Tomographic imaging of the spleen: the role of morphological and metabolic features in differentiating benign from malignant diseases

Clinical Imaging 36 (2012) 559–567

Tomographic imaging of the spleen: the role of morphological andmetabolic features in differentiating benign from malignant diseases

Pier Paolo Mainentia,⁎, Delfina Iodiceb, Immacolata Cozzolinoc, Sabrina Segretob,Sergio Capeceb, Giacomo Sicab, Mario Magliuloa, Giuseppe Cianciac,

Leonardo Paceb, Marco Salvatoreb

aIBB CNR, Napoli, Via Pansini 5, 80131, Naples, ItalybDipartimento di Scienze Biomorfologiche e Funzionali, Sezione di Radiologia, Università di Napoli “Federico II”, Via Pansini 5, 80131 Naples, Italy

cDipartimento di Scienze Biomorfologiche e Funzionali, Sezione di Anatomia Patologica, Università di Napoli “Federico II”, Via Pansini 5,80131 Naples, Italy

Received 14 November 2011; received in revised form 3 January 2012; accepted 5 January 2012

Abstract

To evaluate the tomographic features in differentiating benign from malignant splenic diseases, 54 patients with a cytohistologicalexamination and a contrast-enhanced multidetector computed tomography (ce-MDCT) and/or positron emission tomography/computedtomography (PET/CT) were retrospectively selected. Significant associations were observed between ce-MDCT Pattern 3 (focal hyperdenselesion) and Pattern 4 (infarcts/cysts) as well as PET/CT Pattern 3 (focal photopenia/diffuse uptakebliver) and benign pathologies, and betweence-MDCT Pattern 1 (splenomegaly without focal lesions) and Pattern 2 (focal hypodense lesion) as well as PET/CT Pattern 1 (diffuseuptake≥liver) and Pattern 2 (focal increased uptake) and malignant diseases. No significant association between benign or malignant diseasesand spleen volumes andmaximum standardized uptake value (SUVmax) as well as lesion characteristics was observed. Themetabolic data do notimprove the performance of morphological patterns.© 2012 Elsevier Inc. All rights reserved.

Keywords: Spleen; Benign disease; Malignant disease; CT; PET/CT

1. Introduction

Both contrast-enhanced multidetector computed tomogra-phy (ce-MDCT) and F18-FDG positron emission tomography/computed tomography (PET/CT) have a relevant role indiagnosis and characterization of abdominal pathologies.However, although a number of studies have investigated therole of both tomographic modalities in splenic disease [1–13],no clear conclusions have been reached so far. Themain reasonsare due to patient or disease selection, lack of histologicallyproven diagnosis, use of therapy response as the standard ofreference, and use of old-generation devices [1–13].

⁎ Corresponding author. Corso Vittorio Emanuele, 670, 80122 Naples,Italy. Tel.: +39 0817613060; fax: +39 0817616013.

E-mail address: [email protected] (P.P. Mainenti).

0899-7071/$ – see front matter © 2012 Elsevier Inc. All rights reserved.doi:10.1016/j.clinimag.2012.01.015

Morphological or metabolic abnormalities of the spleenmay be an incidental and unexpected finding on imagingstudies performed for unrelated causes or in a patient withouta definitive clinical diagnosis. In these cases, the clinicallyrelevant point is to define whether splenic changes are due toa benign or to a malignant process.

Therefore, we investigated the role of new-generationtomographic devices, both ce-MDCT and PET/CT, in thedifferential diagnosis of benign and malignant splenicpathologies using cytohistological examination as thestandard of reference.

Thus, the aims of our study were (1) to define ce-MDCT and PET/CT features capable of differentiatingbenign and malignant splenic pathologies; (2) to evaluatewhether PET/CT adds any additional information to ce-MDCT findings.

560 P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

2. Material and methods

2.1. Population

From January 2004 to December 2010, 54 patients whounderwent a diagnostic scan in our department±30 days ofhistology (either after splenectomy or ultrasound-guidedhistological or cytological examination) were retrospectivelyselected. Of these, 47 patients underwent a ce-MDCT and 28a PET/CT, 21 of whom underwent both examinations. Thus,the final study group consisted of 54 patients (26 F, 28 M;mean age: 54.8 years; age range: 11–79 years), whounderwent a ce-MDCT and/or a PET/CT and a splenectomy(n=35) or histological (n=13) or cytological (n=6) ultra-sound-guided examination in our hospital. Only 1 of 54patients had hepatic cirrhosis.

The following pathological diagnosis were observed:hamartoma (n=1), angiolipoma (n=1), hemangioma (n=1),multiple hemangiomatosis (n=2), littoral cell angiosarcoma(n=1), epidermoid cyst (3), pseudocyst (n=1), large hemor-rhagic necrosis in thalassemia (n=1) , multiple inflammatorylesions in systemic lupus erythematosus (n=1), granuloma-tous disease (n=1), inflammatory pseudotumor (n=1),congested spleen with infarctions (n=1), immunologicallyactivated spleen (n=5), extramedullary hematopoiesis (n=2),plasmacellular dyscrasia (n=1), thalassemia (n=1), Gaucherdisease (n=1), Hodgkin lymphoma (n=4), non-Hodgkinlymphoma (n=23), pancreatic cancer metastases (n=1), andcolon cancer metastases (n=1).

2.2. Imaging technique

2.2.1. Contrast-enhanced MDCTA 4-row or 64-row MDCT (Aquilion 4 and 64, Toshiba

Medical System, Otawara, Tochigi, Japan) was used. Scanswere acquired using the following parameters: (1) MDCT 4:4×3 mm beam collimation, pitch 5.5, 120 kV, 300 mA s,rotation time 0.5 s, and field of view to fit; images werereconstructed with a 3-mm effective thickness; (2) MDCT64: 2×16 mm beam collimation, pitch 5.5, 120 kV,modulated milliampere second, rotation time 0.5 s, andfield of view to fit; images were reconstructed with a 2-mmeffective thickness.

In all 47 patients, the procedure was performed 80 s(portal phase) after the intravenous bolus (3 ml/s) adminis-tration of 120 ml of an iodinated nonionic contrast agent,iopromide (Ultravist, 370 mg of iodine per milliliter; BayerSchering, Berlin, Germany), into an antecubital vein bymeans of an 18-gauge needle. Moreover, unenhanced CTimaging, arterial-phase (35 s), and equilibrium-phase (300 s)imaging of the abdomen were performed in 28, 32, and 38patients, respectively.

All images were transferred to a workstation equippedwith the Osirix software and evaluated using a basicabdominal (width: 360 HU; level: 60 HU) window settings,which could be modified by the readers.

2.2.2. Positron emission tomography/computed tomographyDual-modality imaging was performed with a PET/CT

system (Discovery-LS, GE-Medical-Systems, Milwaukee,WI, USA) consisting of a PET scanner and a four-rowMDCT system.

All patients had been instructed to fast for a minimum of6 h prior to the examination. Blood glucose levels were foundto be in the normal range prior to F18-FDG injection by bloodsampling. PET/CT examination was carried out 60 min afterintravenous administration of 370 MBq of F18-FDG.

MDCT scans were acquired from the base of the skullto the upper thighs using the following parameters: 4×5mm collimation (140 kV, 80 mA s), 0.5 s rotation time, apitch of 6.

PET data were acquired with the patient in the sameposition on the table at three to four bed positions (5 min foreach bed position) covering the same field of view as CT.

Data obtained from the CT acquisition were used forattenuation correction of PET emission data. PET imageswere reconstructed with a 4.5-mm thickness.

PET, CT, and fused PET/CT images were reviewed on aworkstation equipped with the Osirix software.

2.3. Image interpretation

The ce-MDCT images were interpreted by tworadiologists (PPM and SS, with 10 and 5 years ofexperience, respectively) and the PET/CT images by aradiologist and a nuclear medicine physician (GS and LP,both with 10 years of experience). The observers readindependently the images blinded to clinical and cytohis-topathological data and to the results of any other test. Incases of disagreement, a consensus panel consisting of thetwo original observers plus a third blinded party (MS, with20 years of experience) reviewed their interpretation andmade a final decision.

All ce-MDCT set of images were evaluated with a zoomdedicated to the spleen to prevent extra-splenic informationfrom possibly influencing the evaluation of the organ.

Each ce-MDCT observer received the portal phaseimages of the 47 patients to prevent the other phases frompossibly influencing the evaluation. Subsequently, theobservers received the 28 unenhanced, the 32 arterial, andthe 38 equilibrium scans in order to evaluate whether theycould change the portal phase splenic interpretation.

All PET/CT examinations were evaluated with a zoomdedicated to the spleen to prevent extra-splenic informationfrom possibly influencing the evaluation of the organ.The two observers had to describe the distribution of theradiotracer as diffuse or focal and had to measure the SUVmax.Only subsequently could they compare the splenic findingswith the liver.

The following MDCT parameters had to be described:

– splenic volume;– splenic morphological findings;

561cal Imaging 36 (2012) 559–567

The splenic morphological findings were classified

Table 1Relationship of the morphological and metabolic patterns with the benignand malignant pathology

CT Benign† Malignant⁎ PET/CT Benign† Malignant⁎

Pattern 1⁎ 5 9 Pattern 1⁎ 3 9Pattern 2⁎ 4 16 Pattern 2⁎ 2 9Pattern 3† 5 0 Pattern 3† 5 0Pattern 4† 8 0(⁎,†P=.0001) (⁎,†P=.004)

as follows:

Pattern 1: homogenous or patchy density without focallesions in splenomegaly;Pattern 2: single/multiple hypodense focal lesions;Pattern 3: single/multiple hyperdense focal lesions;Pattern 4: single/multiple infarctions or cystic lesions;

For Patterns 2 and 3, each observer had to describe thetotal number, the size of the largest lesion, the shape (round-oval, irregular), the density (homogenous, patchy), and thepresence of a capsule and/or calcifications.

The following PET/CT parameters had to be described:

– splenic volume;– splenic metabolic findings;– SUVmax (maximum standardized uptake value).

The splenic metabolic findings were classified as follows:

Pattern 1: splenic F18-FDG uptake superior or equal to thatof the liver (SUVmax of the spleen≥SUVmax of the liver);Pattern 2: single/multiple focal lesions with increasedF18-FDG uptake (SUVmax of the lesionsNSUVmax ofthe liver);Pattern 3: splenic F18-FDG uptake inferior to that of theliver without focal lesions (SUVmax of the spleenbSUVmax

of the liver) or single/multiple focal lesions with decreasedF18-FDG uptake (SUVmax of the lesionsbSUVmax of thesurrounding normal splenic parenchyma).

For both ce-MDCT and PET/CT, in the presence of twocoexisting patterns in the same patient the largest one had tobe indicated.

The splenic volume (expressed in cubic centimeters) wasautomatically calculated with the Osirix software whichtraced the splenic outline in each axial CT and PET/CTimage. A splenic volume greater than 450 cm3 wasconsidered as pathologically increased [1].

The SUVmax was calculated with the Osirix softwarewhich drew a region of interest with a diameter of two thirdsof the diameter of the spleen or of a focal lesion. A SUVmax

greater than 2.5 was considered as pathologically increased.

2.4. Statistical analysis

Chi-square analysis was performed to evaluate thepresence of a significant association of benign or malignantnature with splenic volume (≤450 cm3 and N 450 cm3),morphological and metabolic patterns, splenic SUVmax (≤2.5and N2.5), and, relatively to focal lesion, the number (single,multiple), size (≤5 cm and N 5 cm), shape (round-oval,irregular), and presence of a capsule and/or calcifications. TheMcNemar test for paired proportion was performed tocalculate the P value, as appropriate.

P.P. Mainenti et al. / Clini

A P value of less than .01 was consideredstatistically significant.

A two-tailed Student's t test for unpaired data was used toevaluate the difference of the continuous variables (volumeand SUVmax of the spleen, size and SUVmax of the focallesions) between benign and malignant nature. A P value ofless than .01 was considered statistically significant.

In the presence of a significant association between amorphological and/or a metabolic parameter, the sensitivity,specificity, accuracy, and positive (PPV) and negativepredictive values (NPV) were then calculated.

3. Results

The following splenicmorphologicalfindingswere observedat ce-MDCT: 14 cases with Pattern 1, 20 with Pattern 2, 5 withPattern 3, and 8 with Pattern 4. The 14 cases with Pattern 1consisted of 13 homogenous and 1 patchy densities. The 8 caseswith Pattern 4 consisted of 3 infarcts and 5 cystic lesions. Asignificant association (P=.0001) between ce-MDCT patternsand the presence of benign or malignant disease was observed(Table 1). Since all Patterns 3 and 4 (13/13)were associatedwithbenign pathology (Figs. 1A and B and 2A and B) and Patterns 1and 2 (25/34) withmalignant pathology (Figs. 3A andB and 4Aand B), all patients were grouped according to this finding.Table 2 shows the sensitivity, specificity, accuracy, PPV, andNPV obtained.

No significant associations between benign/malignantpathologies and the following features of focal lesions (Patterns2 and 3) were found: the number, the shape, the density, thepresence of a capsule and/or calcifications, and the size.

The following splenic metabolic findings were found atPET/CT: 12 cases with Pattern 1, 11 with Pattern 2, and 5with Pattern 3. The 12 Pattern 1 cases consisted of 9 patientswith splenic F18-FDG uptake higher than that of the liverand 3 with splenic F18-FDG uptake equal to that of the liver.The 5 Pattern 3 cases consisted of 4 patients with splenicF18-FDG uptake less than that of the liver and of 1 patientwith focal area of photopenia.

A significant association (P=.004) between PET/CTpatterns and the presence of benign or malignant diseasewas observed (Table 1). Since all Pattern 3 cases (5/5) wereassociated with benign pathology (Fig. 2C) and Pattern 1 and2 cases (18/23) with malignant pathology (Figs. 3C, 4C, 5Aand B), all patients were grouped according to this finding.

Fig. 1. (A,B) Splenic hamartoma. (A) ce-MDCT (portal phase): normal splenicvolume (185 cm3) with a homogenous hypervascular focal lesion (Morpho-logical Pattern 3). (B) Histological examination: flattened endothelial cells lineirregular vascular space, which is surrounded by disorganized red pulp tissue(EE, 20×).

Fig. 2. (A–C) Widespread area of hemorrhage in a thalassemic patient. (A)ce-MDCT (portal phase) coronal multiplanar reconstruction (MPR) image:increased splenic volume (2114 cm3) with the presence of focal areacharacterized by a huge hypodense cystic component and a small cranialhypervascular component (Morphological Pattern 4 for the prevalence ofcystic component). (B) Histological examination: splenic parenchymasubverted by widespread areas of hemorrhage (EE, 20×). (C) MPR Coronalreconstruction of a PET image: a huge photopenic area is observed in thespleen (Metabolic Pattern 3).

562 P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

Table 2 shows the sensitivity, specificity, accuracy, PPV,and NPV obtained.

No significant difference in splenic volume was observedbetween benign and malignant pathologies with ce-MDCTas well as with PET/CT (Table 3).

A mean SUVmax of 7.4±7 (range 0.4–30) was observed.No significant difference in SUVmax was found betweenbenign (4.6±4.1) and malignant pathologies (8.9±7.9)considering all the patients as well as stratifying for thepatients with and without focal splenic lesions (Table 4).

In the 21 patients who underwent both ce-MDCT andPET/CT scans, the accuracy of ce-MDCT (75%, 16/21patients) and that of PET/CT (81%, 17/21) were notsignificantly different (Table 5). Both modalities werelimited by false-positive findings. In a patient affected bysystemic lupus erythematosus with multiple splenic inflam-

Fig. 3. (A–C) Non-Hodgkin lymphoma with large B cells (CD20+). (A) ce-MDCT (portal phase): mild increased splenic volume (481 cm3) with thepresence of focal solid hypodense lesion, 40 mm in diameter (MorphologicalPattern 2). (B) Histological examination: large cells with round nucleus withsmall nucleolus and indistinct cytoplasm (EE, 20×); the cells areimmunolabeled with CD20 (inset, immunolabeled 40×). (C) Fused PET/CT image: the hypodense lesion shows a pathological tracer uptake(SUVmax 27) (Metabolic Pattern 2).

Fig. 4. (A–C) Extranodal non-Hodgkin lymphoma with pleomorphic Tcells. (A) ce-MDCT (portal phase), coronal MPR image: mild increasedsplenic volume (600 cm3) with homogenous parenchymal density(Morphological Pattern 1). (B) Histological examination: splenic infiltrationby large pleomorphic atypical cells with irregular multilobulated, sometimesmultiple, nucleus (EE, 10×); the cells have prominent eosinophilic nucleolusand moderate cytoplasm (inset, EE 40×). (C) Fused PET/CT coronal MPRimage: mild increased splenic volume with diffuse tracer uptake (SUVmax

4.5) superior to that of the liver (Metabolic Pattern 1).

563P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

Table 2Sensitivity, specificity, accuracy, positive (PPV), and negative (NPV)predictive value using Morphological Patterns 3 and 4 as benign and 1 and 2as malignant, and Metabolic Pattern 3 as benign and 1 and 2 as malignant

Morphology Metabolism

Sensitivity 100% Sensitivity 100%Specificity 59% Specificity 50%Accuracy 81% Accuracy 82%PPV 74% PPV 78%NPV 100% NPV 100%

Table 3Relationship of the splenic volume with the benign and malignant pathology

CT Volume (cm3) PET/TC Volume (cm3)

Benign⁎ 1060±772 (184–2829) Benign† 1102±674 (400–2186)Malignant⁎ 1142±888 (157–3800) Malignant† 1247±1001 (157–3800)

(⁎,†P=NS)

Values are shown as mean±S.D. (range).

564 P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

matory lesions and in another patient with splenic granulo-matous disease, the ce-MDCT showed multiple focalhypodense lesions (Morphological Pattern 2) which pre-sented radiotracer uptake with a SUVmax N10 on PET/CT(Metabolic Pattern 2) (Fig. 6A and B). In two patients withimmunologically activated spleen, the splenomegaly associ-ated with homogenous parenchymal density (Morphological

Fig. 5. (A, B) Littoral cell angiosarcoma. (A) Fused PET/TC image: increased splenof the liver (Metabolic Pattern 1). (B) Histological examination: on the right, dilendothelial cells with solid areas composed of spindle cells in bundles, Kaposi-like (atypical cuboidal endothelial cells (EE, 10×), sometimes with a hobnail aspect (in

Pattern 1) showed a diffuse increased F18-FDG uptakegreater than or equal to that of the liver (Metabolic Pattern 1).In these four cases both ce-MDCT and PET/CT erroneouslyclassified a benign splenic condition as malignant. Themetabolic information showed an added value with respectto the morphological findings in a patient with animmunologically activated spleen: the ce-MDCT scanoffered a Morphological Pattern 2, while the PET/CT aMetabolic Pattern 3.

ic volume (1450 cm3) with diffuse tracer uptake (SUVmax 30) superior to thatated vascular spaces lined by a single layer of atypical cuboidal or spindleEE, 10×); on the left, sinusoidal spaces with pseudopapillary pattern lined byset, EE 40×).

Table 4Relationship of SUVmax and the benign and malignant pathology considering all the patients and stratifying for the patients with and without focal splenic lesions

All patients SUVmax Patients without splenic focal lesions SUVmax Patients with splenic focal lesions SUVmax

Benign⁎ 4.6±4.1 (0.4–13) Benign† 3.2±1.4 (1.3–5.7) Benign‡ 7.8±6.6 (0.4–13)Malignant⁎ 8.9±7.9 (1.5–30) Malignant† 6.7±8.8 (1.5–30) Malignant‡ 11.1±6.5 (5–27)

(⁎,†,‡P=NS)

Values are shown as median±S.D. (range).

565P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

Using the 28 unenhanced, the 32 arterial, and the 38equilibrium scans, the CT readers changed the portal phasesplenic evaluation only in one case: the inflammatorypseudotumor showed a Morphological Pattern 2 on theportal phase scan which changed into Pattern 3 on theequilibrium phase (Fig. 7A–C).

4. Discussion

In the present study, a similar performance in differenti-ating benign from malignant splenic diseases was shown byce-MDCT and PET/CT. Although both morphological andmetabolic pattern classifications had a quite high accuracy,the specificity and PPV were unsatisfactorily low. The false-positive rate did not significantly decrease even with theaddition of other information as splenic volume and SUVmax,or specific characteristics of the focal lesions (i.e., number,size, shape, density, presence of a capsule and/or calcifica-tion, SUVmax).

Several points have to be considered in the evaluation ofsplenic disease:

– The majority of cystic and solid lesions without theevidence of a primary neoplasm are benign;

– The malignant lesions are less common than the benignand are often observed in a well-defined clinical contextas patients with lymphoma or metastatic disease;

– A primary vascular neoplasm may be hypothesizedin the absence of findings suggestive of metastasisor lymphoma;

– The increase of splenic volume accompanies severalinfective, inflammatory, hematologic, thesaurismotic,and oncological diseases.

A splenic pathological finding may be an incidental andan unexpected event on imaging studies performed for

Table 5Sensitivity, specificity, accuracy, PPV, and NPV using MorphologicalPatterns 3 and 4 as benign and 1 and 2 as malignant, and Metabolic Pattern 3as benign and 1 and 2 as malignant in the 21 patients who underwent bothce-CT and F18-FDG-PET/CT modalities

Morphology Metabolism

Sensitivity 100% Sensitivity 100%Specificity 37% Specificity 50%Accuracy 76% Accuracy 81%PPV 72% PPV 76%NPV 100% NPV 100%

(P=NS)

unrelated causes. In these cases, the main question is on thebenign or malignant nature of the splenic involvement, andthe above-mentioned features may be helpful but not usedfor a definite diagnosis of benign or malignant involvementleading to the appropriate clinical management.

Few studies have investigated the impact of morphologicaland metabolic features in defining the benign or malignantnature of a focal or a diffuse splenic disease. Goerg et al. [14]

Fig. 6. (A, B) Inflammatory systemic lupus erythematosus lesions. (A) FusedPET/TC image: normal splenic volume (410 cm3) with focal areas of traceruptake (SUVmax 10.7) (Metabolic Pattern 2). (B) Histological examination:fibrinoid necrosis (yellow arrow) bordered by engulfing histiocytes andplasma cells (yellow arrowhead) in continuity with the splenic red pulp(green arrow) (EE 10×).

Fig. 7. (A–C) Inflammatory pseudotumor. (A) CT (portal phase): normalsplenic volume (237 cm3) with a focal hypodense lesion. (B) CT (equilibriumphase): the splenic lesion shows a mild hyperdensity. (C) Histologicalexamination: pseudoangiomatous aspect with “expansion” of the red pulp andpresence of fibrosis (EE, 20×).

566 P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

evaluated the role of ultrasound in differentiating benign frommalignant splenic lesions considering the splenic size, the echopattern, the size, and the number of the focal lesions. The

authors concluded that a differential diagnosis was oftenimpossiblewithout contributory clinical data.Metser et al. [13]assessed the role of 18-FDG-PET/CT in evaluating solidsplenic masses considering the lesion size, the presence of asingle vs. multiple splenic lesions, and the 18-FDG uptakeexpressed as SUVmax. The authors observed a significantdifference between the SUVmax of benign and malignantlesions and concluded that 18F-FDG PET can reliablydiscriminate between benign and malignant solid splenicmasses. On the contrary, in our population, we did not find asignificant difference between the SUVmax of malignant andbenign lesions considering the whole population as well asstratifying for patients with and without focal lesions. Themismatch may be in part explained by the followingobservations: Metser et al. [13] included only patients withsplenic focal solid masses and cytohistological examinationwas available only in 16 of 88 patients, while the clinical andimaging follow-up were used in all other patients; in thepresent study, all patients had a cytohistological diagnosis andboth diffuse and focal splenic lesions were included.

In our series, the morphological and metabolic patternswere more accurate than all the other parameters consideredin differentiating benign from malignant splenic diseases;however, they were limited by the false-positive findings.The rate of false-positive findings did not decreasesignificantly when matching the pattern classifications withthe other morphological or metabolic parameters.

On the basis of this study's results, the splenomegaly ofunknown origin and the solid hypodense masses on ce-MDCTrepresent a diagnostic challenge because they may begenerated by malignancies as well as infective or noninfectivegranulomatous/inflammatory diseases or nonneoplastic im-munological/hematological processes. In these challengingcases, PET/CT may exclude the presence of a malignancywhen neither 18F-FDG avid lesions nor a diffuse tracer uptakeequal or superior to that of the liver is observed. Anyway, itshould be borne in mind that non-18F-FDG avid tumors, suchas some renal cell carcinomas or thyroid cancers, maymetastasize to the spleen [13].

A few more points need further consideration:

(1) In our series the three lesions with calcifications wereepidermoid cyst (n=2) or pseudocyst (n=1). Calcifi-cations have been often described in benign spleniclesions as cyst, pseudocyst, hamartoma, infarction,hemangioma, inflammatory pseudotumor, Pneumo-cystis carinii infection, tuberculosis, and histoplas-mosis, while they have been very seldom associatedwith malignant lesions such as metastases fromovarian cancer [15].

(2) The five lesions appearing as hyperdense (Pattern 3)on ce-MDCT were benign vascular neoplasms infour cases and a hamartoma in the remaining case.Moreover, the inflammatory pseudotumor appearedas hyperdense in the equilibrium phase. As a result,Pattern 3 is highly predictive of a benign pathology.

567P.P. Mainenti et al. / Clinical Imaging 36 (2012) 559–567

On the other hand, the only malignant vascularneoplasm of our series appeared as a complex masswhich showed a marked enhancement of theperiphery and a prevalent large central hypodensecomponent. The prevalence of the hypodensecentral component allowed us to describe thelesion with Morphological Pattern 2. Previous CTreports of angiosarcoma have demonstrated similarfindings [10].

(3) Splenic infarction may occur in patients withmyeloproliferative, lymphoproliferative, hematologic,infective, and thromboembolic diseases [14]. As aconsequence, a malignancy has to be excluded in thepresence of an infarction of the spleen.In our series, the three patients with splenicinfarction presented with a congested spleenassociated respectively with a jejunal lymphoma,extramedullary hematopoiesis, and thalassemia.Moreover, a patient with splenic infarctions showedlarge, prevalent, hypodense, non-Hodgkin lympho-matous lesions. The prevalence of the hypodensecomponents allowed us to associate the patient withMorphological Pattern 2.

(4) A CT scan of the spleen is usually performed with amultiphase technique; however, a splenic patholog-ical finding may be an incidental and an unexpectedevent on CT scan executed with only a portal phaseacquisition. In this case, it may raise the diagnosticdilemma if it is necessary to repeat the study with amultiphase technique. In our population, theunenhanced and arterial phases did not show anyadded value with respect to the portal phase in anypatient, while the equilibrium phase changed theportal pattern only in one case.

(5) The aim of our study was to differentiate benignfrom malignant splenic diseases exclusively on thebasis of the splenic tomographic imaging; for thisreason, we did not evaluate the impact either of theclinical and laboratory information or of theextrasplenic imaging findings.

The following limits, mainly related to the retrospectivedesign of the study, have to be acknowledged:

(1) ce-MDCT was performed on either a 4-row or a64-row MDCT, which may have a differentdiagnostic impact.

(2) The CT examinations were not necessarily executedwith a protocol dedicated to the spleen but on the basisof the specific clinical question. This point explainsthe lack of uniformity of the CT acquisition protocols.

(3) The morphological and metabolic parameters weredichotomized into benign or malignant excludingthe possibility to consider a finding as “uncertain”.

This type of analysis may have caused the relativelyhigh rate of false-positive findings.

(4) The direct comparison of ce-MDCT and F18-FDG-PET/CT was available in 21 of 54 patients.Prospective studies in larger populations have tobe designed to confirm our observations.

In conclusion, our data suggest that, in differentiatingbenign from malignant splenic diseases, (1) the patternclassification is more accurate than the other morphologicaland metabolic parameters; (2) CT and PET/CT present asimilar performance; (3) the metabolic features do not showa significant added value with respect to the morphologicalones; however, the high NPV intrinsic to PET/CT modalitymay help to exclude the presence of a malignancy.

References

[1] de Jong PA, van Ufford HM, Baarslag HJ, de Haas MJ, Wittebol SH,Quekel LG, de Klerk JM. CT and 18F-FDG PET for noninvasivedetection of splenic involvement in patients with malignant lymphoma.AJR Am J Roentgenol 2009;192:745–53.

[2] Karaosmanoglu DA, Karcaaltincaba M, Akata D. CT and MRIFindings of sclerosing angiomatoid nodular transformation of thespleen: spoke wheel pattern. Korean J Radiol 2008;9(Suppl):S52–5.

[3] Taylor A, Dodds WJ, Erickson SJ, Stewar ET. CT of the acquiredabnormalities of the spleen. AJR Am J Roentgenol 1991;157:1213–9.

[4] Wang YJ, Li F, Cao F, Sun JB, Liu JF, Wang YH. Littoral cell angiomaof the spleen. Asian J Surg 2009;32:167–71 art.

[5] Rabushka LS, Kawashima A, Fishman Elliot K. Imaging of the spleen:CTwith supplemental MR examination. Radiographics 1994;14:307–32.

[6] Mirowitz S, Brown JJ, Lee JKT, Heiken JP. MR Dynamic gadolinium-enhanced, imaging of the spleen: normal enhancement patterns andevaluation of splenic lesions. Radiology 1991;179:681–6.

[7] Elsayes KM, Narra VR, Govind Mukundan G, Lewis JS, Menias CO,Heiken JP. MR Imaging of the spleen: spectrum of abnormalities.Radiographics 2005;25:967–82.

[8] Abbott RM, Levy AD, Aguilera NS, Gorospe L, ThompsonWM. Fromthe Archives of the AFIP. Primary vascular neoplasms of the spleen:radiologic–pathologic correlation. Radiographics 2004;24:1137–63.

[9] Levy AD, Abbott RM, Abbondanzo SL. Littoral cell angioma of thespleen: CT features with clinicopathologic comparison. Radiology2004;230:485–90.

[10] Thompson WM, Levy AD, Aguilera NS, Gorospe L, Abbott RM.Angiosarcoma of the spleen: imaging characteristics in 12 patients.Radiology 2005;235:106–15.

[11] Semelka RC, Shoenut P, Lawrence PH, Greenberg HM, Madden TP,Kroeker MA. Spleen: dynamic enhancement patterns on gradient-echoMR images enhanced with gadopentetate dimeglumine. Radiology1992;185:479–82.

[12] Ryu JC, Won Um J, Min WB. Inflammatory pseudotumour of the spleen:the findings on F-18 fluorodeoxyglucose positron emission tomography/-computed tomography (FDG-PET/CT). ANZ J Surg 2010;80:650–2.

[13] Metser U, Miller E, Kessler A, Lerman H, Lievshitz G, Oren R, Even-Sapir E. Solid splenic masses: evaluation with 18F-FDG PET/CT. JNucl Med 2005;46:52–9.

[14] Goerg C, SchwerkWB,GoergK. Splenic lesions: sonographic patterns,follow up, differential diagnosis. Eur J Radiol 1991;13:59–66.

[15] Wan YL, Cheung YC, Lui KW, Tseng JH, Lee TY. Ultrasonographicfindings and differentiation of benign and malignant focal spleniclesions. Postgrad Med J 2000;76:488–93.