Differentiation of Malignant From Benign Focal Splenic Lesions ...

Gastrointestinal Imaging • Original Research Jang et al. Addition of DWI for MRI of Focal Splenic Lesions

Downloaded from www.ajronline.org by 2.27.95.35 on 11/03/15 from IP address 2.27.95.35. Copyright ARRS. For personal use only; all rights reserved

Gastrointestinal Imaging Original Research

Differentiation of Malignant From Benign Focal Splenic Lesions: Added Value of Diffusion-Weighted MRI Kyung Mi Jang1 Seong Hyun Kim1 Jiyoung Hwang2 Soon Jin Lee1 Tae Wook Kang1 Min Woo Lee1 Dongil Choi1 Jang KM, Kim SH, Hwang J, et al.

OBJECTIVE. The objective of our study was to evaluate the added value of diffusion-weighted imaging (DWI) for distinguishing between malignant and benign focal splenic lesions. MATERIALS AND METHODS. This study included 53 patients with 11 malignant and 42 benign splenic lesions who underwent gadoxetic acid–enhanced MRI and DWI. Qualitative and quantitative analyses were conducted for splenic lesions. Two blinded observers evaluated the two image sets—that is, the conventional MR images alone versus the combined conventional MR and DW images—and scored their confidence for malignancy of splenic lesions. The Fisher exact test and Mann-Whitney U test were performed, and diagnostic performance values (ROC curve analysis) were evaluated. RESULTS. All malignant lesions showed a progressive hypovascular enhancement pattern, whereas the hypervascular enhancement patterns were shown in only benign lesions (n = 20, 47.6%) (p < 0.05). The mean apparent diffusion coefficient (ADC) of the malignant lesions (0.73 × 10 −3 mm2 /s) was significantly lower than that of the benign lesions (1.21 × 10 −3 mm2 /s) (p < 0.001). The addition of DW images to conventional MR images showed a significant improvement for predicting malignant splenic lesions (area under ROC curve [A z] without DW images vs with DW images: 0.774 vs 0.983 for observer 1 and 0.742 vs 0.986 for observer 2) (p < 0.001). In addition, the diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of combined conventional MR and DW images were higher than those of conventional MR images alone. CONCLUSION. The addition of DWI to conventional MRI improves differentiation of malignant from benign splenic lesions.

A

Keywords: diffusion-weighted imaging, MRI, spleen DOI:10.2214/AJR.13.11914 Received September 15, 2013; accepted after revision December 31, 2013. 1 Department of Radiology and Center for Imaging Science, Samsung Medical Center, Sungkyunkwan University School of Medicine, 50 Ilwon-dong, Gangnam-gu, Seoul 135-710, Korea. Address correspondence to S. H. Kim ([email protected]). 2 Department of Radiology, Soonchunhyang University Hospital, Seoul, Korea.

AJR 2014; 203:803–812 0361–803X/14/2034–803 © American Roentgen Ray Society

wide range of benign and malignant diseases can affect the spleen, thus complicating the development of a differential diagnosis that includes both benign and malignant focal lesions. Therefore, when focal splenic lesions are encountered on imaging studies, a splenic biopsy is often considered, particularly in patients with extrasplenic malignancies. Because most splenic lesions that are found incidentally are benign, it might not be necessary to perform a biopsy in all patients with a focal splenic lesion [1]. In addition, there is a risk of bleeding after a splenic biopsy because of the rich blood supply of the spleen, and focal splenic lesions may not be accessible to ultrasound-guided biopsy because of a poor sonic window [2–5]. When pathologic diagnosis is mandatory but biopsy may be difficult to perform, the classic alternative is splenectomy or laparoscopy-guided splenic biopsy. However, unnecessary splenectomy for be-

nign lesions will lead to the loss of an important immunologic organ. Therefore, it is important to noninvasively differentiate malignant from benign focal splenic lesions on imaging. The noninvasive distinction between malignant and benign focal splenic lesions is usually based on enhancement patterns seen after the injection of a contrast agent on CT and MRI [1, 6–18]. However, accurate differentiation of malignant from benign focal splenic diseases may not be possible solely on the basis of conventional morphologic imaging findings [19–22]. With the current advances in MR technology, diffusion-weighted imaging (DWI) of the abdomen has become more widely used. DWI is an appealing technique because it is noninvasive, can be performed relatively quickly, and does not require exogenous contrast agents or ionizing radiation [23]. The apparent diffusion coefficient (ADC) value, which is reflective of tissue water mobility

AJR:203, October 2014 803


Jang et al. and is calculated from DW images, is helpful for differentiating benign from malignant tumors, assessing treatment response of tumor, and diagnosing cirrhosis and fibrosis [23– 26]. Thus, we hypothesized that DWI could be used to differentiate malignant from benign focal splenic lesions. The purpose of our study was to evaluate the added value of DWI for distinguishing between malignant and benign focal splenic lesions.

Fig. 1—Flowchart shows selection of study population. DWI = diffusion-weighted imaging.

Search radiologic database for abdominal MR examinations performed between October 2008 and April 2013 using the terms “spleen” and “splenic” (n = 689 patients)

Inclusion criteria: 1. Gadoxetic acid–enhanced MRI with DWI 2. Pathologic diagnosis of focal splenic lesions or clinical diagnosis of benign focal splenic lesions and imaging follow-up with abdominal CT or MRI over 24 months

Materials and Methods Study Population This retrospective study was approved by the ethics committee at our institution and informed consent was waived. We searched our radiologic database for abdominal MR examinations performed between October 2008 and April 2013 using the search terms “spleen” and “splenic”; this search yielded 689 patients. To develop a study group of suitable cases in which to compare the MRI findings between benign and malignant focal splenic lesions, we used the following inclusion criteria: patients who underwent abdominal gadoxetic acid–enhanced MRI with DWI and who had a pathologic diagnosis of focal splenic lesions or had a clinical diagnosis of benign focal splenic lesions and imaging follow-up with abdominal CT or MRI over 24 months. On the basis of these inclusion criteria, we excluded 587 patients. An additional 49 patients were excluded for the following reasons: MRI findings suggestive of typical benign cystic lesions (n = 41) and disseminated benign lesions (n = 3 [tuberculosis, n = 2; sarcoidosis, n = 1]) or of disseminated malignant lesions (n = 5 [peritoneal seeding from gastric, colon, or ovarian cancer, n = 4; hepatocellular carcinoma, n = 1]) that involved the extrasplenic upper abdominal organs and lymph nodes on MRI (n = 8), which may have affected decision making during the blinded review. Typical benign cystic lesions were defined as a well-defined lesion with signal intensity identical to CSF on T2-weighted and heavily T2-weighted images with neither an enhancing solid portion nor a discernible wall [1, 15, 18]. In our study, there was no patient with chronic liver disease with splenomegaly. The review of the MR images was performed by one experienced radiologist who had 20 years of experience in abdominal MRI interpretation at the time of the study. Finally, 53 patients (mean age, 48.6 years; age range, 21–75 years), 24 men (mean age, 50.0 years; age range, 22–75 years) and 29 women (mean age, 47.5 years; age range, 21–69 years), were included in our study group. In 11 of the 53 patients, malignant splenic lesions were diagnosed as metastases (n = 9 [colon cancer, n = 6; ovarian cancer,

804

587 patients Exclusion criteria: 1. MRI findings suggestive of typical benign cystic lesions (n = 41) 2. Disseminated benign or malignant lesions in the extrasplenic upper abdominal organs and lymph nodes on MRI (n = 8) 49 patients 53 consecutive patients

n = 2; neuroendocrine carcinoma of the pancreas, n = 1]), non-Hodgkin lymphoma (n = 1), and angiosarcoma (n = 1) by surgery (n = 10) and ultrasound-guided biopsy (n = 1). Among these patients, one of the patients with colon cancer had four metastatic lesions and the patient with angiosarcoma had more than 10 lesions. The remaining nine patients had a solitary lesion. Of the remaining 42 patients with benign focal splenic lesions, benign splenic lesions in 10 patients were diagnosed as inflammatory pseudotumor (n = 3), sclerosing angiomatoid nodular transformation (n = 2), tuberculosis (n = 2), hematoma (n = 1), old hematoma (n = 1), or vascular malformation (n = 1) by surgery (n = 7) and ultrasound-guided biopsy (n = 3). The remaining 32 patients were clinically diagnosed with benign focal splenic lesions after the following findings were observed: disappearance, decrease, no interval change, or minimal interval growth (< 5 mm) in size on follow-up images with no change of morphologic features with enhancement pattern over 24 months. In these 42 patients, five had more than 10 splenic lesions, and the remaining 37 patients had a solitary lesion. The case accrual process is summarized in Figure 1.

MRI Technique All MRI examinations were acquired using a 3-T whole-body MR system (Intera Achieva, Philips Healthcare) and a 16-channel phased-array coil as the receiver coil. Baseline MRI examinations were composed of three sequences. The first sequence was a T1-weighted turbo field-echo in- and opposed-phase sequence (TR/first-echo

TE, second-echo TE, 10/2.3 [in phase], 3.5 [opposed phase]; flip angle, 15°; matrix size, 256 × 194; bandwidth, 434.3 Hz/pixel). The second sequence was a breath-hold multishot T2-weighted sequence with an acceleration factor of 2 (TR/TE, 1623/70; flip angle, 90°; matrix size, 324 × 235; bandwidth, 258.4 Hz/pixel). The third sequence was a respiratory-triggered single-shot heavily T2-weighted sequence with an acceleration factor of 2 (1156/160; flip angle, 90°; matrix size, 256 × 256; bandwidth, 420.9 Hz/pixel). A 5-mm section thickness and an FOV of 32–38 cm were used for all three sequences. DW images were acquired simultaneously before the administration of gadoxetic acid using respiratory-triggered single-shot echo-planar imaging (TR/TE, 1600/70); the TR was matched in each patient to the length of the respiratory cycle. The scanning parameters were as follows: b values of 0, 100, and 800 s/mm2; spectral presaturation with inversion recovery for fat suppression; matrix size, 112 × 112; acceleration factor of sensitivity encoding, 2.0; FOV, 35 × 35 cm; number of signals acquired, 2; slice thickness, 5 mm; slice gap, 1 mm; and 33 axial slices. The acquisition time for this sequence ranged from 3 to 4 minutes depending on the respiratory efficiency of each patient. The ADC was generated using a monoexponential function with b values of 0 and 800 s/mm2. For gadoxetic acid–enhanced imaging, unenhanced, arterial phase (20–35 seconds after contrast administration), portal venous phase (60 seconds), late phase (3 minutes), and hepatobiliary phase (20 minutes) images were obtained using a T1-weighted 3D turbo field-echo se-

AJR:203, October 2014


Addition of DWI for MRI of Focal Splenic Lesions

A

B

C

D

Fig. 2—42-year-old woman with diffuse large B-cell lymphoma in spleen. A–D, Compared with splenic parenchyma, tumor (arrow) has hypovascular enhancement on transverse gadoxetic acid–enhanced arterial phase image (A) and appears as isointense lesion containing central necrosis (arrowhead, B) on transverse fat-saturated T2-weighted image (B), as hyperintense lesion on diffusion-weighted image obtained with b value of 800 s/mm2 (C), and as hypointense lesion on apparent diffusion coefficient (ADC) map (D). ADC value and tumor-to–splenic parenchyma ADC ratio are 0.55 × 10 −3 mm2 /s and 0.58, respectively. Lesion was considered as indeterminate splenic lesion on conventional MR images alone but was diagnosed correctly as malignant splenic lesion after additional review of diffusion-weighted images and ADC map by both observers.

quence (THRIVE [T1-weighted high-resolution isotropic volume examination], Philips Healthcare). The following parameters were used: TR/ TE, 3.1/1.5; flip angle, 10°; matrix size, 256 × 256; bandwidth, 724.1 Hz/pixel; section thickness, 2 mm; and FOV, 32–38 cm. The measured voxel size was 1.5 × 1.5 × 4.0 mm, and the reconstructed voxel size was 1.17 × 1.17 × 2.0 mm. The time for arterial phase imaging was determined by the MR fluoroscopic bolus-detection technique. Contrast agent was administered IV with a power injector at a rate of 2 mL/s for a dose of 0.025 mmol/kg of body weight followed by a 20-mL saline flush. Subtraction images of unenhanced series from each contrast-enhanced series (arterial and portal venous) of each patient were automatically acquired by the software of the MR machine.

Image Analysis Imaging analysis was performed on a PACS (Centricity Radiology RA 1000, GE Healthcare). All reviewers were blinded to the clinical, surgical, and histopathologic results that corresponded with the images. Reviewers evaluated MR images side-by-side using a spatial cursor key on our PACS to facilitate matching corresponding sites on different images. If a patient had multiple similar splenic lesions, the largest lesion was assessed. Qualitative and quantitative analyses—The qualitative analysis of the images was performed by two experienced gastrointestinal radiologists with 21 and 11 years of experience, respectively, in abdominal MRI interpretation in consensus. The margin (well, partially, or ill defined) and signal intensity of the focal splenic lesions compared with that of adjacent normal splenic parenchyma

were assessed on conventional MR and DW images. The focal splenic lesions were graded as hyperintense when the signal intensity of the lesion was greater than that of the normal splenic parenchyma, isointense when the signal intensity of the lesion was similar to that of the normal splenic parenchyma, and hypointense when the signal intensity of the lesion was lower than that of the normal splenic parenchyma on unenhanced T1- and T2-weighted images; gadoxetic acid–enhanced arterial, portal venous, 3-minute late, and 20-minute hepatobiliary phase images; and DW images with a b value of 800 s/mm2. The enhancement patterns of the focal splenic lesions on dynamic imaging were classified into categories as having only rim enhancement, showing progressive hypovascular enhancement, or showing arterial hypervascularity with persistent or washout enhancement.



Jang et al.

A

B

C

D

Fig. 3—32-year-old man with sclerosing angiomatoid nodular transformation in spleen. A–D, Compared with splenic parenchyma, tumor (arrow) has hypervascular enhancement on transverse gadoxetic acid–enhanced arterial phase image (A), appears as heterogeneous hypointense lesion on transverse fat-saturated T2-weighted image (B) and on diffusion-weighted image obtained with b value of 800 s/mm2 (C), and appears as mildly heterogeneous hyperintense lesion on apparent diffusion coefficient (ADC) map (D). ADC value and tumor-to–splenic parenchyma ADC ratio are 1.21 × 10 −3 mm2 /s and 1.18, respectively. Lesion was considered as indeterminate splenic lesion on conventional MR images but was diagnosed correctly as benign splenic lesion after additional review of diffusion-weighted images and ADC map by both observers.

The quantitative analyses in our study were performed by an experienced radiologist with 11 years of experience in abdominal MRI. The size of the lesions, ADC values, and lesion-to-parenchyma ratios of ADC values for focal splenic lesions and normal splenic parenchyma on an ADC map were measured. The solid portions of focal splenic lesions equal to or greater than 10 mm in the shortest diameter were included in the quantitative signal intensity analysis to reduce partial volume averaging effects. Circular or ovoid ROIs were placed to include almost the entire area of the homogeneous solid portion of the focal splenic lesion while avoiding the most peripheral portions to exclude partial volume effects of adjacent extralesional tissue. For heterogeneous lesions, ADC values of the most strongly enhancing portion of the lesions were assessed. The signal intensities were measured three

806

times and averaged. The mean ROI sizes were as follows: 69.4 ± 117.5 (SD) mm2 (range, 6.3–834 mm2) for the lesions and 39.2 ± 1.4 mm2 (12–75 mm2) for the splenic parenchyma. Evaluation of the added value of DWI for differentiating malignant from benign focal splenic lesions—The two image sets (i.e., the conventional MRI set and the combined set of conventional MR images and of DW images with b values of 0, 100, and 800 s/mm2 with an ADC map) were assessed by two gastrointestinal radiologists (observers 1 and 2 with 5 and 13 years of experience, respectively) independently. The reading sessions were separated by an interval of 4 weeks. The observers did not participate in selecting the study population or conducting qualitative or quantitative analysis and they were blinded to the information obtained during clinical, surgical, and pathologic analyses.

At the initial reading session, the observers independently reviewed the conventional MR images and recorded their confidence level with respect to their diagnosis of the lesions as malignant or benign with a 5-point scale designed as follows: A score of 1 indicated a definitely benign lesion; 2, probably benign lesion; 3, indeterminate lesion; 4, probably malignant lesion; and 5, definitely malignant lesion. We did not define specific imaging findings as suggestive of malignant or benign focal splenic lesions because multiple benign and malignant diseases with various imaging findings can affect the spleen. Instead, observers determined each score on the scale based on the results of previous studies [1, 6–18] and their clinical experience. At the second reading session, the observers were asked to score their confidence level using the same 5-point scale with respect to the diagnosis of the le-




A

B

C

D

Fig. 4—40-year-old man with tuberculous abscess in spleen. A–D, Compared with splenic parenchyma, lesion (arrow) has only rim enhancement on transverse gadoxetic acid–enhanced arterial phase image (A) and appears as slightly hyperintense lesion on transverse fat-saturated T2-weighted image (B), as hyperintense lesion on diffusion-weighted image obtained with b value of 800 s/mm2 (C), and as hypointense lesion on apparent diffusion coefficient (ADC) map (D). ADC value and lesion-to–splenic parenchyma ADC ratio are 0.81 × 10 −3 mm2 /s and 0.69, respectively. Lesion was considered as indeterminate splenic lesion on conventional MR images and was misdiagnosed as malignant splenic lesion after additional review of diffusion-weighted images by both observers.

sions as malignant or benign on the basis of findings for the combined imaging set (i.e., conventional MR images and DW images with b values of 0, 100, and 800 s/mm2 and an ADC map). The spleen is a physiologically restricted diffusion organ [26]. We defined splenic lesions as malignant when the lesion showed hyper- or isointense signal on DW images (b = 800 s/mm2) and iso- or hypointense signal on the ADC map compared with the signal intensity of the normal splenic parenchyma; however, we did not define a lesion as malignant when the lesion showed isointense signal on all DW images (b values = 0, 100, and 800 s/mm2) and on the ADC map. When the DWI findings differed from those seen on conventional MR images, the observers were asked to give priority to the DWI findings.

Statistical Analysis For qualitative and quantitative comparisons between malignant and benign lesions, the Fisher ex-

act test and Mann-Whitney U test were used. The diagnostic performance of each observer was calculated with an ROC curve analysis, and the area under the ROC curve (Az) was calculated. For comparing the diagnostic performance before and after review of additional DW images, pairwise comparisons of the ROC curves were performed. The 95% CIs were used to express the statistical precision of the results. We calculated diagnostic accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for both observers on the basis of the assumption that a confidence level of 4 or higher was a positive diagnosis of malignant lesions. The McNemar test was used to compare accuracy, sensitivity, and specificity between interpretation of conventional MR images alone and interpretation of DW and conventional MR images together. We defined the false-negative cases as those assigned confidence levels of 3 or less that were ultimately confirmed to be malignant lesions, and false-positive

cases as those assigned confidence levels of 4 or higher that were ultimately confirmed to be benign lesions. False-negative or false-positive cases that were incorrectly interpreted by both observers 1 and 2 were analyzed. The kappa statistic was used for evaluation of interobserver agreement. A kappa value of less than 0.20 indicated poor agreement; 0.21– 0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, good agreement; and a value equal to or greater than 0.81, excellent agreement [27]. We used statistical software (MedCalc, version 12.2.1.0, MedCalc Software) for Microsoft Windows for all statistical analyses, and p < 0.05 was considered to indicate a significant difference for all analyses.

Results Qualitative and Quantitative Analyses For both malignant and benign focal splenic lesions, the margin and signal intensities of the lesions compared with adja-



Jang et al. cent normal splenic parenchyma on images obtained with various pulse sequences are listed in Table 1. On arterial, portal venous, and 3-minute late phases, almost all malignant lesions (100% [11/11], 100% [11/11], and 90.9% [10/11], respectively) showed hypointensity, whereas hyper- or isointensity was seen on arterial and portal venous phase images in only benign splenic lesions (47.6% [20/42] and 52.4% [22/42], respectively) (p < 0.05) (Figs. 2 and 3). All malignant splenic lesions showed a hypovascular progressive enhancement pattern, and hypervascular enhancement patterns were seen in only benign splenic lesions (47.6% [20/42]) (p < 0.05) (Figs. 2 and 3). A benign lesion with only rim enhancement on contrast-enhanced MR images was diagnosed as a tuberculous abscess at pathology (Fig. 4). On DW images obtained with a b value of 800 s/mm2, 10 of 11 malignant lesions (90.9%) showed hyper- or isointense signal and 26 of 42 benign lesions (61.9%) showed hypointense signal compared with the normal splenic parenchyma (p = 0.003) (Figs. 2 and 3). The margin and signal intensities on unenhanced T1- and T2-weighted images and 20-minute hepatobiliary phase images were not significantly different between malignant and benign splenic lesions (p > 0.05). The sizes of the malignant lesions (36.54 ± 12.27 mm; range, 15–52 mm) and benign lesions (32.04 ± 17.97 mm; range, 10–77 mm) were not significantly different (p = 0.215). The ADC values were significantly different between malignant lesions (0.73 × 10 −3 mm2 /s; range, 0.47–1.02 × 10 −3 mm2 /s) and benign lesions (1.21 × 10 −3 mm2 /s; range, 0.27–2.07 × 10 −3 mm2 /s) (p < 0.001). The lesion-to-parenchyma ADC ratios were also significantly different between malignant lesions (0.85 ± 0.04; range, 0.54–1.14) and benign lesions (1.43 ± 0.55; range, 0.30–3.30) (p < 0.001).

TABLE 1: Qualitative Analysis No. (%) of Lesions Benign (n = 42)

Malignant (n = 11)

Well defined

28 (66.7)

7 (63.6)

Partially defined

10 (23.8)

4 (36.4)

4 (9.5)

0 (0)

Hypointense

11 (26.2)

3 (27.3)

Isointense

31 (73.8)

8 (72.7)

Characteristic Margin

Ill defined

0.454

Signal intensity T1-weighted images

0.755

T2-weighted images

0.93

Hypointense

13 (31.0)

4 (36.4)

Isointense

9 (21.4)

2 (18.2)

Hyperintense

20 (47.6)

5 (45.4)

Gadoxetic acid–enhanced images Arterial phase Hypointense Isointense Hyperintense

0.015 22 (52.4)

11 (100.0)

1 (2.4)

0 (0)

19 (45.2)

0 (0)

Portal venous phase

< 0.001

Hypointense

20 (47.6)

11 (100.0)

Isointense

5 (11.9)

0 (0)

Hyperintense

17 (40.5)

0 (0)

Late phase (3 min)

< 0.001

Hypointense

11 (26.2)

10 (90.9)

Isointense

12 (28.6)

1 (9.1)

Hyperintense

19 (45.2)

0 (0)

1 (2.4)

0 (0)

Progressive hypovascular

21 (50.0)

11 (100.0)

Persistent hypervascular

16 (38.1)

0 (0)

Hypervascular washout

4 (9.5)

0 (0)

2 (4.8)

2 (18.2)

Isointense

18 (42.8)

4 (36.4)

Hyperintense

22 (52.4)

5 (45.4)

Enhancement pattern Only rim enhancement

0.028

Signal intensity Hepatobiliary phase

Evaluation of the Added Value of DiffusionWeighted Imaging for Differentiating Malignant From Benign Focal Splenic Lesions Diagnostic performance for the evaluation of malignant lesions improved significantly for both observers after they reviewed the DW images: A z improved from 0.774 to 0.983 for observer 1 and from 0.742 to 0.986 for observer 2 (p < 0.001) (Table 2). The diagnostic accuracy and sensitivity also improved significantly for both observers (p < 0.05) (Table 2). Both observers gave a score of 3 (indeterminate lesion) to most of the malignant splenic lesions (observer 1, 81.8% [9/11]; observer 2, 72.7% [8/11]) and about 808

pa

Hypointense

0.32

Diffusion-weighted images (b value = 800 s/mm2)

0.003

Hypointense

26 (61.9)

1 (9.1)

Isointense

10 (23.8)

4 (36.4)

Hyperintense

6 (14.3)

6 (54.5)

aThe Fisher exact test was used to calculate the p values.

half of the benign splenic lesions (observer 1, 50.0% [21/42]; observer 2, 59.5% [25/42]) on conventional MRI alone. However, on the

combined interpretation of DW images with ADC maps and conventional MR images, the number of lesions with a score of 3 diminAJR:203, October 2014


Addition of DWI for MRI of Focal Splenic Lesions ished markedly, and only one malignant lesion by observer 1 and two malignant lesions by observer 2 were given a score of 3 (Table 3). The combined interpretation of DW images with ADC maps and conventional MR images at the second reading session enabled the observers to correct several diagnostic errors made on the basis of the conventional MR images alone (observer 1, n = 10; observer 2, n = 8) (Table 4). On the retrospective image review (Table 4), six false-negative findings and one false-positive finding based on review of the conventional MR images alone were corrected by both observers during the review of the additional DW images. There was one false-negative finding, an angiosarcoma with no distinct diffusion restriction, that was not interpreted correctly by either observer on the additional DW image review (Fig. 5). Additionally, one false-positive finding, a tuberculous abscess with distinct diffusion restriction, was not interpreted correctly by either observer on the additional DW image review (Fig. 4). Interobserver agreement of confidence levels was moderate for conventional MRI alone (κ = 0.575) but was excellent for the combined set of DW and conventional MR images (κ = 0.840). Discussion Our results showed that the diagnostic accuracy of MRI for malignant splenic lesions increased significantly for both observers when DWI was added to conventional MRI. In addition, the sensitivity and interobserver agreement for the diagnosis of malignant splenic lesions were both higher after the addition of DWI to conventional MRI (sensitivity, [observer 1, 90.9%; observer 2, 81.8%]; κ = 0.840) compared with conventional MRI alone (sensitivity, [observer 1, 18.2%; observer 2, 27.3%]; κ = 0.575). In previous studies that reported the imaging findings of various splenic lesions on conventional unenhanced MRI, the signal intensity of malignant splenic lesions usually appeared as isointense signals on T1weighted images and as hypo-, iso-, or hyperintense signals on T2-weighted images [15, 16]. Those reported findings of malignant splenic lesions on T1- and T2-weighted images are in agreement with the findings in our study, and the signal intensities of malignant splenic lesions on T1- and T2-weighted images were not significantly different from those of benign splenic lesions in our study (p > 0.75). In the liver, malignant focal lesions, including lymphoma and metastasis, usually show lower signal intensity on T1-

TABLE 2: Comparison of Diagnostic Capability of Combined Conventional MRI and Diffusion-Weighted Imaging (DWI) With Conventional MRI Alone for the Prediction of Splenic Malignancy Performance Value

Conventional MRI Alone

Combined Conventional MRI and DWI

Observer 1 A z valuea (95% CI)

0.774 (0.638–0.877)

0.983 (0.902–1.000)

Accuracyb (%)

81.1 (43/53)

96.2 (51/53)

Sensitivityb (%)

18.2 (2/11)

90.9 (10/11)

Specificity (%)

97.6 (41/42)

97.6 (41/42)

PPV (%)

66.7 (2/3)

90.9 (10/11)

NPV (%)

82.0 (41/50)

97.6 (41/42)

Observer 2 A z valuea (95% CI)

0.742 (0.604–0.853)

0.986 (0.907–1.000)

Accuracyb (%)

81.1 (43/53)

94.3 (50/53)

Sensitivityb (%)

27.3 (3/11)

81.8 (9/11)

Specificity (%)

95.2 (40/42)

97.6 (41/42)

PPV (%)

60.0 (3/5)

90.0 (9/10)

NPV (%)

83.3 (40/48)

95.3 (41/43)

Note—Unless otherwise indicated, numbers in parentheses are raw data. Sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) were calculated according to the assumption that a confidence score of 4 or higher was considered positive for the diagnosis of splenic malignancy. A z = area under the ROC. aDifferences between two imaging sets with both observers were statistically significant (p < 0.001). bDifference between combined conventional MRI and DWI set and conventional MRI set alone by both observers was statistically significant (p < 0.05).

TABLE 3: Distribution of Diagnostic Confidence Scores According to Lesion Diagnosis as Benign or Malignant Confidence Scorea Lesion Diagnosis

1

2

3

4

5

Observer 1

0

0

9

2

0

Observer 2

0

0

8

3

0

Observer 1

0

0

1

4

6

Observer 2

0

0

2

0

9

Observer 1

4

16

21

1

0

Observer 2

3

12

25

2

0

Observer 1

32

9

0

0

1

Observer 2

35

6

0

0

1

Malignancy (n = 11) Conventional MRI alone

Combined conventional MRI and DWI

Benignancy (n = 42) Conventional MRI alone

Combined conventional MRI and DWI

Note—Data are numbers of lesions. DWI = diffusion-weighted imaging. aThe definitions of the confidence scores are as follows: A score of 1 indicated a definitely benign lesion; 2, probably benign lesion; 3, indeterminate lesion; 4, probably malignant lesion; and 5, definitely malignant lesion.

weighted images and higher signal intensity on T2-weighted images compared with that of normal hepatic parenchyma. However, in

the spleen, comparable contrast relationships between malignant focal lesions and background hepatic parenchyma cannot be ex-



Jang et al. pected because the spleen is hypointense to the liver on T1-weighted images and hyperintense to the liver on T2-weighted images as a result of the large fractional heme content of the spleen [16]. Additionally, in cases of primary vascular tumors of the spleen, which constitute most nonhematolymphoid splenic tumors, the MRI appearance reflects the hemorrhagic nature of these tumors [1], and the hemosiderin-laden portions of these tumors are often hypointense on T2-weighted images. Therefore, it may be difficult to distinguish between malignant and benign focal splenic lesions on conventional unenhanced MRI sequences. To improve the distinction between malignant and benign focal splenic lesions on MRI, contrast-enhanced dynamic MRI has been used [15]. In addition, recent reports suggested that contrast-enhanced ultrasound is useful for the differentiation of malignant from benign focal splenic lesions [28–30]. In those studies [15, 28–30], hypovascular enhancement or no enhancement of splenic lesions was suggestive of malignant lesions and the main indicator of a benign lesion was hyperenhancement. However, benign focal splenic lesions such as inflammatory pseudotumors or atypical hamartomas and hemangiomas can show hypovascular enhancement [31, 32], as was seen in our study in which 50% (21/42) of benign lesions showed hypovascular enhancement. In addition, metastasis from renal cell carcinoma may manifest as a hypervascular lesion in the spleen [33]. Therefore, differentiation of malignant from benign focal splenic diseases is difficult when solely based on conventional enhanced MRI findings [19–22]. In our study, both observers could not make a decision for malignancy or benignancy for splenic lesions on the basis of conventional MRI alone, and the majority of splenic lesions were considered to be indeterminate lesions (score of 3) by both observers (observer 1, 56.6% [30/53]; observer 2, 62.3% [33/53]). This finding may be because of various nonspecific conventional MRI findings of benign and malignant splenic lesions. However, with the addition of DWI to conventional MRI, observers were able to correct several diagnostic errors made on the basis of conventional MR images alone (observer 1, n = 10; observer 2, n = 8). Among all the splenic lesions analyzed, 41 of 42 (97.6%) benign focal splenic lesions and nine of 11 (81.8%) or 10 of 11 (90.9%) malignant lesions were correctly diagnosed by the observers after DWI was added to conventional MRI. Thus, our study

810

TABLE 4: Results Obtained With and Without Diffusion-Weighted Imaging (DWI) According to Lesion Diagnosis as Benign or Malignant No. of Lesions Correct Diagnosis Without DW Images

Incorrect Diagnosis Without DW Images

Malignancy

1

9

Benignancy

40

1

Malignancy

0

1

Benignancy

1

0

Malignancy

3

6

Benignancy

39

2

Malignancy

0

2

Benignancy

1

0

Malignancy

1

6

Benignancy

39

1

Malignancy

0

1

Benignancy

1

0

Lesion Diagnosis Observer 1 Correct diagnosis with DW images

Incorrect diagnosis with DW images

Observer 2 Correct diagnosis with DW images


Both observers Correct diagnosis with DW images


suggests that the combination of DWI and conventional MRI improves diagnostic confidence not only for malignant focal splenic lesions but also for benign focal splenic lesions and that unnecessary biopsy or splenectomy for benign lesions might be avoided. In our study, we defined the presence of a malignant splenic lesion as a lesion with hyper- or isointense signal on DW images (b = 800 s/mm2) and isointense or hypointense signal on the ADC map when compared with the signal intensity of the normal splenic parenchyma because the spleen is an organ with physiologically restricted diffusion [26]. Based on this definition, 10 of the 11 (90.9%) malignant splenic lesions in our study showed DWI findings that were consistent with malignancy. In addition to the qualitative results, malignant lesions showed a significantly lower mean ADC value (0.73 × 10 −3 mm2 /s) and lower tumor-to-splenic parenchyma ADC ratio (0.85) than those of benign lesions (1.21 × 10 −3 mm2 /s, 1.43, respectively). Therefore, we believe that the qualitative and quantitative analyses of DWI

may be helpful for differentiating malignant from benign focal splenic lesions. In our study, a false-positive lesion that was not interpreted correctly by either observer on the additional DW image review was identified as a tuberculous abscess with distinct diffusion restriction. An abscess, which is composed of viscous fluid-containing bacteria, inflammatory cells, mucoid proteins, and cellular debris, has the potential to show restricted diffusion on DWI and may be confused with a malignant lesion [34]. A false-negative lesion by both observers on the additional DW image review was an angiosarcoma with marked hypointense signal on T2-weighted and DW images due to the prominent siderotic components. Hemosiderin causes hypointensity on T2-weighted and DW images and susceptibility effects, and measurements of diffusion in areas with susceptibility effects can be problematic because of local field distortions [35]. There are some limitations to this study. First, although the patients in this study met the inclusion criteria, there might be selec-




A

B

C

D

Fig. 5—67-year-old woman with angiosarcomas in spleen. A–D, Compared with splenic parenchyma, tumors (arrows) have hypovascular enhancement on transverse gadoxetic acid–enhanced arterial phase image (A), show hypovascular progressive enhancement on 3-minute late phase image (B), and appear as markedly hypointense lesions on transverse fat-saturated T2-weighted image (C) and diffusion-weighted image obtained with b value of 800 s/mm2 (D). Lesion was considered as indeterminate splenic lesion on conventional MR images and after additional review of diffusion-weighted images by both observers.

tion bias because this study is retrospective. Second, the observers might have gained experience with the grading scheme, which may have contributed to the increased accuracy in the second reading session of the combined image sets. Third, although observers reviewed the image sets twice with a 1-month interval between readings and a reordering of cases for each reading, there is the possibility of a recall bias. Fourth, reviewers evaluated MR images side-by-side instead of using image fusion. This reading strategy might have caused errors in the localization of lesions on various MR image sequences because the small lesions may not have precisely localized on MR images. To lessen this potential error and to facili-

tate matching corresponding sites on different images, we used a spatial cursor key on our PACS. Fifth, our study had a small number of malignant lesions, and the number of malignant lesions might not make our results generalizable to the entire spectrum of malignant splenic lesions. Sixth, although we determined that a clinical follow-up period of longer than 24 months would be sufficient to exclude other diagnoses, 32 benign splenic lesions in our study were not confirmed surgically. Finally, the use of a relatively lower dosage of gadoxetic acid (0.025 mmol/kg of body weight) might have resulted in suboptimal contrast enhancement in the arterial and portal venous phases. However, the greater enhancement effect of gadoxetic acid on 3-T

MRI compared with 1.5-T MRI and the specific characteristics of gadoxetic acid, such as higher protein binding, result in increased relaxivity that compensates for the low gadolinium concentration [36, 37]; thus, the lower dosage may not have weakened the characteristic dynamic enhancement pattern of the spleen and splenic lesions in our study. In conclusion, the addition of DWI to conventional MRI improves the prediction of malignant splenic focal lesions compared with conventional MRI alone. References 1. Abbott RM, Levy AD, Aguilera NS, Gorospe L, Thompson WM. From the archives of the AFIP: primary vascular neoplasms of the spleen—radio-



Jang et al. logic-pathologic correlation. RadioGraphics 2004; 24:1137–1163 2. Gómez-Rubio M, López-Cano A, Rendón P, et al. Safety and diagnostic accuracy of percutaneous ultrasound-guided biopsy of the spleen: a multicenter study. J Clin Ultrasound 2009; 37:445–450 3. Lieberman S, Libson E, Maly B, Lebensart P, BenYehuda D, Bloom AI. Imaging-guided percutaneous splenic biopsy using a 20- or 22-gauge cutting-edge core biopsy needle for the diagnosis of malignant lymphoma. AJR 2003; 181:1025–1027 4. Civardi G, Vallisa D, Bertè R, et al. Ultrasoundguided fine needle biopsy of the spleen: high clinical efficacy and low risk in a multicenter Italian study. Am J Hematol 2001; 67:93–99 5. Di Stasi M, Buscarini L, Cavanna L, Rossi S, Buscarini E, Silva M. Complications of ultrasoundguided fine-needle biopsy of the spleen: report on 110 patients and review of the literature. J Interv Radiol 1996; 11:43–46 6. Robertson F, Leander P, Ekberg O. Radiology of the spleen. Eur Radiol 2001; 11:80–95 7. Paterson A, Frush DP, Donnelly LF, Foss JN, O’Hara SM, Bisset GS. A pattern-oriented approach to splenic imaging in infants and children. RadioGraphics 1999; 19:1465–1485 8. Bean MJ, Horton KM, Fishman EK. Concurrent focal hepatic and splenic lesions: a pictorial guide to differential diagnosis. J Comput Assist Tomogr 2004; 28:605–612 9. Carucci LR, Halvorsen RA. Abdominal and pelvic CT in the HIV-positive population. Abdom Imaging 2004; 29:631–642 10. Schininà V, Rizzi EB, Mazuoli G, David V, Bibbolino C. US and CT findings in focal splenic lesions in AIDS. Acta Radiol 2000; 41:616–620 11. Urrutia M, Mego PJ, Ro LH, Torres GM, Ros PR. Cystic masses of the spleen: radiologic-pathologic correlation. RadioGraphics 1996; 16:107–129 12. Ferrozzi F, Bova D, Draghi F, Garlaschi G. CT findings in primary vascular tumors of the spleen. AJR 1996; 166:1097–1101 13. Komatsuda T, Ishida H, Konno K, et al. Splenic lymphangioma: US and CT diagnosis and clinical manifestations. Abdom Imaging 1999; 24:414–417 14. Rabushka LS, Kawashima A, Fishman EK. Imag-

812

ing of the spleen: CT with supplemental MR examination. RadioGraphics 1994; 14:307–332 15. Luna A, Ribes R, Caro P, Luna L, Aumente E, Ros PR. MRI of focal splenic lesions without and with dynamic gadolinium enhancement. AJR 2006; 186:1533–1547 16. Elsayes KM, Narra VR, Mukundan G, Lewis JS Jr, Menias CO, Heiken JP. MR imaging of the spleen: spectrum of abnormalities. RadioGraphics 2005; 25:967–982 17. Ramani M, Reinhold C, Semelka RC, et al. Splenic hemangiomas and hamartomas: MR imaging characteristics of 28 lesions. Radiology 1997; 202:166–172 18. Ito K, Murata T, Nakanishi T. Cystic lymphangioma of the spleen: MR findings with pathologic correlation. Abdom Imaging 1995; 20:82–84 19. Basso SM, Sulfaro S, Marzano B, Fanti G, Chiara GB, Lumachi F. Incidentally discovered asymptomatic splenic hamartoma with rapidly expansive growth: a case report. In Vivo 2012; 26:1049–1052 20. Piccolo G, Cavallaro A, Antonacci V, et al. Laparoscopic splenectomy for splenic hamartoma in elderly patient: case report and review of the literature. Ann Ital Chir [Epub 2012 Oct 25] 21. Wang HL, Li KW, Wang J. Sclerosing angiomatoid nodular transformation of the spleen: report of five cases and review of literature. Chin Med J (Engl) 2012; 125:2386–2389 22. Falk GA, Nooli NP, Morris-Stiff G, Plesec TP, Rosenblatt S. Sclerosing angiomatoid nodular transformation (SANT) of the spleen: case report and review of the literature. Int J Surg Case Rep 2012; 3:492–500 23. Padhani AR, Liu G, Koh DM, et al. Diffusionweighted magnetic resonance imaging as a cancer biomarker: consensus and recommendations. Neoplasia 2009; 11:102–125 24. Taouli B, Koh DM. Diffusion-weighted MR imaging of the liver. Radiology 2010; 254:47–66 25. Perman WH, Balci NC, Akduman I, Kuntz E. Magnetic resonance measurement of diffusion in the abdomen. Top Magn Reson Imaging 2009; 20:99–104 26. Yoshikawa T, Kawamitsu H, Mitchell DG, et al. ADC measurement of abdominal organs and lesions using parallel imaging technique. AJR

2006; 187:1521–1530 27. Cohen J. Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull 1968; 70:213–220 28. Yu X, Yu J, Liang P, Liu F. Real-time contrastenhanced ultrasound in diagnosing of focal spleen lesions. Eur J Radiol 2012; 81:430–436 29. Stang A, Keles H, Hentschke S, et al. Differentiation of benign from malignant focal splenic lesions using sulfur hexafluoride–filled microbubble contrast-enhanced pulse-inversion sonography. AJR 2009; 193:709–721 30. Stang A, Keles H, Hentschke S, et al. Incidentally detected splenic lesions in ultrasound: does contrastenhanced ultrasonography improve the differentiation of benign hemangioma/hamartoma from malignant lesions? Ultraschall Med 2011; 32:582–592 31. Fernandez-Canton G, Capelastegui A, Merino A, Astigarraga E, Larena JA, Diaz-Otazu R. Atypical MRI presentation of a small splenic hamartoma. Eur Radiol 1999; 9:883–885 32. Bhatt S, Simon R, Dogra VS. Radiologic-pathologic conferences of the University of Rochester School of Medicine: inflammatory pseudotumor of the spleen. AJR 2008; 191:1477–1479 33. Nunes TF, Szejnfeld D, Miiji LN, Goldman SM. Isolated metachronous splenic metastasis from renal cell carcinoma after 5 years. BMJ Case Rep [Epub 2012 Dec 14] 34. Feuerlein S, Pauls S, Juchems MS, et al. Pitfalls in abdominal diffusion-weighted imaging: how predictive is restricted water diffusion for malignancy. AJR 2009; 193:1070–1076 35. Hiwatashi A, Kinoshita T, Moritani T, et al. Hypointensity on diffusion-weighted MRI of the brain related to T2 shortening and susceptibility effects. AJR 2003; 181:1705–1709 36. Goncalves Neto JA, Altun E, Elazzazi M, Vaidean GD, Chaney M, Semelka RC. Enhancement of abdominal organs on hepatic arterial phase: quantitative comparison between 1.5- and 3.0-T magnetic resonance imaging. Magn Reson Imaging 2010; 28:47–55 37. Soher BJ, Dale BM, Merkle EM. A review of MR physics: 3T versus 1.5T. Magn Reson Imaging Clin N Am 2007; 15:277–290
