perfusion imaging of the adiology liver: current challenges · entities of the liver affect blood...

Pari V. Pandharipande, MDGlenn A. Krinsky, MDHenry Rusinek, PhDVivian S. Lee, MD, PhD

Published online10.1148/radiol.2343031362

Radiology 2005; 234:661–673

Abbreviation:HCC � hepatocellular carcinoma

1 From the MRI-Basement, SchwartzBldg, NYU Medical Center, 530 FirstAve, New York, NY 10016. ReceivedAugust 25, 2003; revision requestedNovember 6; revision received Febru-ary 3, 2004; accepted March 23. Ad-dress correspondence to V.S.L. (e-mail: [email protected]).

V.S.L. is on the Speakers’ Bureau forBerlex Laboratories.© RSNA, 2005

Perfusion Imaging of theLiver: Current Challengesand Future Goals1

Improved therapeutic options for hepatocellular carcinoma and metastatic diseaseplace greater demands on diagnostic and surveillance tests for liver disease. Existingdiagnostic imaging techniques provide limited evaluation of tissue characteristicsbeyond morphology; perfusion imaging of the liver has potential to improve thisshortcoming. The ability to resolve hepatic arterial and portal venous componentsof blood flow on a global and regional basis constitutes the primary goal of liverperfusion imaging. Earlier detection of primary and metastatic hepatic malignanciesand cirrhosis may be possible on the basis of relative increases in hepatic arterialblood flow associated with these diseases. To date, liver flow scintigraphy and flowquantification at Doppler ultrasonography have focused on characterization ofglobal abnormalities. Computed tomography (CT) and magnetic resonance (MR)imaging can provide regional and global parameters, a critical goal for tumorsurveillance. Several challenges remain: reduced radiation doses associated with CTperfusion imaging, improved spatial and temporal resolution at MR imaging, accu-rate quantification of tissue contrast material at MR imaging, and validation ofparameters obtained from fitting enhancement curves to biokinetic models, appli-cable to all perfusion methods. Continued progress in this new field of liver imagingmay have profound implications for large patient groups at risk for liver disease.© RSNA, 2005

Perfusion imaging provides the ability to detect regional and global alterations in organblood flow. The utility of hepatic perfusion characterization relies on the resolution ofeach component of its dual blood supply—the portal vein and the hepatic artery—becausecontributions from each are altered predictably in many diseases. As early as 1954, Breedisand Young (1) first described relative increases in arterial supply in the setting of primaryand metastatic liver tumors, thus heralding the possibility of arterial phase imaging todetect malignant tumors. Changes in relative arterial and portal venous blood flow also areknown to be associated with the evolution of cirrhosis (2,3). Because most pathologicentities of the liver affect blood flow regionally, globally, or both (4), perfusion imaging ofthe liver has been invoked as a means of improving the sensitivity and specificity ofdiagnostic liver imaging. Herein, we provide a context for liver perfusion imaging andreview its physiologic basis. After highlighting goals for its use and development, wereview existing methods and discuss their role in the evaluation of cirrhosis, hepatocellularcarcinoma (HCC), and metastatic disease. We conclude with a consideration of futuredirections in this emerging field.

BACKGROUND: NEED FOR IMPROVED LIVER IMAGING

To date, existing imaging methods for the diagnosis of cirrhosis and HCC are falling shortof clinical needs. For the detection of cirrhosis, imaging techniques rely primarily onmorphologic criteria—particularly identification of a nodular liver surface and liver atro-phy with secondary signs of portal hypertension. Ultrasonography (US) has been reportedto have a sensitivity of 74% (79 of 107 patients) (5) to 88% (50 of 57 patients) (6) for thedetection of cirrhosis (5–7) when liver biopsy and/or gross observation at laparoscopy wereused as standards of reference. Multiphasic contrast material–enhanced CT protocols have

Reviews

661

Ra

dio

logy

been reported to have sensitivities of 6%(two of 33 lesions) (8) to 71% (15 of 21lesions) (9) for the detection of HCC and0% (zero of 22 lesions) (10) to 39% (nineof 23 lesions) (9) for the detection of dys-plastic nodules, which are premalignantprecursors of HCC (8–12), when liver ex-plant abnormalities are used as a stan-dard of reference. With multiphasic con-trast-enhanced MR imaging, sensitivityvalues of 33% (39 of 118 lesions) (13) to76.9% (10 of 13 lesions) (12) have beenreported for HCC detection and 7% (oneof 15 lesions) (13) to 56% (22 of 39 le-sions) (8) for dysplastic nodule detection(8,10,12–14) when liver explant abnor-malities were used as a standard of refer-ence. In two studies in which lesionswere stratified by size, however, sensitiv-ity values for detection of HCC smallerthan 1 cm were 4% (three of 72 lesions)(13) and 26.7% (four of 15 lesions) (10).In addition, in one study of nine patientswho had diffuse HCC (all nodules � 1cm), only three lesions (all from one pa-tient) were detected (13).

From a public health perspective, thepotential benefit of higher sensitivityand higher specificity liver imaging hasincreased substantially in recent years.HCC, the most common primary livercancer and fifth most common cancerin the world, is responsible for morethan one-half million deaths annuallyworldwide (15). While the reported3-year survival rate for untreated smallHCC (�5 cm) is 21% (16), the 4-yearsurvival rate in patients with cirrhosiswho undergo transplantation for smallHCC (one lesion of �5 cm or as many asthree lesions of �3 cm) is approxi-mately 75% (17). This discrepancy sug-gests that early detection of small HCCmay be critical to patient outcome.Greater organ availability made possibleby the increasing practice of adult-to-adult living related liver donor transplan-tation (18), combined with improved de-tection of small HCC, has the potentialto dramatically influence the care of pa-tients at risk. The early noninvasive de-tection of cirrhosis may enable earlieridentification of patients at risk for HCCand more timely treatment of the dis-eases that cause cirrhosis; moreover, non-invasive detection may obviate histologicsampling.

For the evaluation of metastatic dis-ease, sensitivity values of 74% (60 of 81metastases) (19) to 85% (247 of 290 me-tastases) (20) have been reported whencontrast-enhanced CT is used (19–21); inthe study by Ward et al (19), 17 metasta-ses missed by all readers were smallerthan 1 cm. Histologic evaluation of sur-gical specimens, as well as intraoperativeUS and gross examination of the unre-sected liver, were used as standards ofreference (19–21). Improved detectionand characterization of metastatic dis-ease to the liver also has profound impli-cations for the prognosis and treatmentof patients at risk. The earlier identifica-tion of patients with micrometastatic dis-ease may facilitate future use of targetedchemotherapeutic strategies. Antiangio-genic therapies targeted against vascularendothelial growth factors already havedemonstrated promising results for thetreatment of multiple cancers in phase IIclinical trials (22,23). Moreover, exclu-sion of the presence of micrometastaticdisease could obviate adjuvant chemo-therapy in some patient groups. Im-proved detection of focal metastatic dis-ease also may affect the outcome incandidates for segmental hepatectomy(24–28).

PERFUSION IMAGING OF THELIVER: PHYSIOLOGIC ANDSTRUCTURAL BASIS

Normal Liver Anatomy:Brief Review

In the normal liver, approximatelythree-quarters of the blood supply is de-rived from the portal vein, whereas onlyabout one-quarter is derived from the he-patic artery (29,30). The organizationand delivery of this blood supply is bestunderstood by first observing the under-lying structure of the liver.

The liver is organized into small unitscalled acini, which consist of a centralincoming blood supply from hepatic ar-terial and portal venous branches, bileducts, cords of liver cells, intervening si-nusoids, and a peripheral outgoing sys-temic venous blood supply (Fig 1). Thelow-pressure sinusoids are lined withhighly fenestrated endothelial cells andare flanked by sheets of hepatocytes. Thisarrangement allows plasma to flow freelythrough the sinusoidal endothelial cellsinto the space between sinusoidal endo-thelium and hepatocytes; this space istermed the space of Disse (Fig 2). Low-molecular-weight compounds within se-rum plasma, including conventionalcontrast materials, typically have free ac-cess to the space of Disse.

Evolution of Cirrhosis

The evolution of cirrhosis is associatedwith a progressive disruption of normalanatomy and physiology, which leads toboth regional and global perfusionchanges. Deposition of collagen in thespace of Disse and alteration of sinusoi-dal architecture through the loss offenestra between sinusoidal endothelialcells together result in an increase in re-sistance to incoming sinusoidal bloodflow. As a result, portal venous flow de-creases and begins to bypass the liver pa-renchyma via small portosystemic ve-nous shunts. The mean transit time, aparameter defined as the traversal time ofa compound from liver entry to exit av-eraged over all possible paths, of low-molecular-weight compounds throughthe liver is increased. This increase is at-tributed to restriction of movementwithin the extravascular space caused bydeposition of collagen (31,32). In cirrho-sis, this alteration of portal venous bloodflow is counteracted by an increase inhepatic arterial flow, a process known asthe hepatic arterial buffer response (2,3).By studying different markers of liverfunction (such as lidocaine metabolism)

ESSENTIALS● Improved therapeutic options for hepa-

tocellular carcinoma and metastaticdisease have placed greater demandson liver imaging.

● Existing diagnostic imaging techniquesprovide limited evaluation of tissuecharacteristics beyond morphology;liver perfusion imaging has the poten-tial to improve this shortcoming.

● The primary goal of perfusion imagingof the liver is to enable resolution ofarterial and portal venous componentsof blood flow on a global and regionalbasis.

● Earlier detection of primary and meta-static hepatic malignancies and cirrho-sis may be possible on the basis of rel-ative increases in hepatic arterial bloodflow in these diseases.

● While several technical challenges re-main, continued progress in this newfield of liver imaging may have pro-found implications for the diagnosisand surveillance of liver diseases in thefuture.

662 � Radiology � March 2005 Pandharipande et al

Ra

dio

logy

in patients with cirrhosis following ade-nosine-induced hepatic arterial vasodila-tation, this arterial reserve has beenfound to benefit hepatic function and toimprove liver oxygenation (33). The he-patic arterial buffer response in cirrhosishas been demonstrated in a rat model (2);supportive data also have been reportedin humans with cirrhosis (31,32,34–39).To date, however, a quantitative thresh-old of change in absolute or fractionalarterial flow that delineates the onset ofthis response and the correlation be-tween flow changes and other markers ofcirrhosis have not been determined.

Development of Primary LiverMalignancy: HCC

Folkman and colleagues (40) first pro-posed the dependence of sustained tumor

growth on angiogenesis in the 1960s, arelationship that continues to be heavilyexplored today (41,42). In patients withcirrhosis, a spectrum of nodules, includingbenign regenerative nodules, dysplasticnodules, and HCC, can form; differences intheir respective blood supplies can assist intheir detection and characterization (43–46). Regenerative nodules, like normalliver parenchyma, continue to receive amajority of their blood supply from theportal vein, whereas the evolution from alow-grade dysplastic nodule to frank HCCis associated with a progression toward in-creasing arterial blood supply (43–46) (Fig3). During this evolution, sinusoidal endo-thelial cells are recruited to create an arte-riolar network that gradually replaces nor-mal sinusoidal architecture, and thisprocess is known as “capillarization” (43).

At histologic analysis, dysplastic nodulesand HCC manifest neoarteriogenesis, whichtakes the appearance of unpaired or non-triadal arteries, that is, arteries not associ-ated with portal vein branches (43,45,46).Vascular endothelial growth factor expres-sion, a marker of angiogenic activity, hasbeen found to increase linearly and paral-lels the development of unpaired arteries(47); it is negligible in regenerative nod-ules, moderate in dysplastic nodules, andstrong in HCC (47). Given this progres-sion, serum vascular endothelial growthfactor levels in patients with HCC havebeen explored as markers of tumor activity(48) and as predictors of postoperative tu-mor recurrence and survival (49).

Development of Secondary LiverMalignancy: Metastatic Disease

The formation of a metastasis is initiatedby the extravasation of a circulating tumorcell within the liver (Fig 3). After tumorenlargement progresses to a point beyondwhich angiogenesis is necessary to sustaingrowth, neovascularization occurs in astepwise fashion (50); proliferation of sinu-soidal endothelial cells serves as the pri-mary basis for angiogenesis (50–52). In theperitumoral region, sinusoidal capillariza-tion has been found to occur through aprocess similar to that seen in hepatocarci-nogenesis (50,53). Angiogenesis associatedwith liver metastases is further assisted byvascular endothelial growth factor expres-sion (53,54). In the presence of metastaticdisease, relative increases in hepatic arterialperfusion have been found on a global ba-sis as well (37,38,55–60). The possibilitythat a tumor-related circulating vasoactivemediator contributes to this global perfu-sion change has been suggested (61), al-though further supportive work is neces-sary.

Other Physiologic and IatrogenicCauses of Liver PerfusionAlterations

Although in this review we focus oncirrhosis and malignancy, several otherfactors may contribute to liver perfusionchanges. Of particular note are normalphysiologic alterations in blood flow,such as in response to meals (62,63), aswell as changes in blood flow that occuras a result of increasingly used minimallyinvasive tumor therapies, such as radio-frequency ablation, transarterial chemo-embolization, and percutaneous ethanolinjection (64–75).

An increase in portal venous bloodflow is known to occur in postprandial

Figure 1. Diagram shows anatomic relationship between hepatic arterial and portal venousbranches and bile duct (portal triad), biliary radicles, cords of hepatocytes, intervening sinusoid,and central draining hepatic vein.

Figure 2. Diagram shows microanatomic relationship of red bloodcells within sinusoid, fenestrated endothelial cells, plasma (p), spaceof Disse, and hepatocyte.

Volume 234 � Number 3 Perfusion Imaging of Liver � 663

Ra

dio

logy

states as a result of increased splanchnicblood flow; however, this increase isblunted in cirrhosis, an effect thought tobe due to increased pressure in thesplanchnic venous bed and subsequentportosystemic shunt formation (62,63).The extent to which these opposingchanges in flow occur is variable and isdifficult to predict. As such, evaluation ofglobal arterial versus portal venous flowto the liver should not be performed inthe immediate postprandial state.

Radiofrequency ablation, transcatheterarterial chemoembolization, and percuta-neous ethanol injection are increasinglyused to treat liver malignancy in a varietyof clinical settings (76–78). The enhance-ment characteristics of tissue after treat-ment play a major role in the evaluation oftreatment success. After tumor treatment,the presence of enhancement at dynamiccontrast-enhanced cross-sectional imag-ing, particularly when seen during the ar-terial phase, is suggestive of residual or re-current tumor (64–68,71,72). Peripheralreactive hyperemia, a benign inflamma-tory response to tissue injury after radiofre-quency ablation, however, may manifestas peripheral hyperattenuation during ar-terial phase imaging. This is usually distin-guishable from tumor on the basis of itsregular marginal configuration and en-hancement characteristics at portal venousphase and equilibrium phase imaging;moreover, this finding typically resolveswithin 1–4 months (64,65). Also, periph-eral wedge-shaped enhancement at arterialphase imaging may be seen as a result ofiatrogenic arterial–portal venous shuntsfrom percutaneous therapies (64,65,75).Direct injury to the portal venous system(74) or indirect injury as a result of primarybiliary injury (70) can also complicate en-hancement patterns. Arterial chemoembo-lization is known to result in altered he-patic perfusion (69); the degree to whichthis occurs may depend on the territory ofembolized vessels and the extent of subse-quent arterial collateral circulation. Perfu-sion imaging ultimately may have a role inthe earlier identification of tumor recur-rence if subtle differences in enhancementof benign versus malignant tissue can becaptured earlier than they are capturedwith conventional multiphasic cross-sec-tional techniques.

GOALS OF PERFUSION IMAGING

The primary goal of liver perfusion imag-ing is to increase the sensitivity and speci-ficity with which liver diseases can be iden-tified. The following requirements for

optimal perfusion imaging outline a com-prehensive approach to this task: (a) accu-rate quantification of arterial and portalvenous perfusion on a global and regionalbasis to evaluate both focal and diffuse ab-normalities; (b) high spatial resolution toidentify perfusion differences in small tu-mors; (c) high temporal resolution to iden-tify kinetic properties of contrast agentsthat may vary across tumors; (d) reliableimage-based tracer (contrast agent) con-centration measurements to facilitate accu-rate perfusion quantification; (e) robusttracer kinetic modeling methods for accu-rate derivation of perfusion parametersfrom enhancement curves, including arte-rial and portal venous perfusion, meantransit time of the tracer in the liver, andtracer distribution volume, all of whichmay be abnormal in liver disease; (f) whole-liver imaging to enable tumor surveillance;and (g) compatibility with existing mor-phologic imaging techniques to enable allimaging to be performed during a singlevisit.

In the next section, we review themain approaches to liver perfusion imag-ing, including scintigraphy, US (withDoppler imaging or transit time interro-gation), CT, and MR imaging.

TECHNIQUES OF LIVERPERFUSION IMAGING

Perfusion Indices with Scintigraphy

The calculation of liver perfusion indi-ces at flow scintigraphy was initially de-scribed in the 1970s and is still used to-day. In the most commonly describedapproach, dynamic images are obtainedat 1–2-second intervals after the intrave-

nous administration of a radiopharma-ceutical agent, such as technetium 99m–labeled tin, sulfur, or albumin colloid ortechnetium 99m–labeled pertechnetate(35,39,55–57,59,60,79). Liver enhance-ment is evaluated through a region-of-interest analysis; arterial and portal ve-nous components of liver enhancementare typically distinguished by assumingthat peak renal enhancement parallelsthe onset of the portal venous phase ofliver enhancement (35,55–57,59,60,79).By using the slopes of the rise in activityduring the arterial phase, designated as A,and the portal venous phase, designatedas P, of liver enhancement, a hepatic ar-terial perfusion index value is generatedand is defined as A/(A � P); this index hasbeen the primary parameter of interest inflow scintigraphy of the liver (35,55–57,59,60,79).

Blood Flow Assessment at US

US with Doppler imaging or transittime interrogation enables flow quantifi-cation (blood volume divided by time)as opposed to perfusion measurement(blood volume divided by time dividedby tissue volume or weight). To calculateflow rates at Doppler US, mean velocity ismultiplied by the cross-sectional area ofthe vessel. Flow measurement with atransit time flowmeter, a technique thatnecessitates direct placement of a USprobe over the blood vessel of interest, isbased on measurement of the differencein time that it takes for an ultrasoundbeam to pass upstream versus the timethat it takes for it to pass downstreamthrough flowing blood; this difference is

Figure 3. Diagram shows physiologic basis of perfusion imaging for tumor surveillance. Pro-gressive increase in arterial versus portal venous supply is associated with both the evolution froma low-grade dysplastic nodule to frank HCC and the development of a metastasis from acirculating tumor cell.


Ra

dio

logy

proportional to blood flow. Three pri-mary approaches to flow measurementsin the liver have been described: intraop-erative placement of US probes directlyover the blood vessels of interest (80,81),intravascular measurements (82,83), andconventional transabdominal imaging(36,84). Invasive approaches are not asfeasible in a surveillance setting. By usingtransabdominal US with Doppler imag-ing, flow measurements in the hepaticartery (H) and portal vein (PV) may bereadily obtained, and a hepatic Dopplerperfusion index value, defined as H/(H �PV), may then be computed (36,84). Twoprimary disadvantages of Doppler US inthe context of perfusion imaging includereported intra- and interobserver vari-ability when the Doppler perfusion index(85,86) is measured and the inability togain regional parenchymal flow measure-ments; these may be overcome at CT orMR perfusion imaging.

Perfusion Assessment with CT

In the early 1990s, Miles et al (37) de-scribed liver perfusion imaging at CT byusing a dynamic single-section techniquein which high-temporal-resolution imag-ing (3–7 seconds) was performed afterrapid intravenous administration of a bo-lus of iodinated contrast material. Re-gions of interest were used to generateenhancement curves over the liver, aorta,and spleen (37). Parenchymal enhance-ment was resolved into arterial and por-tal venous components by assuming thatpeak splenic enhancement marked thebeginning of dominant portal venousperfusion (37). Hepatic arterial and portalvenous perfusion values were calculatedby dividing the slopes of the rise in atten-uation during the arterial and portal ve-nous phases of liver enhancement, re-spectively, by peak aortic enhancement(37). A hepatic perfusion index was de-rived by dividing the calculated arterialperfusion by the sum of arterial and por-tal perfusion values (37). To better esti-mate portal perfusion, Blomley et al (38)revised this approach by subtracting arte-rial phase liver enhancement (modeledafter splenic enhancement) from theoriginal liver enhancement curve to de-rive a “corrected” liver enhancementcurve. With this curve, portal venous per-fusion was calculated by dividing theslope of the rise in attenuation during theportal phase of liver enhancement bypeak portal venous or splenic venous en-hancement (38). Calculation of hepaticarterial perfusion was similar to that de-scribed by Miles et al (37). These slope-

ratio methods of deriving perfusion pa-rameters and indices have dominated thefield of CT perfusion imaging (37,38,58,87).

Two groups have applied tracer kineticmodeling techniques to CT perfusion im-aging of the liver (31,88–91). One group(31,88,89) used a dual-input one-com-partment model that may be summa-rized by the following equation:

dCL�t�/dt � k1aCa�t� � k1pCp�t� � k2CL�t�,

(1)

in which CL, Ca, and Cp reflect the tracer(contrast agent) concentration within theliver, hepatic artery, and portal vein, re-spectively, and k1a, k1p, and k2 reflect arterialand portal venous inflow and liver outflowconstants, respectively. Concentrations ofthe iodinated contrast agents are deter-mined from attenuation measurementsthat are based on their known linear rela-tionship (92). Regions of interest placedover the aorta, portal vein, and liver paren-chyma are used to generate concentration-time enhancement curves (Fig 4), wherethe aorta is used as a surrogate for the he-patic artery (31,88,89). A fixed delay timefor tracer transit from the aorta to intrahe-patic hepatic arterial branches is incorpo-rated; an equal delay time also is used toaccount for transit from the main portalvein to portal venous branches within thehepatic parenchyma (31,88,89). By fittingmeasured CL(t) against predicted values,the rate constants, k1a, k1p, and k2, can beestimated to yield the best fit and canthen be used to calculate hepatic arterialand portal venous perfusion, mean tran-sit time of tracer through the liver, andtracer distribution volume within the liver(31,88,89).

Cuenod et al (90,91) used a deconvolu-tion technique to evaluate CT perfusionimaging of the liver. With this approach, atransfer function (a range of flow valueswith each value having a different proba-bility) is substituted for each single-valuedinflow and outflow rate. These transferfunctions are computed by using a lineardeconvolution method and then reducedto derive the same perfusion parametersdetermined by compartmental modelingtechniques (including arterial and portalvenous perfusion, mean transit time of thetracer through the liver, and tracer distri-bution volume in the liver).

Perfusion Assessment with MRImaging

To date, reports of liver perfusion at MRimaging have been limited and vary con-siderably. In 1999, Scharf et al (93) re-

ported use of a single-section T1-weighted gradient-echo technique at1.0 T to perform gadolinium-enhancedperfusion imaging with a 2-second tem-poral resolution in nine pigs prior toand after partial portal vein occlusion.Intrahepatic thermal diffusion probeswere used as the standards of referencefor perfusion measurement (93). The re-lationship between signal intensity andtissue concentration of contrast mate-rial was assumed to be linear, and asingle-input single-compartment modelwas used, which prevented delineationof arterial and portal venous contribu-tions (93). The MR imaging–based liverperfusion after partial portal vein occlu-sion decreased from 117 mL � min�1 �100 g�1 � 42 (standard deviation) to 48mL � min�1 � 100 g�1 � 22; thermal dif-fusion probes measured a decrease inperfusion from an average of 78 mL �min�1 � 100 g�1 � 7 to 47 mL � min�1 �100 g�1 � 15; a statistically significant cor-relation was observed between MR imag-ing–based liver perfusion values and ther-mal diffusion probe measurements (P �.01) (93).

Jackson et al (94) used three-dimen-sional dynamic contrast-enhanced perfu-sion MR imaging in humans to investigatelesion-specific permeability mapping. Theauthors were able to attain 4.1-second tem-poral resolution by using a 128 � 128 � 25matrix (94). Contrast material concentra-tion was determined by using a T1 map-ping sequence (94). Lesion-specific perfu-sion parameters were calculated by using apharmacokinetic model that characterizedendothelial permeability and relativeblood volume (94). In the input functionof the pharmacokinetic model used, onlythe hepatic arterial supply was consideredand the portal vein supply was ignored;therefore, results of evaluation of tissuethat received contributions from both thehepatic artery and the portal vein, includ-ing normal liver parenchyma, were unreli-able.

Materne et al (95) reported single-sec-tion high-temporal-resolution liver perfu-sion imaging in rabbits by using dynamicT1-weighted gradient-echo MR imaging.Tissue tracer concentration was estimatedwith empiric determination of the rela-tionship between signal intensity and T1values with the pulse sequence used (95).Region-of-interest analyses over the aorta,portal vein, and liver parenchyma enabledgeneration of concentration-time curves,which were fitted to a dual-input single-compartment model for liver perfusion(Eq [1]). Arterial and portal inflow and out-flow rate constants were determined by


Ra

dio

logy

using the model and were used to calculatearterial perfusion (23 mL � min�1 � 100mL�1 � 13) and portal venous perfusion(84 mL � min�1 � 100 mL�1 � 32) (95). Thedistribution volume of contrast material(13% � 3.7) and the mean transit time (8.9second � 4.1) were also calculated. All val-ues correlated well with results obtained byusing microspheres (95). This group subse-quently used this method of perfusion MRimaging to evaluate perfusion parametersin rabbits with and without cirrhosis (96)and in humans (32), as discussed in detailnext in the context of perfusion imaging incirrhosis.

PERFUSION IMAGING INCIRRHOSIS

Relative increases in arterial versus portalvenous flow in cirrhosis, as predicted bythe hepatic arterial buffer response, havebeen demonstrated at flow scintigraphy,Doppler US, and CT perfusion imaging(31,35–39). At flow scintigraphy, al-though hepatic perfusion index valueshave correlated well with severity of cir-rhosis, they have not been shown to behelpful in the prediction of major com-plications of cirrhosis, such as ascites,variceal rebleeding, and death (35). Byusing transabdominal Doppler US, Leenet al (36) reported an increase in the he-patic Doppler perfusion index in patientswith cirrhosis.

At CT perfusion imaging, Miles et al(37) and Blomley et al (38) reported in-creased arterial perfusion in patients withcirrhosis. Blomley et al (38) reported avalue of 25 mL � min�1 � 100 mL�1, andMiles et al (37) reported that of 36mL � min�1 � 100 mL�1; in healthy con-trol groups, Miles et al (37) reported avalue of 17 mL � min�1 � 100 mL�1, andBlomley et al (38) reported that of 19mL � min�1 � 100 mL�1. These research-ers also reported diminished portal per-fusion in patients with cirrhosis. Miles etal (37) reported a value of 17 mL � min�1 �100 mL�1, and Blomley et al (38) reportedthat of 43 mL � min�1 � 100 mL�1; inhealthy control groups, Miles et al (37)reported a value of 34 mL � min�1 � 100mL�1, and Blomley et al (38) reportedthat of 93 mL � min�1 � 100 mL�1. Calcu-lation of perfusion parameters was basedon the slope-ratio methods described byMiles et al (37) and Blomley et al (38).

When Van Beers et al (31) used CTperfusion imaging with dual-input sin-gle-compartment modeling (Eq [1]), theyfound increases in fractional arterial per-fusion and mean transit time in a group

with cirrhosis (41% � 27 and 51 sec-ond � 79, respectively), compared with

values in healthy controls (17% � 16 and16 second � 5, respectively); however, a

Figure 4. CT liver perfusion images and graph that shows enhancement curves. (a) Representativetransverse CT images (120 kVp, 60 mAs) from four time points in a CT perfusion study performed in a39-year-old healthy male volunteer (temporal resolution, 3.5 seconds). Rapid sequential single-sectionimaging was performed after intravenous administration of iodinated contrast material at 9.9 mL/sec byusing low-radiation-dose technique. Liver perfusion imaging afforded excellent resolution of arterial (t �15.5 seconds) and portal venous (t � 26 seconds) phases of imaging. Top right: Arterial phase. Arterialvessels, including aorta (a), celiac trunk (c), hepatic artery (h), and splenic artery (sp), selectively enhanced.Bottom left: Portal venous phase. Selective enhancement of the portal vein (p) is seen. (b) Graph showssample enhancement curves derived from analyses of regions of interest placed over aorta, portal vein,and liver. After 57.5 seconds, temporal resolution was decreased substantially to minimize radiationexposure. Use of dual-input single-compartment modeling techniques for derivation of perfusion param-eters required enhancement data from both inputs (aorta and portal vein), as well as the liver.


Ra

dio

logy

statistically significant difference in dis-tribution volume between the twogroups was not observed. Global liverperfusion was lower in the group withcirrhosis (0.67 mL � min�1 � mL�1 �0.23) compared with that in healthy con-trols (1.08 mL � min�1 � mL�1 � 0.34)(31). Increased mean transit time of con-trast material through the liver in pa-tients with cirrhosis was attributed to de-creased motility of standard contrastmaterials (low molecular weight) in theextravascular space of Disse because of itscollagenization (31). The unchanged dis-tribution volume of contrast material wasattributed to continued passage of smallmolecules into the space of Disse, despitechanges of cirrhosis (31).

By using a similar CT perfusion imag-ing technique with dual-input single-compartment modeling, the same groupdemonstrated decreased mean transittime and distribution volume with use ofa higher-molecular-weight contrast agentcompared with a lower-molecular-weightagent in rabbits with hepatic fibrosis (89).These findings were attributed to the pro-gressive restriction of high-molecular-weight molecules to the intravascularspace in cirrhosis (89), as previously re-ported during assessment of the hepaticmicrocirculation of albumin in rats withcirrhosis (97). This work was followed byuse of a perfusion MR imaging techniquewith dual-input single-compartmentmodeling in normal rabbits and rabbitswith liver fibrosis (96). Van Beers et al(96) reported decreased distribution vol-ume with two high-molecular-weightcontrast materials and increased meantransit time for a lower-molecular-weightcontrast material in rabbits with liver fi-brosis. The clearance of a xenobioticagent, indocyanine green, correlated withthe distribution volume of both high-molecular-weight contrast materials stud-ied, and the collagen content of the liverwas inversely related to the distributionvolume of the highest-molecular-weightcontrast material studied (96). More re-cently, this group reported increased frac-tional arterial perfusion and decreasedmean transit time when they evaluated hu-mans with cirrhosis by using dual-inputsingle-compartment perfusion MR imag-ing with a standard low-molecular-weightcontrast material (32).

The limited spatial resolution of flowscintigraphy restricts its current applicabil-ity. Doppler US is restricted by its relianceon measurements of global flow as op-posed to global and regional perfusion.One major limitation of CT perfusion im-aging with slope-ratio methods is that the

mean transit time and distribution vol-ume of contrast material in the liver, pa-rameters found to be altered in cirrhosisby using compartmental modeling tech-niques (31,32,89,96), cannot be assessed.Disadvantages of all methods of CT per-fusion imaging include the required useof iodinated contrast material and its at-tendant risks of allergic reactions andnephrotoxicity (98,99), with the latter ofmajor concern in the setting of end-stagecirrhosis because these patients are at riskfor hepatorenal syndrome (100), and theradiation doses associated with dynamicCT imaging. The latter will be discussedin detail next in the context of HCC sur-veillance. Perfusion MR imaging studies,while promising, need to be validatedfurther in humans.

PERFUSION IMAGING FORDETECTION OF HCC

To date, true perfusion imaging tech-niques have not been used for the sur-veillance of HCC or dysplastic nodules.This is because optimal HCC surveillancenecessitates whole-liver perfusion imag-ing with high spatial resolution to detectand characterize small (�1-cm) lesions.Conventional flow scintigraphic tech-niques and perfusion positron emissiontomographic imaging (101–103) have in-adequate spatial resolution for detectionof regional perfusion abnormalities. Flowmeasurements at Doppler US do not en-able characterization of regional flow ab-normalities.

Whole-liver CT imaging has the poten-tial to provide both high-temporal-reso-lution and high-spatial-resolution imag-ing of the entire liver for the detection ofHCC; however, the radiation doses mustbe considered. By using a reduced-dosewhole-liver imaging technique (120 kVp,100 mAs, 5-mm collimation), the radia-tion dose to the liver with a multi–detec-tor row scanner (Volume Zoom; Siemens,Forchheim, Germany) is approximately2.8 mSv for a man and 3.4 mSv for awoman for a single scan, as indicatedwith commercially available software(WinDose; Wellhofer Dosimetry, Schwar-zenbruck, Germany). With the assump-tion that 15–30 3-second whole-liverscans are obtained, the dose equivalentto the liver would be 42–84 mSv for menand 51–102 mSv for women. In compar-ison, a three-phase liver study performedwith 120 kVp and 165 mAs would resultin a dose equivalent to the liver of 13.8mSv for men and 16.8 mSv for women.Thus, a whole-liver CT perfusion study

would result in radiation doses that areapproximately three to six times greaterthan those of a routine diagnostic CTexamination. A dedicated analysis of therisks versus the benefits of whole-liver CTperfusion imaging may provide greaterinsight into the significance of this in-creased dose in the care of specific patientgroups.

Perfusion MR imaging represents apromising technique for HCC surveil-lance. The closest technique to perfusionMR imaging has been use of a double–arterial phase MR imaging technique en-abled by a rapid parallel imaging method(termed sensitivity encoding, or SENSE);one group reported sensitivity values of91.7% (33 of 36 lesions) for the detectionof HCC and 78.6% (11 of 14 lesions) forthe detection of HCC of 1 cm or smallercompared with sensitivity values of76.3% (29 of 38 lesions) and 27.3% (threeof 11 lesions), respectively, by using aprotocol that included a single arterialphase (104). These findings suggest thepotential of high-temporal-resolutionperfusion imaging to improve detectionof small HCC.

PERFUSION IMAGING FORMETASTATIC DISEASE

In patients with known metastatic dis-ease, increased arterial perfusion hasbeen shown by using CT perfusion imag-ing with the slope-ratio analytic methodsdescribed by Miles et al (37) and Blomleyet al (38). Blomley et al (38) reported avalue of 43 mL � min�1 � 100 mL�1, andMiles et al (37) reported that of 50mL � min�1 � 100 mL�1 in patients withknown metastatic disease versus values of19 mL � min�1 � 100 mL�1 (38) and 17mL � min�1 � 100 mL�1 (37), respec-tively, in healthy control groups. Leggettet al (58) also demonstrated increased ar-terial perfusion (25 mL � min�1 � 100mL�1) in 82% (nine of 11 patients) of acohort of patients with overt colorectalmetastases examined by using CT perfu-sion imaging with slope-ratio analyticmethods. Similarly, at flow scintigraphy,the relationship between an elevated he-patic perfusion index value and overtmetastatic disease has been well estab-lished (55–57,59,60). For example, Per-kins et al (57) determined a hepatic per-fusion index threshold value of 0.37; atvalues above that level, metastatic dis-ease should be suspected.

The detection of micrometastatic or oc-cult metastatic disease with perfusion im-aging has also been studied (56,84,90). By


Ra

dio

logy

using Doppler US, Leen et al (84) showedthat the 5-year survival of patients afterpotentially curative colon surgery for colo-rectal carcinoma was 91% in patients withnormal hepatic Doppler perfusion indexvalues and 29% in patients with abnormalhepatic Doppler perfusion index values. Byusing CT perfusion imaging with a decon-volution technique for data analysis, Cue-nod et al (90) reported decreased portal

perfusion and increased mean transit timein rats with occult liver metastases; arterialperfusion was unchanged. The authors hy-pothesized that arterial perfusion may beincreased in humans with occult metasta-ses secondary to a hepatic arterial bufferresponse; this response may not have beendetectable in rats because of their smallbaseline arterial contribution to liver per-fusion (90).

FUTURE CHALLENGES OFPERFUSION IMAGING

Improvements in perfusion imaging willstem largely from technical advances. CTperfusion imaging with multi–detectorrow technology has the potential to pro-vide high-spatial-resolution and high-temporal-resolution whole-liver perfu-

Figure 5. Images demonstrate trade-off between temporal and spatial resolution when whole-liver perfusion MR images andconventional MR images are compared in a healthy 29-year-old woman suspected of having liver hemangioma (not shown).(a) Representative images from high-temporal-resolution (3.35 seconds) coronal whole-liver perfusion MR imaging per-formed with three-dimensional T1-weighted gradient-echo sequence (repetition time msec/echo time msec, 2.3/0.8; 9° flipangle) sequentially at 1.5 T after administration of intravenous contrast material (7 mL at 5 mL/sec, followed by 20-mLnormal saline flush). Left: Image obtained immediately after initiation of contrast material injection (t � 3.35 seconds).Middle: Peak arterial enhancement (t � 20.1 seconds). Right: Peak portal venous enhancement (t � 30.2 seconds). Perfusionimaging enables excellent temporal resolution of the arterial and portal venous phases as seen by temporal separation inenhancement of the aorta (a) (including celiac trunk [c], superior mesenteric artery [s] branches, and portal vein [p] ). Spatialresolution is suboptimal in that interpolated section thickness is 5.0 mm and voxels (3.1 � 1.8 � 5.0 mm) are anisotropic onimages in a. (b) Representative images obtained from coronal reconstructions from conventional three-dimensional T1-weighted gradient-echo MR images (3.4/1.6, 12° flip angle) obtained after contrast material administration (acquisition time,22 seconds). Left: Arterial phase. Celiac trunk (c), superior mesenteric artery (s), and portal vein (p) are seen. Right: Portalvenous phase. Although a timing sequence was performed to achieve temporal resolution of arterial phase, the arterial phasewas contaminated by portal venous enhancement. Although temporal resolution was poor, spatial resolution was excellent,with in-plane pixel size of 2.3 � 1.4 mm and interpolated section thickness of 2.0 mm, on images obtained during bothphases.


Ra

dio

logy

sion imaging. Necessary methods forradiation dose reduction are in the earlystages of development. Perfusion MR im-aging has the potential to enable dy-namic whole-liver three-dimensional im-aging without the risks of radiation and,therefore, is likely the most promisingfuture approach for perfusion imaging.Effective implementation of perfusionMR imaging, however, requires explora-tion of optimal techniques for fasterhigher-resolution whole-liver imagingand subsequent image processing. More-over, estimation of tissue contrast mate-rial concentration at MR imaging needsfurther work to optimize quantitativeperfusion measurements. Finally, valida-tion of existing methods of liver perfu-sion imaging and perfusion quantifica-tion is necessary to determine whichtechniques are most accurate. The nextsection focuses on the challenges that arerelated to perfusion MR imaging at thistime.

Improved Spatial and TemporalResolution at Perfusion MR Imaging

Strategies that are under developmentto improve spatial and temporal resolu-tion include improvements in hardwareto reduce repetition times, undersam-pling strategies (105), improved interpo-lation algorithms, and new parallel imag-ing methods such as simultaneousacquisition of spatial harmonics (orSMASH) (106) and sensitivity encoding(104), with new coil designs to optimizetheir implementation. The aim of thesestrategies should be to obtain volumetricimaging of the liver with isotropic pixels(of approximately 2 mm or less in allthree dimensions) in a time frame of 3seconds or less. We have implemented afast interpolated three-dimensional gra-dient-echo technique to perform coronalwhole-liver imaging in 3.35 seconds witha pixel size of 3.1 � 1.8 � 5.0 mm (inter-polated section thickness) (107) (Fig 5)

and transverse whole-liver imaging in3.22 seconds with a pixel size of 3.1 �1.8 � 5.0 mm (interpolated section thick-ness) (Fig 6).

Quantification of Concentration ofGadolinium-based Contrast Agents

Unlike CT, in MR imaging the relation-ship between the concentration of gado-linium-based contrast agents and signalintensity is nonlinear and complex. Al-though this relationship may be nearlylinear with specific imaging conditions(primarily related to magnetic fieldstrength, sequences employed, and truetissue concentration of the gadolinium-based agent) (108), more robust tech-niques for the quantification of gadolin-ium-based contrast agents have beenreported (109,110). One such methodused in the context of contrast-enhancedMR imaging of the kidneys relies on theadditivity of T1 relaxivity and requires

Figure 6. Representative images from high-temporal-resolution transverse whole-liver perfusion MR imaging study in a55-year-old woman with history of hemangioma and metastatic melanoma. For perfusion imaging, three-dimensionalT1-weighted gradient-echo whole-liver imaging (2.3/0.8, 9° flip angle) was performed sequentially at 1.5 T with a temporalresolution of 3.22 seconds following the administration of intravenous contrast material (7 mL at 5 mL/sec followed by a20-mL normal saline flush). This enabled an in-plane spatial resolution of 3.1 � 1.8 mm, with an interpolated sectionthickness of 5.0 mm. Representative images from four series in the perfusion study are shown. Top left: Arterial phase (t �23 seconds). Aortic (a) and splenic artery (s) enhancement are noted. Top right: Early portal venous phase (t � 35 seconds).Bottom left: Late portal venous phase (t � 55 seconds). Bottom right: Delayed phase (t � 80 seconds). A large right lobe lesion(arrowhead) is seen that demonstrates peripheral nodular enhancement with progressive central filling over time, featurescharacteristic of a hemangioma.


Ra

dio

logy

determination of the relationship be-tween signal intensity and T1 relaxationtimes for a given sequence, as well as theunenhanced T1 values for tissues of in-terest (109). An alternative and more di-rect approach relies on T1 mapping forperfusion imaging calculations of theconcentrations of gadolinium-based agents(110). These approaches, however, arepresently limited to evaluation of a limitednumber of two-dimensional sections.

Image Processing

For volumetric CT or perfusion MR im-aging of the liver to be clinically useful,automated or semiautomated image pro-cessing tools are needed to register mul-tiple three-dimensional data sets and toanalyze results in terms of useful clinicalparameters, such as arterial perfusionfraction (Fig 7). For the accurate detec-tion of small transient perfusion changes,precise registration of large data sets (on apixel-by-pixel basis) is critical. Typically,manual drawing of regions of interest re-quires adjustment to compensate for re-spiratory motion artifacts. Manual imageanalysis is not only time-consuming andlaborious but also introduces operator-dependent variability. With semiauto-mated image processing methods, liverperfusion parameters could be calculatedwith efficiency and perfusion parametricmaps could be generated on a pixel-by-pixel basis for the entire liver, as deter-mined with application of tracer kineticmodels. Volumetric color-based paramet-ric maps could then facilitate rapid detec-

tion of areas of perfusion abnormality;parametric maps already have been dem-onstrated for single-section CT perfusionimaging (37,87). These tools would facil-itate the diagnosis of global and regionalabnormalities and the efficient extrac-tion of key parameters of diagnostic im-portance from large data sets.

Validation Studies of QuantitativePerfusion Imaging

Validation studies are critical not onlyfor the assessment of methods of dataacquisition but also for evaluation of an-alytic techniques, such as the applicabil-ity of dual-input single-compartmentmodeling to liver perfusion. Moreover,they will help to define the significanceof different factors that may contributeto error in perfusion quantification; thesefactors include parameters such as bolusquality and cardiac function. Computersimulations have been used to estimatethe effect of noise on perfusion parameterswhen compartmental modeling techniquesare used (95). In animal studies, radiolabeledmicrospheres (95) and implantation ofthermal diffusion probes (93) have beenused for validation of perfusion parameters.To validate perfusion studies in humans,alternative approaches must be developed.In addition, validation of sensitivity for thedetection of disease necessitates correla-tion with whole-liver explant pathologywithin a short period following imaging.Thin-section pathologic evaluation of theliver can provide an invaluable reference

standard in the assessment of diagnosticaccuracy for cirrhosis or HCC (13,14).

CONCLUSION

Perfusion imaging of the liver has thepotential to improve detection and char-acterization of liver disease beyond thatenabled by conventional imaging tech-niques. Preliminary efforts in perfusionimaging are promising; however, severalchallenges remain. Although technicallythe most straightforward, clinical imple-mentation of whole-liver CT perfusionimaging requires exploration of methodsto reduce radiation exposure. PerfusionMR imaging permits sequential whole-liver imaging without radiation. Tech-niques that enable improved spatial andtemporal resolution and accurate con-trast material quantification, however,need to be developed. Future efforts toestablish whole-liver perfusion imagingas a clinically feasible and reliable tech-nique may have profound implicationsfor several patient populations. Thus, al-though perfusion imaging remains in theearly stages of development, throughcontinued implementation and explora-tion of advanced imaging technologies,its clinical value in imaging of the livershould begin to unfold in the near future.

Acknowledgment: We gratefully acknowl-edge the assistance of Martha Helmers, BS, inthe creation of the diagrams shown in thisreview.

Figure 7. Sample color segmentation demonstrated with representative coronal and transverse images from dynamicthree-dimensional gradient-echo whole-liver perfusion MR imaging studies (2.3/0.8, 9° flip angle) in a 39-year-old manwithout cirrhosis (transverse and coronal imaging studies were performed independently). Semiautomated segmentationalgorithms enable efficient color-based discrimination of aorta (red), portal vein and its branches (green), liver parenchyma(blue), and hepatic veins and inferior vena cava (dark yellow). Region-of-interest analyses of aortic, portal venous, and liverparenchymal enhancement on the basis of semiautomated segmentation algorithms subsequently can be used for compart-mental modeling.


Ra

dio

logy

References1. Breedis C, Young G. The blood supply of

neoplasms in the liver. Am J Pathol1954; 30:969–977.

2. Richter S, Mucke I, Menger MD, VollmarB. Impact of intrinsic blood flow regula-tion in cirrhosis: maintenance of hepaticarterial buffer response. Am J PhysiolGastrointest Liver Physiol 2000; 279:G454–G462.

3. Gulberg V, Haag K, Rossle M, Gerbes AL.Hepatic arterial buffer response in pa-tients with advanced cirrhosis. Hepatol-ogy 2002; 35:630–634.

4. Itai Y, Matsui O. Blood flow and liverimaging. Radiology 1997; 202:306–314.

5. Colli A, Fraquelli M, Andreoletti M,Marino B, Zuccoli E, Conte D. Severeliver fibrosis or cirrhosis: accuracy of USfor detection—analysis of 300 cases. Ra-diology 2003; 227:89–94.

6. Di Lelio A, Cestari C, Lomazzi A, BerettaL. Cirrhosis: diagnosis with sonographicstudy of the liver surface. Radiology1989; 172:389–392.

7. Hung CH, Lu SN, Wang JH, et al. Corre-lation between ultrasonographic andpathologic diagnoses of hepatitis B andC virus-related cirrhosis. J Gastroenterol2003; 38:153–157.

8. Libbrecht L, Bielen D, Verslype C, et al.Focal lesions in cirrhotic explant livers:pathological evaluation and accuracy ofpretransplantation imaging examina-tions. Liver Transpl 2002; 8:749–761.

9. Lim JH, Kim CK, Lee WJ, et al. Detectionof hepatocellular carcinomas and dys-plastic nodules in cirrhotic livers: accu-racy of helical CT in transplant patients.AJR Am J Roentgenol 2000; 175:693–698.

10. de Ledinghen V, Laharie D, Lecesne R, etal. Detection of nodules in liver cirrho-sis: spiral computed tomography ormagnetic resonance imaging? a prospec-tive study of 88 nodules in 34 patients.Eur J Gastroenterol Hepatol 2002; 14:159–165.

11. Peterson MS, Baron RL, Marsh JW Jr,Oliver JH 3rd, Confer SR, Hunt LE. Pre-transplantation surveillance for possiblehepatocellular carcinoma in patientswith cirrhosis: epidemiology and CT-based tumor detection rate in 430 caseswith surgical pathologic correlation. Ra-diology 2000; 217:743–749.

12. Rode A, Bancel B, Douek P, et al. Smallnodule detection in cirrhotic livers: eval-uation with US, spiral CT, and MRI andcorrelation with pathologic examina-tion of explanted liver. J Comput AssistTomogr 2001; 25:327–336.

13. Krinsky GA, Lee VS, Theise ND, et al.Transplantation for hepatocellular carci-noma and cirrhosis: sensitivity of mag-netic resonance imaging. Liver Transpl2002; 8:1156–1164.

14. Krinsky GA, Lee VS, Theise ND, et al.Hepatocellular carcinoma and dysplasticnodules in patients with cirrhosis: pro-spective diagnosis with MR imaging andexplantation correlation. Radiology2001; 219:445–454.

15. Parkin DM, Bray F, Ferlay J, Pisani P.Estimating the world cancer burden:GLOBOCAN 2000. Int J Cancer 2001;94:153–156.

16. Barbara L, Benzi G, Gaiani S, et al. Nat-

ural history of small untreated hepato-cellular carcinoma in cirrhosis: a multi-variate analysis of prognostic factors ontumor growth rate and patient survival.Hepatology 1992; 16:132–137.

17. Mazzaferro V, Regalia E, Doci R, et al.Liver transplantation for the treatmentof small hepatocellular carcinomas inpatients with cirrhosis. N Engl J Med1996; 334:693–699.

18. Brown RS Jr, Russo MW, Lai M, et al. Asurvey of liver transplantation from liv-ing adult donors in the United States.N Engl J Med 2003; 348:818–825.

19. Ward J, Naik KS, Guthrie JA, Wilson D,Robinson PJ. Hepatic lesion detection:comparison of MR imaging after the ad-ministration of superparamagnetic ironoxide with dual-phase CT by using alter-native-free response receiver operatingcharacteristic analysis. Radiology 1999;210:459–466.

20. Valls C, Andia E, Sanchez A, et al. He-patic metastases from colorectal cancer:preoperative detection and assessmentof resectability with helical CT. Radiol-ogy 2001; 218:55–60.

21. Valls C, Lopez E, Guma A, et al. HelicalCT versus CT arterial portography in thedetection of hepatic metastasis of colo-rectal carcinoma. AJR Am J Roentgenol1998; 170:1341–1347.

22. Fernando NH, Hurwitz HI. Inhibition ofvascular endothelial growth factor inthe treatment of colorectal cancer. Se-min Oncol 2003; 30:39–50.

23. Rosen LS. Inhibitors of the vascular en-dothelial growth factor receptor. Hema-tol Oncol Clin North Am 2002;16:1173–1187.

24. Lang H, Nussbaum KT, Kaudel P, Fru-hauf N, Flemming P, Raab R. Hepaticmetastases from leiomyosarcoma: a sin-gle-center experience with 34 liver resec-tions during a 15-year period. Ann Surg2000; 231:500–505.

25. Carlini M, Lonardo MT, Carboni F, et al.Liver metastases from breast cancer: re-sults of surgical resection. Hepatogastro-enterology 2002; 49:1597–1601.

26. Elias D, Maisonnette F, Druet-CabanacM, et al. An attempt to clarify indica-tions for hepatectomy for liver metasta-ses from breast cancer. Am J Surg 2003;185:158–164.

27. Murata S, Moriya Y, Akasu T, Fujita S,Sugihara K. Resection of both hepaticand pulmonary metastases in patientswith colorectal carcinoma. Cancer 1998;83:1086–1093.

28. Miyazaki M, Itoh H, Nakagawa K, et al.Hepatic resection of liver metastasesfrom gastric carcinoma. Am J Gastroen-terol 1997; 92:490–493.

29. Chiandussi L, Greco F, Sardi G, VaccarinoA, Ferraris CM, Curti B. Estimation of he-patic arterial and portal venous blood flowby direct catheterization of the vena portathrough the umbilical cord in man: pre-liminary results. Acta Hepatosplenol1968; 15:166–171.

30. Schenk WG Jr, McDonald DJ, McDonaldDK, Drapanas T. Direct measurement ofhepatic blood flow in surgical patients:with related observations on hepaticflow dynamics in experimental animals.Ann Surg 1962; 156:463–471.

31. Van Beers BE, Leconte I, Materne R,

Smith AM, Jamart J, Horsmans Y. He-patic perfusion parameters in chronicliver disease: dynamic CT measurementscorrelated with disease severity. AJRAm J Roentgenol 2001; 176:667–673.

32. Annet L, Materne R, Danse E, Jamart J,Horsmans Y, Van Beers BE. Hepatic flowparameters measured with MR imagingand Doppler US: correlations with de-gree of cirrhosis and portal hyperten-sion. Radiology 2003; 229:409–414.

33. Zipprich A, Steudel N, Behrmann C, etal. Functional significance of hepatic ar-terial flow reserve in patients with cir-rhosis. Hepatology 2003; 37:385–392.

34. Koranda P, Myslivecek M, Erban J, Sei-dlova V, Husak V. Hepatic perfusionchanges in patients with cirrhosis indi-ces of hepatic arterial blood flow. ClinNucl Med 1999; 24:507–510.

35. Bolton RP, Mairiang EO, Parkin A, WareF, Robinson P, Losowsky MS. Dynamicliver scanning in cirrhosis. Nucl MedCommun 1988; 9:235–247.

36. Leen E, Goldberg JA, Anderson JR, et al.Hepatic perfusion changes in patientswith liver metastases: comparison withthose patients with cirrhosis. Gut 1993;34:554–557.

37. Miles KA, Hayball MP, Dixon AK. Func-tional images of hepatic perfusion ob-tained with dynamic CT. Radiology1993; 188:405–411.

38. Blomley MJ, Coulden R, Dawson P, et al.Liver perfusion studied with ultrafastCT. J Comput Assist Tomogr 1995; 19:424–433.

39. Rypins EB, Fajman W, Sarper R, et al.Radionuclide angiography of the liverand spleen: noninvasive method for as-sessing the ratio of portal venous to totalhepatic blood flow and portasystemicshunt patency. Am J Surg 1981; 142:574–579.

40. Folkman J, Cole P, Zimmerman S. Tu-mor behavior in isolated perfused or-gans: in vitro growth and metastases ofbiopsy material in rabbit thyroid andcanine intestinal segment. Ann Surg1966; 164:491–502.

41. Folkman J. The role of angiogenesis intumor growth. Semin Cancer Biol 1992;3:65–71.

42. Folkman J. Role of angiogenesis in tu-mor growth and metastasis. Semin On-col 2002; 29:15–18.

43. Park YN, Yang CP, Fernandez GJ, Cu-bukcu O, Thung SN, Theise ND. Neoan-giogenesis and sinusoidal “capillariza-tion” in dysplastic nodules of the liver.Am J Surg Pathol 1998; 22:656–662.

44. Roncalli M, Roz E, Coggi G, et al. Thevascular profile of regenerative and dys-plastic nodules of the cirrhotic liver: im-plications for diagnosis and classifica-tion. Hepatology 1999; 30:1174–1178.

45. Krinsky GA, Theise ND, Rofsky NM,Mizrachi H, Tepperman LW, WeinrebJC. Dysplastic nodules in cirrhotic liver:arterial phase enhancement at CT andMR imaging—a case report. Radiology1998; 209:461–464.

46. Krinsky GA, Zivin SB, Thorner KM, LeeVS, Theise ND, Weinreb JC. Low-gradesiderotic dysplastic nodules: determina-tion of premalignant lesions on the basisof vasculature phenotype. Acad Radiol2002; 9:336–341.


Ra

dio

logy

47. Park YN, Kim YB, Yang KM, Park C. In-creased expression of vascular endothe-lial growth factor and angiogenesis inthe early stage of multistep hepatocarci-nogenesis. Arch Pathol Lab Med 2000;124:1061–1065.

48. Zhao J, Hu J, Cai J, Yang X, Yang Z.Vascular endothelial growth factor ex-pression in serum of patients with hep-atocellular carcinoma. Chin Med J(Engl) 2003; 116:772–776.

49. Chao Y, Li CP, Chau GY, et al. Prognos-tic significance of vascular endothelialgrowth factor, basic fibroblast growthfactor, and angiogenin in patients withresectable hepatocellular carcinoma af-ter surgery. Ann Surg Oncol 2003; 10:355–362.

50. Terayama N, Terada T, Nakanuma Y. Amorphometric and immunohistochem-ical study on angiogenesis of humanmetastatic carcinomas of the liver.Hepatology 1996; 24:816–819.

51. Gervaz P, Scholl B, Mainguene C, PoitryS, Gillet M, Wexner S. Angiogenesis ofliver metastases: role of sinusoidal endo-thelial cells. Dis Colon Rectum 2000; 43:980–986.

52. Paku S, Lapis K. Morphological aspectsof angiogenesis in experimental livermetastases. Am J Pathol 1993; 143:926–936.

53. Terayama N, Terada T, Nakanuma Y. Animmunohistochemical study of tumourvessels in metastatic liver cancers andthe surrounding liver tissue. Histopa-thology 1996; 29:37–43.

54. Takeda A, Stoeltzing O, Ahmad SA, et al.Role of angiogenesis in the developmentand growth of liver metastasis. Ann SurgOncol 2002; 9:610–616.

55. Leveson SH, Wiggins PA, Nasiru TA,Giles GR, Robinson PJ, Parkin A. Im-proving the detection of hepatic metas-tases by the use of dynamic flow scintig-raphy. Br J Cancer 1983; 47:719–721.

56. Warren HW, Gallagher H, HemingwayDM, et al. Prospective assessment of thehepatic perfusion index in patients withcolorectal cancer. Br J Surg 1998;85:1708–1712.

57. Perkins AC, Whalley DR, Ballantyne KC,Hardcastle JD. Reliability of the hepaticperfusion index for the detection of livermetastases. Nucl Med Commun 1987;8:982–989.

58. Leggett DA, Kelley BB, Bunce IH, MilesKA. Colorectal cancer: diagnostic poten-tial of CT measurements of hepatic per-fusion and implications for contrast en-hancement protocols. Radiology 1997;205:716–720.

59. Ballantyne KC, Charnley RM, PerkinsAC, et al. Hepatic perfusion index in thediagnosis of overt metastatic colorectalcancer. Nucl Med Commun 1990; 11:23–28.

60. Leveson SH, Wiggins PA, Giles GR, Par-kin A, Robinson PJ. Deranged liver bloodflow patterns in the detection of livermetastases. Br J Surg 1985; 72:128–130.

61. Carter R, Anderson JH, Cooke TG, BaxterJN, Angerson WJ. Splanchnic blood flowchanges in the presence of hepatic tu-mour: evidence of a humoral mediator.Br J Cancer 1994; 69:1025–1026.

62. Pugliese D, Ohnishi K, Tsunoda T, SabbaC, Albano O. Portal hemodynamics after

meal in normal subjects and in patientswith chronic liver disease studied byecho-Doppler flowmeter. Am J Gastro-enterol 1987; 82:1052–1056.

63. Gaiani S, Bolondi L, Li Bassi S, Santi V,Zironi G, Barbara L. Effect of meal onportal hemodynamics in healthy hu-mans and in patients with chronic liverdisease. Hepatology 1989; 9:815–819.

64. Lim HK, Choi D, Lee WJ, et al. Hepato-cellular carcinoma treated with percuta-neous radio-frequency ablation: evalua-tion with follow-up multiphase helicalCT. Radiology 2001; 221:447–454.

65. Dromain C, de Baere T, Elias D, et al.Hepatic tumors treated with percutane-ous radio-frequency ablation: CT andMR imaging follow-up. Radiology 2002;223:255–262.

66. Katyal S, Oliver JH, Peterson MS, ChangPJ, Baron RL, Carr BI. Prognostic signif-icance of arterial phase CT for predictionof response to transcatheter arterial che-moembolization in unresectable hepa-tocellular carcinoma: a retrospectiveanalysis. AJR Am J Roentgenol 2000;175:1665–1672.

67. Kim HC, Kim AY, Han JK, et al. Hepaticarterial and portal venous phase helicalCT in patients treated with transcatheterarterial chemoembolization for hepato-cellular carcinoma: added value of un-enhanced images. Radiology 2002; 225:773–780.

68. Kubota K, Hisa N, Nishikawa T, et al.Evaluation of hepatocellular carcinomaafter treatment with transcatheter arte-rial chemoembolization: comparison ofLipiodol-CT, power Doppler sonogra-phy, and dynamic MRI. Abdom Imaging2001; 26:184–190.

69. Makita O, Yamashita Y, Arakawa A, et al.Diffuse perfusion abnormality of theliver parenchyma on angiography-as-sisted helical CT in relation to cirrhosisand previous treatments: a potential di-agnostic pitfall for detecting hepatocel-lular carcinoma. Clin Imaging 2000; 24:292–297.

70. Yu JS, Kim KW, Park MS, Yoon SW. Bileduct injuries leading to portal vein oblit-eration after transcatheter arterial che-moembolization in the liver: CT find-ings and initial observations. Radiology2001; 221:429–436.

71. Bartolozzi C, Lencioni R, Caramella D,Falaschi F, Cioni R, DiCoscio G. Hepato-cellular carcinoma: CT and MR featuresafter transcatheter arterial embolizationand percutaneous ethanol injection. Ra-diology 1994; 191:123–128.

72. Bartolozzi C, Lencioni R, Caramella D,Mazzeo S, Ciancia EM. Treatment ofhepatocellular carcinoma with percuta-neous ethanol injection: evaluationwith contrast-enhanced MR imaging.AJR Am J Roentgenol 1994; 162:827–831.

73. Bartolozzi C, Crocetti L, Cioni D, DonatiFM, Lencioni R. Assessment of therapeu-tic effect of liver tumor ablation proce-dures. Hepatogastroenterology 2001; 48:352–358.

74. Yoshikawa J, Matsui O, Kadoya M, et al.Hepatocellular carcinoma: CT appear-ance of parenchymal changes after per-cutaneous ethanol injection therapy.Radiology 1995; 194:107–111.

75. Yu JS, Kim KW, Jeong MG, Lee JT, YooHS. Nontumorous hepatic arterial-portalvenous shunts: MR imaging findings.Radiology 2000; 217:750–756.

76. Livraghi T. Radiofrequency ablation,PEIT, and TACE for hepatocellular carci-noma. J Hepatobiliary Pancreat Surg2003; 10:67–76.

77. Livraghi T, Solbiati L, Meloni F, Ierace T,Goldberg SN, Gazelle GS. Percutaneousradiofrequency ablation of liver metas-tases in potential candidates for resec-tion: the “test-of-time approach.” Can-cer 2003; 97:3027–3035.

78. Camma C, Schepis F, Orlando A, et al.Transarterial chemoembolization forunresectable hepatocellular carcinoma:meta-analysis of randomized controlledtrials. Radiology 2002; 224:47–54.

79. Parkin A, Robinson PJ, Baxter P, LevesonSH, Wiggins PA, Giles GR. Liver perfu-sion scintigraphy: method, normalrange and laparotomy correlation in 100patients. Nucl Med Commun 1983;4:395–402.

80. Doi R, Inoue K, Kogire M, et al. Simulta-neous measurement of hepatic arterialand portal venous flows by transit timeultrasonic volume flowmetry. Surg Gy-necol Obstet 1988; 167:65–69.

81. Jakab F, Rath Z, Schmal F, Nagy P, FallerJ. Changes in hepatic hemodynamicsdue to primary liver tumours. HPB Surg1996; 9:245–248.

82. Kleber G, Steudel N, Behrmann C, et al.Hepatic arterial flow volume and reservein patients with cirrhosis: use of intra-arterial Doppler and adenosine infusion.Gastroenterology 1999; 116:906–914.

83. Hubner GH, Steudel N, Kleber G, Behr-mann C, Lotterer E, Fleig WE. Hepaticarterial blood flow velocities: assessmentby transcutaneous and intravascularDoppler sonography. J Hepatol 2000;32:893–899.

84. Leen E, Goldberg JA, Angerson WJ,McArdle CS. Potential role of Dopplerperfusion index in selection of patientswith colorectal cancer for adjuvant che-motherapy. Lancet 2000; 355:34–37.

85. Oppo K, Leen E, Angerson WJ, CookeTG, McArdle CS. Doppler perfusion in-dex: an interobserver and intraobserverreproducibility study. Radiology 1998;208:453–457.

86. Fowler RC, Harris KM, Swift SE, Ward M,Greenwood DC. Hepatic Doppler perfu-sion index: measurement in ninehealthy volunteers. Radiology 1998;209:867–871.

87. Dugdale PE, Miles KA. Hepatic metasta-ses: the value of quantitative assessmentof contrast enhancement on computedtomography. Eur J Radiol 1999; 30:206–213.

88. Materne R, Van Beers BE, Smith AM, etal. Non-invasive quantification of liverperfusion with dynamic computed to-mography and a dual-input one-com-partmental model. Clin Sci (Lond) 2000;99:517–525.

89. Materne R, Annet L, Dechambre S, et al.Dynamic computed tomography withlow- and high-molecular-mass contrastagents to assess microvascular perme-ability modifications in a model of liverfibrosis. Clin Sci (Lond) 2002; 103:213–216.


Ra

dio

logy

90. Cuenod C, Leconte I, Siauve N, et al.Early changes in liver perfusion causedby occult metastases in rats: detectionwith quantitative CT. Radiology 2001;218:556–561.

91. Cuenod CA, Leconte I, Siauve N, et al.Deconvolution technique for measuringtissue perfusion by dynamic CT: appli-cation to normal and metastatic liver.Acad Radiol 2002; 9(suppl 1):S205–S211.

92. Axel L. Cerebral blood flow determina-tion by rapid-sequence computed to-mography: theoretical analysis. Radiol-ogy 1980; 137:679–686.

93. Scharf J, Zapletal C, Hess T, et al. Assess-ment of hepatic perfusion in pigs bypharmacokinetic analysis of dynamicMR images. J Magn Reson Imaging 1999;9:568–572.

94. Jackson A, Haroon H, Zhu XP, Li KL,Thacker NA, Jayson G. Breath-hold per-fusion and permeability mapping of he-patic malignancies using magnetic reso-nance imaging and a first-pass leakageprofile model. NMR Biomed 2002; 15:164–173.

95. Materne R, Smith AM, Peeters F, et al.Assessment of hepatic perfusion param-eters with dynamic MRI. Magn ResonMed 2002; 47:135–142.

96. Van Beers BE, Materne R, Annet L, et al.Capillarization of the sinusoids in liverfibrosis: noninvasive assessment withcontrast-enhanced MRI in the rabbit.Magn Reson Med 2003; 49:692–699.

97. Varin F, Huet PM. Hepatic microcircula-tion in the perfused cirrhotic rat liver.J Clin Invest 1985; 76:1904–1912.

98. Katayama H, Yamaguchi K, Kozuka T,

Takashima T, Seez P, Matsuura K. Ad-verse reactions to ionic and nonioniccontrast media: a report from the Japa-nese Committee on the Safety of Con-trast Media. Radiology 1990; 175:621–628.

99. Prince MR, Arnoldus C, Frisoli JK. Neph-rotoxicity of high-dose gadoliniumcompared with iodinated contrast. JMagn Reson Imaging 1996; 6:162–166.

100. Moreau R. Hepatorenal syndrome in pa-tients with cirrhosis. J GastroenterolHepatol 2002; 17:739–747.

101. Taniguchi H, Oguro A, Koyama H, Ma-suyama M, Takahashi T. Analysis ofmodels for quantification of arterial andportal blood flow in the human liverusing PET. J Comput Assist Tomogr1996; 20:135–144.

102. Taniguchi H, Oguro A, Takeuchi K, et al.Difference in regional hepatic bloodflow in liver segments: non-invasivemeasurement of regional hepatic arte-rial and portal blood flow in human bypositron emission tomography withH2(15)O. Ann Nucl Med 1993; 7:141–145.

103. Yamaguchi A, Taniguchi H, KunishimaS, Koh T, Yamagishi H. Correlation be-tween angiographically assessed vas-cularity and blood flow in hepatic me-tastases in patients with colorectalcarcinoma. Cancer 2000; 89:1236–1244.

104. Yoshioka H, Takahashi N, YamaguchiM, Lou D, Saida Y, Itai Y. Double arterialphase dynamic MRI with sensitivity en-coding (SENSE) for hypervascular hepa-

tocellular carcinomas. J Magn Reson Im-aging 2002; 16:259–266.

105. Barger AV, Block WF, Toropov Y, GristTM, Mistretta CA. Time-resolved con-trast-enhanced imaging with isotropicresolution and broad coverage using anundersampled 3D projection trajectory.Magn Reson Med 2002; 48:297–305.

106. Sodickson DK, McKenzie CA, OhligerMA, Yeh EN, Price MD. Recent advancesin image reconstruction, coil sensitivitycalibration, and coil array design forSMASH and generalized parallel MRI.MAGMA 2002; 13:158–163.

107. Pandharipande PV, Krinsky GA, RusinekH, Lee VS. Perfusion imaging of the liverwith time-resolved, contrast-enhanced,whole-liver MRI (abstr). In: Proceedingsof the 11th Meeting of the InternationalSociety for Magnetic Resonance in Med-icine. Berkeley, Calif: International Soci-ety for Magnetic Resonance in Medi-cine, 2003; 1251.

108. Takeda M, Katayama Y, Tsutsui T, Kom-eyama T, Mizusawa T. Does gadolinium-diethylene triamine pentaacetic acid en-hanced MRI of kidney represent tissueconcentration of contrast media in thekidney? in vivo and in vitro study.Magn Reson Imaging 1994; 12:421–427.

109. Rusinek H, Lee VS, Johnson G. Optimaldose of Gd-DTPA in dynamic MR stud-ies. Magn Reson Med 2001; 46:312–316.

110. McKenzie CA, Prato FS, Thornhill RE,Drost DJ. T(1) fast acquisition relaxationmapping (T[1]-FARM): optimized dataacquisition. Magn Reson Imaging 2000;18:129–138.


Ra

dio

logy

perfusion imaging of the adiology liver: current challenges · entities of the liver affect blood...

Documents