| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Gastrointestinal Imaging |
1 From the Department of Medical Imaging, University Health Network and Mount Sinai Hospital, University of Toronto, Toronto, Ontario, Canada. Received August 1, 2003; revision requested October 14; final revision received April 24, 2004; accepted June 16. Address correspondence to U.M., Department of Nuclear Medicine, Tel-Aviv Sourasky Medical Center, 6 Weizman St, Tel-Aviv, Israel 64239 (e-mail: umetser@tasmc.health.gov.il).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: The institutional ethics review board approved the study and waived the requirement for informed consent. Sixty multiphasic hepatic CT examinations were performed in 39 immunocompromised patients who fulfilled the criteria for having probable or proved fungal liver infection. The detection and conspicuity of focal liver lesions were assessed on scans obtained during each CT phase. The lesion enhancement pattern was determined, and, accordingly, lesions were stratified into two groups: lesions suggestive of infection (with ring enhancement patterns or high attenuation) and nonspecific hypoattenuating lesions. Statistical analyses were performed by using logistic regression with generalized estimating equations.
RESULTS: A total of 536 liver lesions detected at 36 CT examinations with results positive for fungal infection were assessed. All 36 (100%) examinations yielded positive results during the arterial phase, whereas 25 (69%) of them yielded positive results during the portal venous phase (P < .001). At lesion-by-lesion analysis, the arterial phase scans depicted significantly more lesions (483 of 536 [90%]) than the portal venous phase (329 of 536 [61%]) and nonenhanced (265 of 465 [57%]) scans (P < .001 for both comparisons). In addition, on arterial phase scans, 386 of 483 lesions, as compared with 134 of 329 lesions on portal venous phase scans (P < .001), were judged to have an enhancement pattern suggestive of infection. The CT phases did not differ significantly in terms of the conspicuity of detected lesions.
CONCLUSION: In patients suspected of having hepatic fungal infection, arterial phase CT depicts significantly more hepatic lesions than does CT performed during the other phases, and it reveals more lesions with enhancement patterns suggestive of infection. Arterial phase CT should be performed in addition to portal venous phase CT in patients suspected of having hepatic fungal infection.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
To our knowledge, the studies on the performance of CT in the detection of fungal microabscesses in the liver that have been published to date pertain to portal venous phase CT of the liver. Thus, the purpose of this study was to retrospectively assess multiphasic (nonenhanced, arterial phase, and portal venous phase) CT of the liver for the depiction of hepatic fungal infection in immunocompromised patients.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Because this study was performed to evaluate the performance of multiphasic CT in the detection of hepatic fungal microabscesses, positive imaging findings, which are an additional criterion for the diagnosis of probable fungal liver infection according to the EORTC-MSG guidelines, were not considered inclusive criteria. However, we elected to use CT-depicted proved response to antifungal therapy, along with clinical response, as an ancillary sign in patients whose cultures were negative for fungi.
Patient Population
This retrospective study was approved by our institutional ethics review board, which waived the requirement for informed consent. Our search of a computerized clinical database of records for a 55-month period (January 1998 to August 2002) that was available for retrospective review revealed 481 patients with leukemia. Sixty-eight of these patients underwent biphasic or triphasic liver CT because they had a neutropenic fever that was nonresponsive to broad-spectrum antibiotics and were clinically suspected of having hepatosplenic fungal infection. Thirty-nine of these patients fulfilled our study inclusion criteria—specifically, proved (n = 4) or probable (n = 35) hepatic fungal infection—and comprised our study population. Patient ages ranged from 17 to 77 years (median, 55 years). Eighteen patients were male (age range, 21–67 years; median age, 48 years), and 21 were female (age range, 17–77 years; median age, 57 years).
Twenty-four patients underwent one CT examination, and 15 underwent more than one. Thirteen patients underwent CT scanning for two separate episodes of neutropenic fever. An average of 1.54 CT examinations per patient (median, 1; range, 1–5) were performed. Overall, the results of 60 liver CT examinations—50 triphasic (nonenhanced, arterial phase, and portal venous phase) and 10 biphasic (arterial and portal venous phases only) examinations—were analyzed.
CT Examinations
CT scans were obtained by using a four–detector row helical CT system (QX/i LightSpeed; GE Medical Systems, Milwaukee, Wis) with the following parameters: 120 kV, 230–330 mA, a table speed of 7.5 mm per rotation, high-quality mode, and a pitch of 3:1. The scans were reconstructed at collimations of 5 mm with 50% overlap. With use of a power injector (Medrad, Indianola, Pa), nonionic intravenous contrast material (iohexol [Omnipaque 300], 30 mg of iodine per milliliter; Amersham Health, Buckinghamshire, England) was administered at a dose of 2 mL per kilogram of body weight (up to a maximum of 200 mL) at a rate of 3 mL/sec, with a 25-second delay for the arterial phase and a 60-second delay for the portal venous phase. For triphasic scanning, a preliminary nonenhanced CT examination of the liver was performed by using the same imaging parameters.
Scan Interpretation and Analysis
First panel analysis: blind reading.—The objective of the first panel analysis was to test the overall performance of each of the three CT phases in the detection of liver lesions. For this purpose, each CT examination was divided into separate phases: nonenhanced, arterial, and portal venous phases. The scans obtained during all phases of the examination were randomized and presented to a panel of two experienced abdominal radiologists with 8 (M.A.H.) and 5 (M.D.M.) years of experience reading abdominal CT images; interpretative decisions were made in consensus. To minimize recall bias, the different phase scans obtained in each patient were read at least 2 weeks apart.
The scans were reviewed in stack mode on a picture archiving and communication system workstation (Merge eFilm, Milwaukee, Wis). The technical parameters used and patient identifiers were hidden from the reviewers at the time of interpretation. Although the readers knew the purpose of the study, they were blinded to the findings on the other phase scans and to the patients’ clinical and/or laboratory data. The radiologists were allowed to choose the window width and window level for each CT phase, as they saw fit.
On separate standard questionnaires pertaining to each CT phase, the panel was asked to note whether the phase yielded findings positive or negative for focal liver lesions. If focal splenic lesions were present, this was also noted. The total number of focal liver lesions was recorded. For the CT examinations revealing 25 or fewer lesions, the specific number of lesions was given. For patient cases in which more than 25 lesions were depicted, the panel had to determine the appropriate range of the number of lesions: 26–50 lesions or more than 50 lesions. For cases in which more than 13 lesions were identified, the readers performed a complete analysis of the three largest lesions, the three smallest lesions, and at least eight additional lesions (at least one from each liver segment, if present). For each lesion, the following parameters were recorded:
1. Size: Lesion size was determined on the basis of the largest single dimensional measurement on the scan section on which the lesion was largest.
2. Conspicuity score: A score of 0 meant no lesions were identified; 1, the finding probably was not a lesion; 2, the finding probably was a lesion or an ill-defined lesion; and 3, the finding definitely was a lesion.
3. Lesion enhancement pattern (for arterial and portal venous phases only): A type 1 lesion was hypoattenuating and either subcentimeter in diameter (type 1a) or at least 1 cm in diameter (type 1b). A type 2 lesion had a hypoattenuating center and a hyperattenuating rim, the thickness of which was equal to or less than the radius of the hypoattenuating center (type 2a) or greater than the radius of the hypoattenuating center (type 2b). A type 3 lesion was any lesion with associated transient hepatic enhancement, which was defined as an adjacent wedge-shaped region of increased enhancement. A type 4 lesion was any hyperattenuating lesion—that is, one that appeared brighter than the surrounding liver on the scan.
The study coordinator (U.M.), who was not involved in the scan interpretations, summarized the responses recorded on the randomized questionnaires that were completed for the first panel analysis and determined whether a given examination yielded results that were positive or negative for fungal liver infection. An examination result was considered to be positive for fungal liver infection when the readers judged at least one of the CT phases to have yielded positive findings. A positive phase result was defined as one in which at least three type 1 lesions were seen or at least one type 2, 3, or 4 lesion was seen (discussed in following text).
The study coordinator also reviewed the available abdominal CT scans obtained in each patient and excluded lesions with CT characteristics of cysts (low attenuation, no enhancement) or hemangiomas (peripheral nodular enhancement with centripetal filling on delayed images) and lesions proved to be cysts or hemangiomas at other imaging examinations (eg, US findings of cysts or magnetic resonance [MR] imaging findings of cysts or hemangiomas). For 25 patients, previously obtained or follow-up CT scans (obtained 3–25 months before or after current acquisitions; mean, 6 months) were available. Because the findings of available studies on serial imaging of fungal liver infections show that fungal abscesses are evolving lesions (10–12), the liver lesions that were stable in size and appearance on the previously obtained or follow-up scans also were omitted. Overall, 27 lesions were excluded from the final study analysis: 10 hemangiomas (two depicted at MR imaging and eight with typical CT features), one case of focal nodular hyperplasia, five cysts (three depicted at US and two depicted at MR imaging), and 11 stable nonspecific hypoattenuating lesions.
Second panel analysis: correlative reading.—The study coordinator reviewed the findings seen during all of the phases together to ensure that all lesions mentioned in the blinded reading matched. At examinations in which not all lesions were analyzed, a certain lesion may have been analyzed on scans obtained during only one or two phases. Lesions that during the initial reading were evaluated on scans obtained during only one or two phases were evaluated on scans obtained during the additional phase or phases in a second reading by the study coordinator and the two readers (M.A.H., M.D.M.).
Data and Statistical Analyses
The rates of diagnosis of fungal liver infection at each CT phase were calculated. This diagnosis rate was defined as the number of lesions diagnosed as positive for infection at each phase (ie, on scans obtained during that phase) divided by the total number of lesions evaluated during that phase. The variances and corresponding 95% confidence intervals were estimated by using logistic regression models for each phase with an overdispersion parameter. The confidence interval for the arterial phase was based on exact methods because all liver infections were detected correctly and an asymptotic model could not be fit. The diagnosis rates for the three phases were tested for differences by using exact conditional logistic regression, with each reference-standard liver infection as a stratum. The variation between livers was thus treated as a nuisance parameter. A binary variable indicating whether the fungal infection was correctly diagnosed was the outcome variable, and the CT phase was the independent variable.
The detection rate for each phase was also calculated on a lesion-by-lesion basis. The numerator was the number of lesions seen during a given phase (ie, seen on scans obtained during a given phase), and the denominator was the total number of lesions assessed during all of the phases. For the nonenhanced CT phase, which was not used in all cases, the denominator was the total number of lesions assessed during all of the phases for those cases in which nonenhanced CT was performed. To address any bias introduced by the second reading performed to match lesions, we repeated the statistical analysis to determine the lesion detection rate without a second reading. Detection rates were compared between phases by using generalized linear regression with a binomial error distribution, a logit link function, and generalized estimating equations to account for the correlation due to multiple lesions per patient. Responses from individual patients were assumed to be equally correlated. Responses from different patients were assumed to be statistically independent. Ninety-five percent confidence intervals were calculated by using the variances from the generalized estimating equations modeling. The value of the dependent variable was 0 if a lesion was not detected during a given phase and 1 if a lesion was detected; the independent variable was a categorical variable indicating the phase type. Testing whether the independent variable had a coefficient significantly different from 0 was considered to be a test of whether the detection rates were different.
The frequencies of the different enhancement patterns on the CT scans obtained during each of the contrast material–enhanced phases also were determined. The enhancement patterns were grouped into patterns suggestive of infection and nonspecific patterns, and the frequency of lesions with patterns suggestive of infection that were seen during each of the contrast-enhanced phases was determined. Because small hypoattenuating lesions—those with the type 1 enhancement pattern—are seen relatively frequently on liver CT scans (13) and confidence to prospectively diagnose them as microabscesses is lower (14), these lesions were termed nonspecific lesions. Lesions with a hyperattenuating rim or associated segmental perfusion abnormalities on contrast-enhanced phase scans are more likely to represent abscesses (15). In addition, a hypervascular enhancement pattern seen on spoiled gradient-echo MR images of liver candidiasis that were obtained immediately after gadolinium enhancement has been described (10). Therefore, for the purposes of this study, lesions with type 2–4 enhancement patterns were considered to be suggestive of infection. The proportions of lesions suggestive of infection that were seen during the arterial and portal venous phases were compared by using logistic regression with generalized estimating equations, as described earlier.
To determine whether lesion conspicuity differed between the different phases, the proportions of lesions with a conspicuity score of 3 (definitely a lesion) that were seen during each phase were calculated. The modeling methods described for detection rates were used to compare the different phases.
SAS version 8.2 statistical software (SAS, Cary, NC) was used for all analyses, except the exact logistic regression analysis, for which LogXact 4 (Cytel Software, Cambridge, Mass) was used. P < .05 was considered to indicate statistical significance.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Lesion Detection
Of the 36 positive-result examinations, 11 yielded positive results during the arterial phase but negative results during the portal venous phase. The rate of diagnosis of hepatic fungal infection achieved by using the arterial phase scans (100%) was significantly higher than the diagnosis rates achieved by using the portal venous phase (69%) and nonenhanced (67%) scans (P < .001 for both comparisons) (Fig 1). The diagnosis rates for the portal venous and nonenhanced phases were not significantly different (P = .68) (Table 1). Of the 11 examinations that yielded positive results during only the arterial phase, four revealed only type 4 lesions. For two examinations, at which a total of eight lesions were depicted, there were follow-up examinations, which revealed changes in the morphologic features or resolution of the lesions.
|
|
|
|
|
|
|
|
Twenty-two of the spleens examined had findings positive for fungal infection. The rates of detecting focal splenic lesions on the arterial phase (20 [91%] of 22 lesions) and portal venous phase (19 [86%] of 22 lesions) scans were not significantly different (P = .66), but the detection rates for both contrast-enhanced phases were significantly higher than the detection rate for the nonenhanced phase (six [30%] of 20 lesions, P < .001 for both comparisons).
Lesion Type Distribution
The distributions of lesion types seen on arterial phase and portal venous phase CT scans are shown in Figure 3. The arterial phase and portal venous phase scans differed significantly in terms of the number of lesions suggestive of infection (types 2–4) that were detected (P < .001) (Table 3). On the arterial phase scans, 386 (79.9%) of the 483 lesions detected were suggestive of infection. On the portal venous phase scans, 134 (40.7%) of the 329 lesions detected were suggestive of infection (Fig 4). Of the 145 type 1a lesions detected on portal venous phase scans, 66 (45.5%) were suggestive of infection on arterial phase scans. A large difference in the type 4 lesion detection rate also was noted (Fig 5). In a few of the patients for whom follow-up CT scans were available (Fig 2), we documented the regression of type 4 lesions to type 2a or type 2b lesions.
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Imaging has an important role in the diagnosis and follow-up of hepatosplenic fungal infection (6,8,10,21). In routine clinical practice, US, CT, and MR imaging may be used. Five patterns of US findings of liver candidiasis (21), which correlate with the stages of the disease, have been described. Data reported in the available literature that correlates these morphologic US patterns with portal venous phase CT findings indicate limited success in correlating US and portal venous phase CT findings. The "wheel within a wheel" and "wagon wheel" patterns represent early disease and usually are not seen at CT, whereas the "bull’s eye" pattern, which may also be seen during the acute stage of disease, is seen only occasionally at CT (21,22). It has been suggested, however, that overall, CT is superior to US in depicting fungal liver microabscesses (23) and that dynamic contrast-enhanced MR imaging is superior to dynamic contrast-enhanced portal venous phase CT (10). In another study involving patients with leukemia, portal venous phase CT depicted 60% of the lesions depicted by dynamic contrast-enhanced MR imaging (24); these results are in line with the 61% detection rate with portal venous phase CT observed in this study (compared with cumulative data for all three phases).
Results of studies on hypervascular hepatic lesions such as hepatoma, carcinoid metastasis, and melanoma have shown an increase in lesion detection with multiphasic CT, as compared with the lesion detection achieved with portal venous phase CT only (25–27). Inflammatory liver lesions may be associated with transient hepatic parenchymal enhancement during the arterial phase, probably owing to associated hyperemia and portal venous flow stoppage (15). For example, in cholangitis, transient hepatic parenchymal enhancement may be seen in a periportal location (28). Thus, the host response to the inflammatory process influences the imaging features of inflammatory liver lesions.
In addition, it has been well described that fungal lesions in the liver or spleen may not be visible until the patient recovers from the neutropenic stage of disease (11,29,30) and that they may become invisible if neutropenia recurs. These lesions may reappear 1–6 weeks after neutrophils recover; this reappearance indicates that fungal infection is still present and antifungal treatment needs to be continued (29). These facts suggest that the inflammatory response of the host has an important role in defining the characteristic appearance of these lesions on images (29) and may explain the increased diagnostic yield of the arterial phase in this study.
As the results indicate, without the addition of an arterial phase the overall diagnosis of fungal liver infection would have been missed in 11 (31%) of 36 cases, and 173 (32.3%) of 536 lesions would have been missed with use of only the portal venous phase. No difference in the detection of focal splenic lesions between the arterial phase and portal venous phase scans was noted, and both types of phase scans were more sensitive than the nonenhanced scans. Overall lesion conspicuity did not differ between the arterial, portal venous, and nonenhanced phases.
The use of an arterial phase also caused a significant alteration in the characterization of focal liver lesions with enhancement patterns that were potentially indicative of infection. Specifically, 45.5% of the subcentimeter hypoattenuating lesions seen on portal venous phase CT scans were depicted as lesions suggestive of infection on the arterial phase scans.
It is interesting to note that almost a third of all lesions detected on the arterial phase scans were type 4—that is, hyperattenuating without central hypoattenuation. Since only four patients had biopsy results positive for fungal infection, we do not have histologic correlation for the different morphologic lesion types encountered. However, biopsy and autopsy study results have shown that the nidus of fungal infection in the liver is walled off by inflammatory cells (22,29). Therefore, a possible explanation for the regression of type 4 lesions to type 2 lesions may be that during the early phase of disease, during neutropenia, the patient lacks the necessary inflammatory response to form an identifiable hypoattenuating granuloma. Instead, hyperemia secondary to the nidus of infection is seen. As the patient’s neutrophil count increases and a granuloma forms, the typical hypoattenuating lesion is seen, with or without a hyperattenuating rim or associated transient parenchymal enhancement.
A limitation of this study was the lack of histologic proof in the majority of patients. However, as has been shown in the literature, obtaining histologic proof in this patient population is difficult (6). Therefore, our study inclusion criteria were based on the suggested EORTC-MSG guidelines for research (9). The true sensitivities of all imaging modalities in the diagnosis of fungal liver infection are difficult to determine. In this study, 16 (41%) of 39 patients who fulfilled the EORTC-MSG criteria for probable or proved fungal liver infection had negative CT scans. It is uncertain how many of these patients had occult fungal liver infection.
We excluded all lesions that could be definitively diagnosed as not related to infection on the basis of imaging findings or previous or follow-up imaging results, when available. Thus, we believe that contamination of the pool of lesions evaluated in this study by nonrelated lesions was minimized. It is conceivable that some of the type 4 lesions detected on the arterial phase scans were incidental hypervascular lesions such as focal nodular hyperplasia or the uncommon nodular form of transient hepatic arterial perfusion disorders (28). However, almost 30% of the lesions detected on arterial phase scans in this study were type 4 abnormalities; this proportion far exceeds the reported prevalence of incidental hypervascular lesions in noncirrhotic livers (28,31). Similarly, the frequency of type 1 lesions far exceeded the incidental prevalence of multiple hepatic cysts or biliary hamartomas (32).
Lack of histologic proof also prevented us from specifically defining the different CT-depicted morphologic structures encountered with pathologic correlates. Instead, we used the available literature to define lesions that potentially represented microabscesses and thereby determine the increased specificity achieved with the arterial phase.
In the patients with liver lesions there was a large number of lesions per patient. To minimize bias in such cases, the readers were asked to select at least one lesion per segment so that the liver could be sampled as homogeneously as possible. With use of this selection technique, some bias is introduced. For example, the most conspicuous lesions are selected, and this practice can cause an artificial increase in conspicuity scores for some phases at lesion-by-lesion analysis. The statistical bias introduced by having multiple lesions in a given patient, as well as the bias introduced by having a given patient included at more than one time point in the study, was taken into account in the statistical analyses.
Some patients did not undergo nonenhanced CT, whereas all patients underwent arterial phase and portal venous phase scanning. This factor may have led to a reduction in the reported detection rate for the nonenhanced phase. The decision to perform a biphasic instead of triphasic CT examination was arbitrary and based on a transient change in protocol at our institution; thus, we expected the associated bias to be small.
In summary, the significant increase in sensitivity and lesion conspicuity at arterial phase CT indicates that a multiphasic technique is needed for the assessment of focal liver lesions in immunocompromised patients suspected of having hepatosplenic fungal infection. The addition of an arterial phase may also yield additional imaging features that could aid the radiologist in making a more confident diagnosis of this disease.
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, U.M., M.A.H.; study concepts, U.M., M.A.H., M.A., M.M.; study design, U.M., M.A.H., G.L.; literature research, U.M.; clinical studies, M.M., U.M.; data acquisition, U.M., M.A.H., M.D.M., G.L.; data analysis/interpretation, U.M., M.A.H., M.D.M., M.A., G.L.; statistical analysis, G.L.; manuscript preparation, U.M., M.A.H.; manuscript definition of intellectual content and editing, U.M., M.A.H., M.A.; manuscript revision/review and final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. S. Levine, O. M. Navarro, G. Chaudry, J. J. Doyle, and S. I. Blaser Imaging the Complications of Bone Marrow Transplantation in Children RadioGraphics, March 1, 2007; 27(2): 307 - 324. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
