| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gastrointestinal Imaging |
1 From the Department of Radiology, Division of Body Imaging, UCSD Medical Center San Diego, 200 W Arbor Dr, San Diego, CA 92103-8756. Received September 12, 2004; revision requested November 18; revision received January 24, 2005; accepted February 24. Address correspondence to C.B.S. (e-mail: csirlin{at}ucsd.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: The institutional review board approved the study and waived informed consent. The study complied with the Health Insurance Portability and Accountability Act. Ten patients (seven women, three men; mean age, 78 years; range, 31–78 years) with fatty infiltration surrounding hepatic veins and/or portal tracts were retrospectively identified by searching the abdominal imaging teaching file of an academic hospital. The patients' medical records were reviewed by one author. Computed tomographic (CT), magnetic resonance (MR), and ultrasonographic (US) imaging studies were reviewed by three radiologists in consensus. Fatty infiltration of the liver on CT images was defined as absolute attenuation less than 40 HU without mass effect and, if unenhanced images were available, as relative attenuation at least 10 HU less than that of the spleen; on gradient-echo MR images, it was defined as signal loss on opposed-phase images compared with in-phase images; and on US images, it was defined as hyperechogenicity of liver relative to kidney, ultrasound beam attenuation, and poor visualization of intrahepatic structures. Perivascular fatty infiltration of the liver was defined as a clear predisposition to fat accumulation around hepatic veins and/or portal tracts. For multiphase CT images, the contrast-to-noise ratio was calculated for comparison of spared liver with fatty liver in each imaging phase.
RESULTS: Fatty infiltration surrounded hepatic veins in three, portal tracts in five, and both hepatic veins and portal tracts in two patients. Six of the 10 patients had alcoholic cirrhosis, two reported regular alcohol consumption (one of whom had acquired immunodeficiency syndrome and hepatitis B), one was positive for human immunodeficiency virus, and one had no risk factors for fatty infiltration of the liver. In three of the 10 patients, fatty infiltration was misdiagnosed as vascular or neoplastic disease on initial CT images but was correctly diagnosed on MR images.
CONCLUSION: Perivascular fatty infiltration of the liver has imaging features that allow its recognition.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The images of the 50 patients were reviewed on a picture archiving and communication system (PACS) (Impax; Agfa Healthcare, Ridgefield Park, NJ) with a 2000 x 2000-pixel-resolution gray-scale monitor by three authors (C.B.S.; D.A.A., an abdominal imaging research fellow with 6 years of experience; and O.W.H., an abdominal imaging research fellow with 4 years of experience) in consensus and without knowledge of the patients' clinical data other than age and sex. All computed tomographic (CT), magnetic resonance (MR), and ultrasonographic (US) images were reviewed. Using predefined imaging criteria, the authors selected 10 patients (mean age, 47 years; range, 31–78 years) with perivascular fatty infiltration of the liver from the original 50. Seven were women (mean age, 51 years; range, 31–78 years), and three were men (mean age, 40 years; range, 35–45 years). In 36 of the original 50 patients, fatty infiltration of the liver was confirmed at the consensus reading, but the pattern was not unequivocally perivascular; these patients were excluded. In four of the 50 patients, the hepatic findings had a perivascular pattern but did not meet the imaging criteria for fatty infiltration of the liver; findings in these four patients (one woman, three men; mean age, 47 years; range, 24–64 years) are described separately.
Each of the 10 patients with imaging evidence of perivascular fatty infiltration underwent at least one examination at an urban academic hospital with a multi–detector row CT scanner with four detector rows. A total of 22 CT examinations were performed in the 10 patients (mean per patient, 2.3; range, 1–4). Four patients also underwent one MR imaging examination with a phased-array torso coil at 1.5 T during the study period, and one patient underwent two MR imaging examinations. Eight patients underwent a total of 27 US examinations: Two patients underwent one examination each, two underwent two each, two underwent three each, one underwent four, and one underwent 11.
Review of Clinical Data
One of the authors (O.W.H.) retrospectively reviewed the patients' medical records (including all outpatient clinic notes, discharge summaries, pathologic and surgical reports, and prospective imaging interpretations) 1 month after the retrospective image review. The following characteristics were documented: age, sex, indications for imaging examinations, prospective interpretation in dictated imaging reports, and relevant medical history. Relevant medical history included the following information, if available: clinical diagnoses; presence of risk factors for fatty infiltration of the liver (eg, diabetes mellitus, inborn error of metabolism, bowel resection, inflammatory bowel disease); history of drug abuse or regular alcohol consumption; medication history; history of chemotherapy; history of liver biopsy; results of pathologic analysis of liver; evidence of liver cirrhosis (clinical, imaging, and histologic data); serum triglyceride levels; and results of serologic testing for hepatitis B, hepatitis C, and human immunodeficiency viruses.
CT Technique
CT images were obtained on a multi–detector row scanner with four detector rows (LightSpeed; GE Medical Systems, Milwaukee, Wis) and with collimation of 2.5 mm (reconstructed section thickness, 5 mm). Two CT examinations were performed without any use of intravenous contrast material (with unenhanced CT). The other 20 CT examinations included the acquisition of contrast material–enhanced images after injection of 125 mL of an iodinated contrast medium (ioversol, Optiray 320; Mallinckrodt, St Louis, Mo) by using a power injector (CT 9000; Liebel-Flarsheim, Cincinnati, Ohio). Four of the 20 examinations involved scanning during a single phase (portal venous phase), and nine involved scanning during two phases (portal venous phase and equilibrium phase). One examination involved scanning in three phases: before contrast material injection, during the hepatic arterial phase, and during the portal venous phase. Six examinations involved scanning in four phases: before contrast material injection, in the hepatic arterial phase, in the portal venous phase, and in the equilibrium phase. Overall, six patients (patients 1, 5, 6, 7, 9, and 10) underwent at least one CT examination that included scanning in the unenhanced phase. Timing was achieved by using bolus tracking software (SmartPrep; GE Medical Systems). In general, the hepatic arterial phase images were obtained at approximately 30 seconds, the portal venous phase images at 60–70 seconds, and the equilibrium phase images at 3–5 minutes after contrast material injection, but exact delays were not recorded. For most contrast-enhanced examinations, intravenous contrast material was administered at a rate of 3–4 mL/sec. For examinations that included hepatic arterial phase imaging, contrast material was injected at a rate of 4–5 mL/sec.
MR Imaging Technique
MR images were obtained with a 1.5-T magnet (Symphony; Siemens Medical Solutions, Erlangen, Germany) and with a torso phased-array surface coil positioned to cover the patient's upper abdomen. Although individual protocols varied, all examinations included at least one set of 8–10-mm-thick T1-weighted spoiled gradient-echo in-phase and opposed-phase images. For in-phase images, an echo time of 4.76 msec was used; for opposed-phase images, an echo time of 2.65 msec was used. For both in-phase and opposed-phase image acquisitions, the repetition time was 204 msec, the flip angle was 30°, the matrix was 128 x 256, and the intersection gap was 0%–25%.
US Imaging Technique
Each US examination was performed after the patient fasted for at least 10 hours, by one of 10 certified sonographers (1–25 years of experience) using one of four US systems (HDI 3000, Advanced Technologies Laboratories, Bothell, Wash; Elegra or Antares, Siemens Medical Solutions; or Logic9, GE Medical Systems). Parameters such as transducer type and frequency, focal zone placement, gain, and tissue harmonics application were optimized by the sonographer on a case-by-case basis; the chosen parameters were not recorded for purposes of this study. Color Doppler images were obtained routinely to enhance vessel depiction. Representative images of the liver (usually, at least six transverse and six longitudinal images) were saved to the PACS. These were then reviewed on a 19-inch 1024 x 1024-pixel color monitor (MWD 421; Barco Display Systems, Kortrijk, Belgium), as well as on the higher-resolution gray-scale monitor used for the CT and MR examinations.
Imaging Criteria of Fatty Infiltration
On unenhanced CT images, fatty infiltration of the liver was defined as absolute attenuation of less than 40 HU and relative attenuation at least 10 HU less than that of spleen without mass effect on the appearance of vessels or other intrahepatic structures (20–24). If unenhanced CT images were not available, attenuation was measured on contrast-enhanced images. The same absolute attenuation threshold (40 HU) was applied to unenhanced and contrast-enhanced CT images; this conservative decision was made because histologic proof was not routinely available. The relative attenuation threshold (10 HU less than the attenuation of spleen) was not applied to contrast-enhanced CT images. Visually hypoattenuating regions in which measured attenuation was more than 40 HU did not qualify as fatty infiltration of the liver, except in one patient in whom the hypoattenuating region in the liver had a measured attenuation of 48 HU on contrast-enhanced CT images (unenhanced CT was not performed). In this patient, MR imaging, performed 2 weeks after CT, confirmed fatty infiltration with the same distribution in the liver. Hypoattenuating regions that subjectively were considered to be consistent or possibly consistent with hepatic or periportal edema in the final consensus reading also did not qualify as fatty infiltration. Attenuation values were measured as the mean number of pixels in user-defined regions of interest (ROIs) on CT images by using the PACS software.
At visual inspection of MR images, fatty infiltration of the liver was defined as unequivocal signal intensity reduction in liver parenchyma relative to signal intensity in spleen and renal cortex on opposed-phase images compared with in-phase images (25–28). Quantification of signal loss on MR images was not necessary, because signal loss was obvious in each case.
On US images, fatty infiltration of the liver was defined as hyperechogenicity of liver in comparison with kidney, attenuation of the ultrasound beam by liver, and subjectively poor visualization of intrahepatic structures (12,29).
Perivascular fatty infiltration of the liver was defined, in the final consensus reading of CT and MR images, as a clear and unequivocal predisposition of liver parenchyma to fatty infiltration around visible intrahepatic vessels. In addition, liver parenchyma remote from visible vessels had to be either completely or almost completely spared. US images were not used to define perivascular fatty infiltration of the liver, because the distribution was not reliably depicted sonographically. The distribution, extent, and delineation of the perivascular fatty infiltration of the liver were further defined as follows: (a) Distribution was defined as perivenous if fatty infiltration surrounded one or more hepatic veins, periportal if it surrounded one or more portal tracts, or combined (both perivenous and periportal) if it surrounded both the hepatic vein and the portal tract; (b) extent was defined as bilobar if perivascular fatty infiltration of the liver involved both lobes, as lobar if it involved one lobe, and as focal if it involved a segment or subsegment; and (c) delineation was described either as well defined (tight) or as ill defined (loose), depending on how closely the fatty infiltration cloaked the hepatic vessels. Fatty infiltration with well-defined or tight delineation formed glovelike, parallel, uniformly thin (5–15 mm) halos around the vessels. Depending on vessel orientation, halos were tramlike (vessel segments parallel to scanning plane) or round (vessel segments perpendicular to scanning plane). In the ill-defined form, halos were fan shaped or wedge shaped and variable in thickness (5–35 mm).
Image Analysis
Qualitative analysis.—The image review was performed by three authors in consensus, as noted above. If various imaging modalities had been used in a patient, all of the patient's images were evaluated in the same session. CT images were initially reviewed with a window level and window width of 70 and 200 HU for unenhanced CT images and 50 and 400 HU for contrast-enhanced CT images. The observers were free to adjust the window width and window level at will, however, to improve the subjective conspicuity of the findings. The morphologic structure of perivascular fatty infiltration of the liver was evaluated for vessel segments parallel and perpendicular to the imaging plane and was compared between CT and MR images (if images from both modalities were available for a given patient). The images were inspected for any potential mass effect exerted by a hypoattenuating area on the appearance of vessels or other hepatic structures. The observers were free to reformat CT and/or MR images by using the PACS three-dimensional display software.
The subjective conspicuity of perivascular fat on CT images was ranked, in consensus, for each examination in patients who underwent more than one CT examination and for each phase in examinations in which scanning was performed in more than one phase. The CT image on which the findings subjectively were most conspicuous (image with the highest ranking for perivascular fat) was chosen as the index CT image. The pattern of fatty infiltration of the liver on the index CT image was then classified according to the criteria specified earlier.
Findings were reviewed to confirm the presence (on MR and US images) and pattern (on MR images) of fatty infiltration of the liver and to compare these MR and US features subjectively with CT findings.
Quantitative analysis.—The thickness of the perivascular zones of fatty infiltration of the liver was measured (to the nearest millimeter) with digital cursors by using the PACS software.
User-defined ROIs were placed in the same section, in the hypoattenuating perivascular area (mean number of pixels per ROI, 220; range, 130–400), in the most normal-appearing liver region (mean number of pixels per ROI, 800; range, 600–1200), in the air outside the patient's body (mean number of pixels per ROI, 600; range, 500–800 pixels), and in the spleen (mean number of pixels per ROI, 800; range, 600–1200 pixels). The size of each ROI was chosen to minimize partial-volume averaging artifacts. ROIs were placed on the most subjectively representative transverse image. Thus, ROI sizes and locations were based on imaging findings on a patient-by-patient basis and were not kept constant between patients. When CT scanning was performed in several phases for a given examination in a given patient, attenuation was determined in every phase, and ROI size and placement were kept constant. With use of the PACS software, the mean attenuation per pixel in each ROI was recorded, as was the standard deviation of the mean attenuation per pixel in air. The contrast-to-noise ratio (CNR) was calculated for comparison of spared liver with fatty tissue in each imaging phase by using the following equation: CNR = (SLmean – FTmean)/ASD, where SLmean is the mean attenuation of spared liver, FTmean is the mean attenuation of fatty tissue, and ASD is the standard deviation of the mean air attenuation.
The difference between attenuation in liver with fatty infiltration and attenuation in the spleen also was calculated on the basis of unenhanced CT images.
For each CT examination, the phase (unenhanced phase, hepatic arterial phase, portal venous phase, or equilibrium phase) in which the attenuation of fatty infiltration in the liver was lowest was chosen as the index phase.
Data Analysis
Patterns of fatty infiltration of the liver were analyzed descriptively in all patients.
Images from the six four-phase CT examinations (even if they were not the index phase images) were further analyzed. Data from comparisons between subjects were pooled, and the mean CNR was calculated for each phase. Rankings for subjective conspicuity were analyzed descriptively.
The time course of perivascular fatty infiltration of the liver was analyzed descriptively by comparing subjective conspicuity rankings and quantitative CNR measurements in similar enhancement phases of successive follow-up CT or MR examinations. US examinations were not used for assessment of temporal changes.
As this was a preliminary study involving only preliminary observations, hypothesis testing and statistical analyses were not performed.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
Indications for the index CT examinations were the following: abdominal pain (patient 1), surveillance for breast carcinoma (patient 2, with no chemotherapy), upper gastrointestinal tract bleeding (patients 3 and 9), acute pancreatitis (patient 4), jaundice and history of Kaposi sarcoma (patient 5, with no chemotherapy), pancreatic cancer (patient 6, with no chemotherapy), surveillance imaging for hepatocellular carcinoma (patients 7 and 10), and perforated peptic ulcer (patient 8).
Retrospective Imaging Findings
Of the 10 patients, three (patients 1–3) had perivenous fatty infiltration, five (patients 4–8) had periportal fatty infiltration, and two (patients 9 and 10) had combined (perivenous and periportal) fatty infiltration of the liver. Fatty infiltration of the liver was confirmed with histologic analysis in two patients (patients 7 and 8), with chemical shift gradient-echo MR imaging in four patients (patients 1, 2, 4, and 9), and with US in eight patients (patients 1–5 and 8–10). Images from three consecutive CT examinations in one patient (patient 6) that met the imaging criteria for a finding of perivascular fatty infiltration of the liver, as outlined earlier, served as the reference standard. The attenuation of perivascular fatty infiltration on the index phase images (the unenhanced phase image in six patients, the portal venous phase image in three patients, and the equilibrium phase image in one patient) from the index CT examination ranged from –17 to 48 HU (mean, 16 HU ± 22). The CNR for comparison of the attenuation of spared liver with that of fatty liver on the same images ranged from 2.3 to 12.3 (mean, 5.9 ± 4.9). One patient (patient 2) had perivascular fatty infiltration of the liver with attenuation greater than 40 HU on portal venous phase images (unenhanced CT images were not obtained). In this patient, perivascular fatty infiltration of the liver was confirmed at MR imaging 2 weeks later. In the six patients for whom the index CT examination included an unenhanced scanning phase, the mean attenuation of unenhanced perivascular fatty infiltration of the liver was 37 HU ± 20 (range, 12–64 HU) less than the mean attenuation of unenhanced spleen.
Perivascular fatty infiltration of the liver had a tramlike configuration for vessel segments parallel to the imaging plane (Fig 1, A; Fig 2, A and B) and a ringlike or round configuration for vessel segments perpendicular to the imaging plane (Fig 1, B; Fig 2, C and D). Although the ringlike or round configuration superficially mimicked a focal lesion, the perivascular pattern became obvious as the reviewer scrolled through the images on the PACS workstation or reformatted the images (Fig 3). Regardless of orientation, distribution, extent, or delineation, the fatty areas caused no mass effect on the appearance of vessels or other hepatic structures. Appearances were virtually identical on CT and MR images, with consideration given to potential differences in technique and changes over time (Figs 4, 5).
|
|
|
|
|
|
Periportal fatty infiltration.—One of the five patients (patient 4) had 10–15-mm ill-defined zones of perivascular fatty infiltration around the left portal vein and its visible branches, with sparing of the liver parenchyma around the hepatic and right portal veins. The second patient (patient 5) with a loose and ill-defined periportal distribution had 10–25-mm halos of severe perivascular fatty infiltration of the liver surrounding all visible segments of the portal tracts in both lobes. In the left lobe, the halos were partially confluent and mimicked a diffuse pattern. In the right lobe, however, the parenchyma surrounding the hepatic veins and along the liver periphery was conspicuously spared (Fig 7). In the third case (patient 6), perivascular fatty infiltration of the liver was well defined, and it tightly cloaked two left-lobe lateral segments of the portal tract; the rest of the liver was completely spared (Fig 8). The fourth patient (patient 7) had well-defined halos of perivascular fatty infiltration that tightly cloaked portal veins of both lobes. In the fifth patient (patient 8), ill-defined zones of perivascular fatty infiltration up to 35 mm thick surrounded the portal tracts in both lobes. Thus, of the five patients with periportal fatty infiltration of the liver, three had a pattern of infiltration that was ill defined and bilobar, one had a pattern that was ill defined and lobar, and one had a pattern that was well defined and focal.
|
|
Time Course of Imaging Findings
In all 10 patients, images from more than one imaging examination were available for assessment of the course of fatty infiltration over time. Five general temporal patterns were observed.
Complete resolution.—In one patient with periportal fatty infiltration (patient 7), perivascular fatty infiltration had completely resolved at the follow-up imaging examination performed with MR imaging 24 weeks after the initial imaging examination. In two patients, one with perivenous fatty infiltration (patient 3) and one with periportal fatty infiltration (patient 5), progressive improvement was observed on images from one and two successive CT examinations, respectively, with complete resolution by the last follow-up examination (at 14 and 36 weeks, respectively). For patient 5, the CNR of spared liver to fatty liver decreased from 9 at the first CT examination to 0 at the last CT examination. For patient 3, a comparison of CNR values was not possible because the follow-up CT images were acquired in a different phase.
Improvement with incomplete resolution.—Three patients, two with periportal (patients 6 and 8) and one with combined (patient 10) perivascular fatty infiltration of the liver, showed progressive improvement on images from two or three successive CT examinations, with incomplete resolution by the last follow-up examination (at 54, 32, and 28 weeks, respectively). The CNR decreased from 7 to 1 (Fig 8), 17 to 3, and 16 to 8, respectively.
Stable infiltration.—Two patients, both with perivenous fatty infiltration (patients 1 and 2), had subjectively stable imaging findings at one or two successive follow-up examinations, with the final follow-up examination at 3 and 26 weeks, respectively. A comparison of CNR values was not possible, because follow-up examinations were performed with MR imaging.
Progression from perivascular to diffuse infiltration.—In one patient with periportal fatty infiltration (patient 4), the perivascular component of infiltration remained stable at 10-week follow-up CT, while infiltration increased in the remainder of the liver, so that the perivascular pattern was less conspicuous at follow-up. The CNR, which was initially 11, decreased to 5 at follow-up.
Progression of perivascular infiltration.—In one patient with combined fatty infiltration of the liver (patient 9), the perivascular fat became progressively more obvious over two successive imaging examinations (one performed with CT and one with MR imaging) in a span of 33 weeks. CNR at the initial CT examination was 3, and at follow-up CT, 7.
Effect of Contrast Phase on Conspicuity and CNR
Six patients underwent at least one four-phase CT examination (patients 1, 5, 6, 7, 9, and 10). In the four-phase CT scans, perivascular fatty infiltration of the liver subjectively was most obvious on the hepatic arterial phase image in two patients, the portal venous phase image in one patient, and the equilibrium phase image in one patient. In one patient, the depiction of infiltration on the hepatic arterial phase image and that on the portal venous phase image were considered equal to one another and superior to that on images from the other phases (Fig 7); in another patient, the portal venous phase and equilibrium phase images were considered equal with regard to the quality of depiction of infiltration (Fig 4). Perivascular fatty infiltration was not considered most conspicuous on unenhanced CT images in any of the patients.
The mean CNR for comparison of spared liver with fatty liver was 4.9 ± 3.8 (unenhanced CT), 5.0 ± 1.3 (hepatic arterial phase), 6.9 ± 3.0 (portal venous phase), and 4.2 ± 1.7 (equilibrium phase).
Prospective Interpretations
Evidence of fatty infiltration of the liver was misinterpreted on initial CT images in three of the 10 patients. In one patient (patient 1), perivenous fatty infiltration was mistaken for Budd-Chiari syndrome (Figs 4, 5); in another (patient 2), it was mistaken for metastatic disease (Figs 2, 6). In a third patient (patient 9), combined perivascular fatty infiltration was mistaken for multifocal hepatocellular carcinoma. In these three patients, the correct diagnosis was made on the basis of MR images acquired 2 to 3 weeks later, which showed definitive signal loss on opposed-phase images compared with in-phase images.
In the other seven of 10 patients, fatty infiltration of the liver was diagnosed correctly on CT images at the prospective interpretation, although it is unclear from the dictated reports whether the perivascular pattern was appreciated.
US images of all eight patients were correctly interpreted prospectively as showing fatty infiltration of the liver, but, in addition, findings in two patients were falsely interpreted as suspicious for superimposed neoplasia, and those in another patient, for hepatic vein thrombosis. Findings in two other patients were considered indeterminate for neoplasia. Neoplasia in these patients was ruled out on the basis of clinical follow-up in conjunction with findings at follow-up CT and MR imaging.
Perivascular Hypoattenuation That Did Not Meet Criteria for Fatty Infiltration
Periportal (three patients) and perivenous (one patient) halos of hypoattenuation were noted that did not meet the criteria for perivascular fatty infiltration of the liver because the contrast-enhanced liver attenuation exceeded 40 HU in all patients and because unenhanced CT was not performed.
One of these patients (a 64-year-old man with a distant history of alcohol abuse) had images that showed bilobar well-defined (10–12-mm) periportal halos (70 HU) that were distinctly lower in attenuation than the surrounding liver tissue (145 HU) on the portal venous phase images. In the equilibrium phase, the periportal tissue was hyperattenuated (102 HU) when compared with the spared liver (92 HU). The findings were stable at 4-week follow-up. In another patient (a 24-year-old man with biopsy-proved cirrhosis due to chronic hepatitis B), fan-shaped bilobar areas of periportal hypoattenuation were visible only on the initial two-phase CT images (the portal venous phase and equilibrium phase images) but not on the follow-up images from two contrast-enhanced CT examinations and one MR imaging examination over 74 weeks. The periportal halos were up to 25 mm thick, were partially confluent, and had attenuation of 80 HU on the portal venous phase images, whereas the attenuation of the spared liver was 140 HU. In the third patient (a 64-year-old woman with recurrent cervical carcinoma metastases to the liver after chemotherapy and radiation therapy), faint fan-shaped bilobar perivenous halos with attenuation of 160 HU (attenuation of spared liver, 190 HU) were noted on two-phase CT images. In the fourth patient (a 34-year-old man with colon cancer), fan-shaped confluent periportal halos up to 40 mm thick and with attenuation of 60 HU (attenuation of surrounding spared liver, 90 HU) were seen on the portal venous phase CT images but imperceptible on equilibrium phase images. In all patients, the cause of the areas of periportal hypoattenuation was unclear; it is possible that one or more of the patients may have had perivascular fatty infiltration of the liver that did not meet imaging criteria on the basis of contrast-enhanced CT images.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Findings in 10 patients with perivascular fatty infiltration of the liver on CT images are described. The perivascular pattern was characterized by hypoattenuating halos that surrounded the hepatic veins, the portal vessels, or both. The configuration was tramlike for vessel segments in the imaging plane and ringlike or round for vessel segments perpendicular to the imaging plane. The round halos were prospectively mistaken for metastatic disease in one patient and for hepatocellular carcinoma in another. In our retrospective consensus review, however, the complete absence of mass effect, the presence of multiple similar hypoattenuating zones surrounding vessels, and the tubular appearance on reformatted images permitted easy diagnosis. In a third patient, perivascular fatty infiltration of the liver was misinterpreted as Budd-Chiari syndrome. In all three patients, the correct diagnosis was confirmed at chemical shift MR imaging, which showed the characteristic features of perivascular fatty infiltration of the liver.
Because histologic confirmation of fatty infiltration of the liver was available in only two patients, we relied on imaging diagnosis. Based on previously published data (20–24), two criteria were used for a finding of fatty infiltration on unenhanced CT images: relative attenuation at least 10 HU less than that of spleen, and absolute attenuation less than 40 HU. Although other pathologic processes (eg, ischemic metastasis or pyogenic abscess) may have attenuation less than 40 HU, other entities were easily excluded on the basis of imaging and clinical findings. A reliable threshold attenuation for diagnosis of fatty infiltration of the liver on contrast-enhanced CT images has not been established (35,36). Therefore, we made a conservative decision to use the same absolute threshold criterion for contrast-enhanced CT images as for unenhanced CT images. Fatty infiltration of the liver documented on CT images was confirmed at biopsy in two patients, at chemical shift gradient-echo MR imaging in four patients, and at US in eight patients (findings in some patients were confirmed at more than one examination). In one patient without confirmation at biopsy, MR imaging, or US, findings at three consecutive CT examinations served as the reference standard.
Although the CT diagnosis of fatty infiltration of the liver was based on objective criteria, the perivascular pattern was identified subjectively. For this reason, a consensus of three radiologists was required. The pattern was described by its distribution (perivenous, periportal, or both), extent (bilobar, lobar, or focal and segmental), and delineation (well or ill defined). Sample sizes were too small to determine whether there were etiologic or other clinical differences between the different types. In general, well-defined perivascular fatty infiltration of the liver (five patients) had a dramatic imaging appearance, with glovelike areas of sharply delineated fatty infiltration of the liver that tightly cloaked vessels and that were separated by areas of complete or near-complete sparing. This appearance was clearly different from previously reported descriptions of fatty infiltration of the liver. Ill-defined or loose cloaking in five patients manifested as broad and vaguely delineated zones of fatty infiltration that superficially mimicked the commonly described diffuse pattern. Careful review of the images, however, demonstrated the unambiguous predisposition of perivascular zones to fatty infiltration.
In the six patients who underwent four-phase CT, the conspicuity of perivascular fatty infiltration of the liver consistently was ranked lowest by the three readers on unenhanced CT images; conspicuity was ranked highest on each of the other three phases on at least one occasion. As opposed to subjective conspicuity, the mean CNR was similar in all four phases, and the unenhanced CT images had the highest CNR in two of the six patients. One explanation for the discrepancy between subjective and quantitative findings is that the perivascular pattern was difficult to appreciate without enhancing vessels. There was no phase in which the perivascular fatty infiltration of the liver was completely obscured by parenchymal enhancement.
Temporal changes in imaging findings were assessed subjectively in all 10 patients and were corroborated quantitatively in the six patients with follow-up CT scans obtained in similar phases. As expected, numerous temporal patterns were observed, including stability, progression, and resolution. Importantly, although the severity of findings changed in some patients, the pattern and distribution of perivascular fatty infiltration of the liver was constant; this reproducibility validates our assumption that the findings represented a true hepatic pathologic process rather than artifacts. However, imaging follow-up was determined by the referring physician and was not prospectively planned. Thus, the type of follow-up, the interval between studies, and the total follow-up period were not uniform among patients. This fact limits our ability to analyze the time course of the fatty changes quantitatively.
When differences in technique and time course are considered, CT and MR findings were highly congruent. US consistently confirmed the presence of fatty infiltration of the liver but did not reliably depict the infiltration pattern or, in five of the eight patients, exclude superimposed neoplasia or other important pathologic processes.
The differential diagnosis of hypoattenuating perivascular halos in the liver is broad and includes, in published reviews and abstracts, edema, hemorrhage, perfusion abnormalities, fibrosis, lymphatic dilatation, extramedullary hematopoiesis, and neoplasia (lymphoma, neurofibroma) (37,38). In the patients in our study, these possibilities can be reliably excluded on the basis of the constellation of multimodality imaging findings, time course of findings, and clinical data. Moreover, the most common differential process, which is periportal edema (postulated to be due in part to looseness of the tissues around the portal triads in young individuals), becomes less frequent with increasing age and would be unusual in a patient sample with a mean age of 47 years.
An important limitation of our study was that the sample selected for inclusion was derived from a teaching file of interesting cases, which is a highly biased population. Consequently, the prevalence of the pattern of perivascular fatty infiltration of the liver cannot be determined from the data of this study. Another limitation was the retrospective and descriptive rather than hypothesis-driven nature of the study. Analysis of the distribution, extent, and delineation of perivascular fatty infiltration of the liver was based on the consensus decision of three readers. Because the reading was made in consensus, inter- and intraobserver variability were not assessed. Histologic proof of fatty infiltration of the liver was available in only two patients. Moreover, even in these two patients, biopsy results could not confirm the distribution, extent, and delineation of perivascular fatty infiltration of the liver, because the liver was sampled randomly. Statistical analysis was not performed. The pathophysiology of perivascular fatty infiltration of the liver also was not addressed by our study. Preferential accumulation of fat in perivascular zones may, in theory, be related to zonal differences in hepatocyte function or perfusion. Interestingly, eight of 10 patients had regular alcohol consumption as a major risk factor for fatty infiltration of the liver. Further study is required to determine whether a perivascular accumulation is more characteristic of alcohol-induced fatty infiltration than of infiltration in patients with other predisposing factors. It also is possible that perivascular fatty infiltration of the liver merely represents one end of the diffuse fatty infiltration spectrum, at least in patients with ill-defined delineation and in whom the pattern superficially resembles the diffuse type.
In conclusion, fatty infiltration of the liver can manifest in various forms. We have described a perivascular pattern of infiltration that, to our knowledge, has not previously been reported. Radiologists should be aware of this pattern to avoid potential diagnostic pitfalls. On the basis of our preliminary observations in conjunction with previously published data, CT attenuation of less than 40 HU, or unenhanced CT attenuation at least 10 HU less than that of spleen, and the absence of mass effect in the perivascular zones, suggests the diagnosis. In difficult cases, MR imaging with chemical shift sequences can be performed. US may help to confirm fatty infiltration of the liver but may be insufficient to confirm the perivascular pattern or to exclude more ominous disease processes.
|
FOOTNOTES |
|---|
Abbreviations: CNR = contrast-to-noise ratio • PACS = picture archiving and communication system • ROI = region of interest
2 Current address: Department of Diagnostic Radiology, University Hospital of Regensburg, Regensburg, Germany 
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, O.W.H., C.B.S.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, O.W.H., D.A.A., G.C.; clinical studies, all authors; manuscript editing, all authors
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Macari, R. Yeretsian, and J. Babb Assessment of Low Signal Adjacent to the Falciform Ligament on Contrast-Enhanced MRI Am. J. Roentgenol., December 1, 2007; 189(6): 1443 - 1448. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. W. Hamer, D. A. Aguirre, G. Casola, J. E. Lavine, M. Woenckhaus, and C. B. Sirlin Fatty Liver: Imaging Patterns and Pitfalls RadioGraphics, November 1, 2006; 26(6): 1637 - 1653. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||
