysts are
present, in the setting of multiple cysts (polycystic kidney
disease), or when multiple liver lesions are present and all
lesions need to be characterized.
Liver cysts should be very high signal on T2WI with a thin,
well-circumscribed wall. No contrast enhancement should occur. MR
is particularly helpful in demonstrating intracystic hemorrhage.
--Hepatocellular adenoma (HCA) is a benign tumor of hepatocytes.
Oral contraceptive pills (OCP) and androgenic steroids are
causative agents in HCA. The incidence is 3 per year per 100,000
long-term OCP users, but only 1 per million in nonusers or women
using OCPs for less than 2 years.
HCAs are often found incidentally in young women on OCPs. They have
a propensity to cause life-threatening hemorrhage and there is a
controversial risk of malignant transformation.
Therefore, surgical resection is the treatment of choice in most
HCAs have variable signal characteristic on T1WI and T2WI due to
the presence of fat, blood, and hepatocytes within the lesion.
Dynamic gadolinium demonstrates most lesions to be hypervascular in
the arterial phase.
As with hepatic cysts, MRI is particularly useful in demonstrating
hemorrhage within the lesion. If clinical history and MRI features
are still nondiagnostic, biopsy or surgical resection is
MRI characteristics of common malignant lesions
--The most common cause of malignant liver lesions in the United
States (Table 4) is metastatic disease; they outnumber primary
hepatic neoplasms by anywhere from 18:1 to 40:1.
The most common primary sites that metastasize to the liver are
colon, stomach, pancreas, breast, and lung. The MRI characteristics
vary with each tumor, but are generally classified as hypervascular
Hypervascular metastases (figure 4) are usually from renal cell
carcinoma, pancreatic islet cell tumors, breast carcinoma,
melanoma, thyroid carcinoma, sarcomas, choriocarcinoma, and
carcinoid tumors. These lesions draw their blood supply from the
hepatic artery, which is reflected in the enhancement pattern seen
on dynamic gadolinium imaging. These lesions will show marked
peripheral rim enhancement during the arterial phase with delayed
peripheral or central washout in the portal venous and equilibrium
Hypovascular hepatic metastases (figure 5) usually arise from
colon carcinoma, pancreatic carcinoma, transitional cell carcinoma,
and lung carcinoma. These lesions derive little blood supply from
the hepatic artery or portal vein and are best imaged in the portal
venous phase of gadolinium enhanced imaging. These lesions will
have low signal compared to liver parenchyma.
--Hepatocellular carcinoma (HCC) is the most common primary liver
neoplasm (figure 6). HCC usually occurs in the settings of chronic
liver disease and cirrhosis,
which include hepatitis B and C, aflatoxins, alcoholism, and
hemochromatosis. The incidence of HCC is greatest in the Far East
and Africa due to the high prevalence of hepatitis B and C and
aflatoxins. Alpha fetoprotein is elevated in 80% of cases.
The liver responds to chronic insults and inflammation by
forming regenerative nodules. Some of these regenerative nodules
become dysplastic and evolve into HCC.
This sequence of dedifferation of regenerative nodules to
dysplastic nodules to HCC has important imaging implications.
HCCs present as a solitary liver mass in 50% of cases,
multifocal liver masses in 40% of cases, and diffuse infiltration
in 10% of cases. They have a propensity for vascular invasion,
which is seen in 30% to 50% of cases.
The most important factors affecting long-term survival are the
presence of vascular invasion, the number of tumors, and size
greater than 5 cm.
MRI has become the modality of choice to characterize the number
and size of lesions and to assess vascular invasion.
The T1 and T2 signal intensity of HCC is highly variable. The
pattern and degree of enhancement with gadolinium is directly
related to tumor differentiation, histologic subtype, and degree of
Well-differentiated HCC has minimal arterial enhancement that
washes out on portal venous images. As stated earlier,
well-differentiated HCC contains Kupffer cells and will take up
ferumoxides. Moderate and poorly differentiated HCCs show marked
hypervascularity and therefore enhance on the arterial phase. In a
cirrhotic liver, any abnormal enhancement should be considered
suspicious for HCC.
--Cholangiocarcinoma (CAC) is a malignant tumor arising from bile
Risk factors include Clonorchis sinensis infection, prior exposure
to Thorotrast, primary sclerosing cholangitis, and Carolis disease.
The prognosis is poor and patients typically die within months of
Although not technically considered an intraheptic neoplasm, the
most common type of CAC is hilar cholangiocarcinoma or Klatskin
tumor. These tumors may be small and obstruct the biliary tree. As
the tumor grows, the hilum of the liver may be involved, resulting
in biliary obstruction. Dynamic gadolinium enhanced imaging
demonstrates a characteristic pattern of slow, gradual enhancement
best seen on the equilibrium phase.
Intrahepatic cholangiocarcinoma is a rare primary malignant
These lesions usually present as a large solitary mass without
biliary obstruction. Dynamic gadolinium imaging demonstrates
peripheral enhancement on the arterial phase, and incomplete
central enhancement on the delayed images.
Modern MRI scanners with improved gradients and fast imaging
techniques have become the modality of choice for accurate
detection and characterization of focal liver lesions. The accurate
detection, characterization, and localization of liver lesions is
imperative for appropriate patient care and surgical planning.
Currently dynamic gadolinium imaging combined with T1WI and T2WI
provides the best technique to evaluate focal liver lesions. The
role of liver-specific contrast agents is still unclear and future
research will better define their appropriate clinical utility.
Future directions in liver MRI include the development of
improved pulse sequences to better characterize lesions in a time-
and cost-efficient manner. In addition, the clinical utility of the
current and future liver-specific contrast agents needs to be
The author wishes to thank Drs. Chris Sistrom, Pat Mergo,
Patricia Abbitt, Walter Drane, and Scott Klioze of the University
of Florida for their assistance with this manuscript.