It is accepted that hypereosinophilia may be associated with tissue damage. All organ systems may be susceptible to the effects of sustained eosinophilia. The common organ systems are hematologic (100%), cardiovascular (58%), utaneous (56%), neurologic (54%), pulmonary (49%), and gastrointestinal (38%) systems [1]. Liver is involved in only 20%–30% of patients [2]. Eosinophilic infiltration into the liver (mainly into the periportal space) is an uncommon entity that is characterized by multiple focal lesions. The resultant focal lesion is an abscess or granuloma with marked eosinophilic infiltrates on pathology, also called eosinophilic abscess or eosinophilic granuloma. Although the mechanism of eosinophil-related tissue damage is poorly understood, the process might involve the infiltration of eosinophils into tissue damage related to eosinophil function and products (e.g., eosinophil major basic protein and eosinophil cationic protein), and the occurrence of thromboembolic phenomena [3]. Eosinophilic organ infiltration has been described secondary to identifiable causes such as drug hypersensitivity, allergic diseases, malignancies, hypereosinophilic syndrome, collagen vascular diseases, and, most commonly, to parasitic infections [4, 5]. Some researches have revealed pathologic changes in various organs such as heart, lung, and brain caused by eosinophilia [6,7,8,9]. Most radiologic reports regarding eosinophilia have focused on pulmonary changes.

Some previous reports have described some specific radiologic findings of FEI of the liver on ultrasound (US) and computed tomography (CT) images. However, few studies describe magnetic resonance imaging (MRI) features [10,11,12]. Compared with CT, MR provides better soft tissue contrast resolution and thus can distinguish enhancing focal lesions from hepatic tissue based on differences in the lesion-to-liver contrast enhancement during both the equilibrium and portal venous phases [13]. Diffusion-weighted imaging (DWI) has become a practical tool for the detection and the characterization of hepatic lesions. Malignant liver tumors usually have lower apparent diffusion coefficients (ADCs) than benign liver tumors. Thus, combined imaging modalities are superior to single modality in evaluating this disease.

Generally, it is not easy to discriminate FEI from other lesions in patients with underlying malignancies by radiologic findings alone [14]. Several identifying radiologic features of FEI also appear in other diseases, particularly with hepatic metastases. Therefore, the diagnosis of eosinophilic infiltration in the liver is of importance radiologically and clinically.

The most important clinical feature of eosinophilic infiltration is the range of eosinophilis in the peripheral blood, for which the upper limit of normal is 3%–5%, with a corresponding absolute eosinophil count of 350–500/mm3 [15]. Hypereosinophilia has generally been defined as a peripheral blood eosinophil count greater than 1500/mm3 and may be associated with tissue damage [1]. However, few reports focused on the imaging features caused by hypereosinophilia.

Because appropriated surgical treatment might otherwise not be undertaken, the preoperative diagnosis of hepatic eosinophilic infiltration presenting as hepatic nodules is important. Moreover, it occurs with a higher incidence among patients with an underlying malignancy. To our best of knowledge, there is no study described joint CT and MRI features of hepatic FEI and uncovered the association between the number of hepatic FEI and the increase of eosinophils in peripheral blood.

Accordingly, the purpose of this study was to find some specific CT and MRI findings in 29 patients with confirmed FEI of the liver and thus improve the radiologists’ diagnostic ability of this disease.

Materials and methods

Patients

This retrospective study was approved by our Institutional Review Board, and informed consent was waived for the review of clinical records and radiologic images. We retrospectively reviewed and collected the medical records from the radiology department between January 2005 and June 2015. A total of 237 patients had been diagnosed with hypereosinophilia by laboratory examination and bone marrow biopsy. Of these, 29 patients with a confirmed hepatic eosinophilic infiltration were enrolled in this study. The diagnosis of all patients was based on percutaneous needle biopsy.

All patients underwent contrast-enhanced upper abdominal CT and/or MRI scan pre-treatment. The initial imaging detecting the hepatic eosinophilic infiltration was by a helical triple-phase enhanced CT scan (n = 24), and a triple-phase enhanced MRI scan (n = 2), of which, three cases underwent CT and MRI scan at the same time.

CT protocol

All the triphasic dynamic helical CT scannings were performed on Lightspeed Ultra (GE Medical Systems, Milwaukee, WI) with a bolus injection of 90 mL of nonionic contrast media (Iopromide 370; Schering AG, Berlin, Germany) via the anticubital vein at a rate of 4–5 mL/s by power injector, with images obtained 36–44, 70–78, and 180 s after the start of contrast injection during the hepatic arterial, portal, and equilibrium phase, respectively. The parameters were as follows: detector configuration = 64 × 0.625 mm2, slice thickness = 3 mm, reconstruction interval = 2 mm, table speed = 46.9 mm/rotation, rotation time = 0.5–0.75 s, tube current = 250 mAs, tube voltage = 120 kVp, matrix = 512 × 512.

MRI protocol

MR imaging was performed on 5 patients with a 1.5 T (Signa Horizon; GE Medical Systems, Milwaukee, WI) MRI unit. All images were obtained in the transverse plane using a four-channel phased-array body coil. Baseline MR sequences included a breath-hold T1-weighted, spoiled gradient recalled echo (GRE) sequence; a respiratory-triggered, T2-weighted, rapid acquisition relaxation-enhanced (RARE) sequence; and a single-shot fast spin-echo (SSFSE) sequence. The T2-weighted RARE imaging was performed using the parameters: TR/TE = 3000–7500/97.1 ms; echo- train length = 12; matrix = 512 × 384; chemical shift fat suppression; array spatial sensitivity encoding technique (ASSET); acceleration factor of 2; and a signal average of 3. T2-weighted, SSFSE imaging was performed using the parameters: TR/TE = infinite/86.9 ms; flip angle = 90°; echo-train length = 240; and matrix = 384 × 192. Breath-hold T1-weighted, spoiled GRE (in-phase and out-of-phase) sequence was performed with the parameters: TR/TE = 140/2.4 and 5.8 ms; flip angle = 70°; matrix = 512 × 192; and a signal average of 1. For all sequences, a 7-mm slice thickness was used with a 25% intersection gap.

Dynamic MR imaging was performed with IV administration of Gd-DTPA (MultiHance; Bracco SpA, Milan, Italy) or Gd-EOB-DTPA (Primovist; Bayer Healthcare, Berlin, Germany) at a rate of 2 mL/s, at a dose of 0.1 mmol/kg of body weight. For contrast-enhanced dynamic images, arterial (20–35 s), portal (45–60 s), and equilibrium (3 min) phases were obtained using a T1-weighted GRE sequence with liver acquisition with volume acceleration (LAVA). The parameters for LAVA were as follows: TR/TE = 4.2/1.9 ms; flip angle = 10°; bandwidth = 83.3 Hz/Px; matrix = 320 (read) × 256 (phase) × 68–84 (partition); slice thickness = 2.5 mm; and FOV = 32 × 35 cm. ASSET with an acceleration factor of 2 was applied in an in-plane phase-encoding direction for 3D-dynamic imaging. Additional hepatocyte-phase (HP) images were acquired 20 min after the injection of Gd-EOB-DTPA.

Respiratory-triggered (using a navigator-echo technique) fat-suppressed single-shot echo-planar diffusion-weighted imaging (DWI) was performed in the transverse plane with tridirectional diffusion gradients using 3 values (0, 100, and 600 s/mm2). The other parameters for DWI were as follows: TR/TE, 5500/83 ms; flip angle, 90°; field of view, 400 mm; matrix, 120 × 177; section thickness, 5 mm; slice spacing, 6 mm; 1 signal acquired).

Image analysis

All CT and MR images were independently reviewed by two abdominal radiologists in consensus, who had 16 and 10 years of experience in liver CT/MRI, respectively. The total number of liver lesions was counted and the size of each lesion was measured using pre-contrast images on picture archiving and communication system (PACS) monitor. In addition, the margin (poorly defined or well defined), shape (nodular or irregular), and distribution (subcapsular or non-subcapsular) of the lesions were analyzed. Poorly defined margin was defined as lesion has a unclear margin that could not be separated from the liver. Irregular shape was defined as random shape. The subcapsular distribution was defined as the area within 2.0 cm of the liver surface [13]. We also evaluated the signal intensity on the pre-contrast T1- and T2-weighted images. The signal intensity of the lesions was categorized into hypointense, isointense, or hyperintense relative to the surrounding liver parenchyma on all MR sequences. The following two enhancement features were evaluated for the post-gadolinium (Gd) dynamic images of each liver lesion: appearance of enhancement (rim or homogeneous) and enhancement pattern. We classified the enhancement patterns as hypointense, isointense, hyperintense rim or nodule, and homogeneous hyperintense relative to normal liver parenchyma.

Statistical analysis

Data are n (%) unless otherwise indicated. The SPSS statistics package (version 23.0 for window) was used for statistical analysis. The Spearman rank correlation test is used to correlate the number of hepatic lesions and eosinophilic counts. The interrater reliability between two readers in terms of imaging features was assessed using simple Kappa coefficient. A p value <0.05 was considered statistically significant.

Results

Clinical features

A total of 29 patients were enrolled in this present study (25 men and 4 woman; mean age, 33.9 years, range, 2–74 years) from January 2005 to June 2016. Clinical symptoms included recurrent fever (n = 18), abdominal pain (n = 12), diarrhea (n = 3), recurrent fever and abdominal pain (n = 11). Other organs of the eosinophilic infiltration included gastrointestinal eosinophilic infiltration (n = 3), eosinophilic infiltration of the skin (n = 1), eosinophilic infiltration of the spleen (n = 1), eosinophilic myocardial infiltration (n = 1). In all, 14 of the 29 patients had parasitic infections. All patients had an increased eosinophil counts more than 5% of the white blood cells (mean percentage, 49.7%; range, 20%–72%) in the peripheral blood, with a corresponding absolute eosinophil count of 1530–40,160/mm3 (mean, 14,260/mm3). Spearman analysis indicated a significant correlation between the number of lesions and the increased eosinophils in peripheral blood (r = 0.627, p = 0.0003) (Fig. 1).

Fig. 1
figure 1

The Spearman correlation analysis between the number of eosinophilic infiltration lesions of the liver and the eosinophil counts in peripheral blood

Full size image

Imaging findings

A total of 108 lesions of eosinophilic infiltration were detected in 29 cases. Of the 29 cases, 2 were with single lesion and the remaining 27 were with multiple lesions (1–4 lesions per case). The mean size of all lesions was 34 mm (range, from 3 to 61 mm). A total of 101 (93.5%) lesions were located in subcapsular parenchyma or surrounding the portal vein. Most (66%) subcapsular lesions were wedge shaped, and all lesions surrounding portal vein were round shaped. However, the hepatic parenchymal lesions were irregular or round shaped (Table 1). Table 2 shows the appearances of pre-contrast and contrast-enhanced CT and MRI images.

Table 1 The summary of lesion shape and location
Full size table
Table 2 The appearances of pre-contrast and contrast-enhanced CT and MRI images
Full size table

CT findings

All lesions were hypo-attenuating or iso-attenuating with ill-defined margins on pre-contrast CT images. Only 65 (64%) were detected on pre-contrast CT images. On arterial phase images, only 36 (36%) lesions with low or high attenuation were found and the 65 (64%) lesions showed iso-attenuation that could not be seen. On portal phase images, 101 lesions were detected, 90 (89%) of which showed low attenuation with well-defined margins, 4 (4%) showed high attenuation and 7 (7%) showed iso-attenuation. Special features including branches of portal vein went through the lesions; ‘stripe sign’ and ‘halo ring sign’ were observed (Figs. 2, 3). On equilibrium phase images, 86 (80%) lesions showed iso-attenuation or nearly iso-attenuation. The detection rates of CT scanning were 64%, 36%, 93%, and 20% on pre-contrast images, arterial, portal, and equilibrium phases, respectively. The detection rates of pre-contrast images and postcontrast images were significantly different (p < 0.012, for all).

Fig. 2
figure 2

A 43-year-old man with hypereosinophilic syndrome. A The pre-contrast CT image shows multiple isodense lesions in subcapsular parenchyma or surrounding the portal vein. B The lesions are slightly hypo-attenuating or iso-attenuating relative to the liver in the arterial phase. C In the portal phase, the lesions became clear and hypo-attenuating relative to the liver. White arrow shows branches of portal vein through the lesions, white arrowhead shows ‘stripe sign’ accompanied the portal vein branches, black arrow shows ‘halo ring sign’ around the portal vein. D In the equilibrium phase, lesions show iso-attenuating

Full size image
Fig. 3
figure 3

A 33-year-old man with hypereosinophilic syndrome and numerous eosinophilic infiltrations at 68% serum eosinophils (16.45 × 109/L). A Pre-contrast scan shows slightly low attenuated lesions with ill-defined margins. B–D On arterial, portal, and equilibrium phases, the lesion shows prominent low attenuation in contrast to the adjacent liver with size reduction. We could observe branches of portal vein went through the lesion (blue arrow)

Full size image

MRI findings

MR images showed 21 lesions of FEI in 5 patients. On T1-weighted images, 12 (57%) lesions were isointense and 9 (43%) were hypointense. On T2-weighted images, 17 (81%) lesions were hyperintense and 4 (19%) lesions were isointense. All lesions were with poorly defined margins. After Gd-DTPA injection, 17 (81%) lesions were isointense relative to the liver on the arterial phase. On the portal phase, 18 (86%) lesions became hypointense relative to the liver. On the equilibrium phase, 17 (81%) lesions were isointense or nearly isointense, only 1 (5%) lesion showed hypointense and 3 (14%) lesions showed hyperintense. One case showed mixed hypointensity in the hepatobiliary phase (Fig. 5). All cases showed hyperintensity in both low and high b value ranges on DWI, with ADCs of less than 1.00 × 10−3 mm2/s (Fig. 5). Special features including branches of portal vein went through the lesions, and ‘stripe sign’ were observed (Figs. 4, 5). The detection rates of MR scanning were 81%, 20%, 96%, and 19% in pre-contrast images, the arterial, portal, and equilibrium phases, respectively. The detection rate of lesions in portal phase images was the highest (p < 0.001 for each comparison).

Fig. 4
figure 4

A 42-year-old with hypereosinophilic syndrome. A The unenhanced T1-weighted image shows multiple isointense lesions in subcapsular parenchyma or surrounding the portal vein. B The lesions are isointense relative to the liver in the arterial phase. C During the portal phase, the lesions became hypointense relative to the liver. White arrow shows branches of portal vein went through the lesions, white arrowhead shows ‘stripe sign’ accompanied with the portal vein branches. D Lesions show isointense on the equilibrium phase

Full size image
Fig. 5
figure 5

A 31-year-old man with hypereosinophilic syndrome and numerous eosinophilic infiltrations at 72% serum eosinophils (16.29 × 109/L). A On T1-weighted image, the lesion shows iso-/hypointensity relative to the liver. B On T2-weighted image, the lesion shows mixed hyperintensity. C On arterial phase, the lesion shows slight enhancement and hypo-intensity relative to the liver. D–E On portal and equilibrium phases, we could observe ‘halo ring sign’ (blue arrow). F–I On hepatobiliary phase (3, 10, 15, 20, and 30 min after the contrast injection), the lesion shows mixed hypointensity with size reduction. J, K On DWI, the lesion shows hyperintensity in both low and high b value ranges, with the corresponding ADC of less than 1.00 × 10−3 mm2/s

Full size image

A total of 14 lesions were detected in 3 patients underwent both. For CT, there were 8 (57.1%), 5 (35.7%), 13 (92.9%), and 3 (21.4%) lesions detected on pre- contrast, arterial phase, portal phase, and equilibrium phase images, respectively, whereas for MRI, there were 11 (78.6%), 3 (21.4%), 14 (100%), and 3 (21.4%) lesions detected on pre-contrast, arterial phase, portal phase, and equilibrium phase images, respectively.

Table 3 shows the enhancement pattern and special signs on dynamic contrast-enhanced images, and 23 (79.3%) cases were gradually enhanced. Branches of portal vein went through the lesions were found in 29 cases; 12 found hypodense or hypointense ‘stripe sign’ accompanying with the portal vein branches, and 16 found low attenuating or hypointense ‘halo ring’ around the portal vein.

Table 3 The enhancement pattern and special signs
Full size table

The interrater reliability between two readers in terms of imaging features was assessed using simple Kappa coefficient, ranging from 0.87 to 1.00.

Discussion

Several reports have described CT or MR findings of eosinophilic liver diseases, including eosinophilic infiltration and eosinophilic abscesses [10, 17, 18]. In this study, we retrospectively analyzed the CT and MRI findings of hepatic eosinophilic infiltration and identified some specific features that are helpful in differentiating it from the other liver diseases, especially metastatic tumor. In addition, spearman correlation analysis indicated a strong correlation between the number of lesions and the increased eosinophils in peripheral blood (r = 0.627, p = 0.0003), which was consistent with the previous studies [16].

By analyzing the 108 lesions in 29 patients, we found that most (88.6%) lesions were located in subcapsular parenchyma or surrounding the portal vein. For subcapsular lesions, wedge shape was seen in most cases. All lesions surrounding portal vein were round shaped. However, the hepatic parenchymal lesions were irregular or round shaped. These findings support the view that the lesions located around portal area and peripheral vascular [19, 20].

On pre-contrast and unenhanced T1-weighted images, about 20%–60% lesions are iso-attenuated or isointense, which are likely to be missed during the diagnosis, but most of them were hyperintense on T2-weighted images [21]. Regarding the enhancement pattern of the dynamic study, it should be noted that the lesions appeared poorly defined low attenuating or hypointensity during the portal phase, which was regarded as a main imaging finding of FEI [14]. In this present study, 23 (79.3%) cases appeared as iso-attenuating or isointense on arterial phase, poorly defined low attenuating or hypointense on portal phase, and iso-attenuating or isointense on equilibrium phase, that is, ‘iso-hypo-iso’ enhancement pattern. Thus, one of the characteristic imaging findings of hepatic eosinophilic infiltration is reported to be most conspicuous during the portal phase and to gradually become obscured during the equilibrium phase in CT scan. However, the dynamic-enhanced MRI of hepatic FEI is somewhat variable according to the previous reports. A study found that the lesions (60%) presented nodular or rim or perilesional enhancement during the arterial phase [22]. Sun et al. observed that eosinophilic infiltration was hyperintense on the portal phase (82%) and equilibrium phase (77%) after the injection of Gd-DTPA [13]. Kim et al. [11] reported that 50% of focal eosinophilic liver diseases were isointense during the portal phase. Our results showed that most focal lesions showed hypointensity on the portal (86%) and equilibrium phase (5%) in relation to the surrounding liver parenchyma after a bolus injection of contrast [11]. It indicated that portal phase was the optimal phase to detect the FEI [23]. All lesions in our study showed hyperintensity in both low and high b values on DWI, which means that water molecules were being restricted within the lesions. However, the results suggest that DWIs with ADCs on lesions might demonstrate malignant features. Therefore, DWIs and ADCs are not likely to be helpful in differentiating FEIs from other malignant tumors, such as liver metastases.

The other important findings in terms of enhancement pattern were as follows: (1) most lesions (79.3%) showed continuous homogenous enhancement pattern during the dynamic study; (2) in all cases, branches of portal vein went through the lesions; (3) 12 cases were found low attenuating or hypointense ‘stripe sign’ accompanying with the portal vein branches; (4) and 16 cases were found low attenuating or hypointense ‘halo ring sign’ around the portal vein. The exact mechanisms for ‘stripe sign’ and ‘halo ring sign’ remain unclear. These findings had not been reported by the previous studies and were helpful to differentiate FEI from other liver diseases, such as metastasis. However, in patients with malignancy, it is difficult to exclude metastasis based only on radiologic images. Histologic confirmation is advised despite the presence of metastasis.

Our study also had some limitations. First, although we used 18-gauge automated biopsy guns, the specimens were too small to provide an exact pathologic correlation with the imaging findings. Second, various patterns in the dynamic studies depended on a variety of conditions, such as stage, age, and severity of eosinophilic infiltration. Lastly, fibrous stroma in the focal lesion might play a role in contrast enhancement in an equilibrium phase scan. Despite some fibrous stroma in our pathology specimen, it is difficult to explain the definite relation between MR findings and pathology in patients.

In summary, CT and MR imagings play the important role in the detection and characterization of focal eosinophilic infiltrations of the liver. Subcapsular location, obscure margins, continuous homogenous enhancement, branches of portal vein through the lesions, ‘halo ring sign,’ ‘stripe sign,’ and hypereosinophilia could be characteristic features of FEI of the liver. These results concerning the diagnosis of this disease may be useful in daily practice.