September 2009, VOLUME 193

Recommend & Share

September 2009, Volume 193, Number 3

Gastrointestinal Imaging

Original Research

Comparative Performance of MDCT and MRI With MR Cholangiopancreatography in Characterizing Small Pancreatic Cysts

+ Affiliations:
1Division of Abdominal Imaging and Interventional Radiology, Department of Radiology, Massachusetts General Hospital, 55 Fruit St., White 270, Boston, MA 02114-2696.

2Department of Pathology, Massachusetts General Hospital, Boston, MA.

3Department of Surgery, Massachusetts General Hospital, Boston, MA.

Citation: American Journal of Roentgenology. 2009;193: 722-731. 10.2214/AJR.08.1253

Next section

OBJECTIVE. The objective of our study was to compare MDCT with MRI–MR cholangiopancreatography (MRCP) in characterizing small pancreatic cysts (≤ 3 cm) and predicting aggressiveness.

MATERIALS AND METHODS. In a retrospective analysis, contrast-enhanced MDCT and MRI examinations of 30 patients with 38 pathologically confirmed small pancreatic cysts were reviewed. MDCT and MRCP studies were independently evaluated by two readers for cyst morphology, cyst characterization, and prediction of lesion aggressiveness, which included lesions with moderate-grade dysplasia, high-grade dysplasia (carcinoma in situ), and invasive carcinomas. The sensitivity of MDCT and MRI for the detection of each morphologic feature, accuracy for cyst characterization, and predictive values for aggressiveness were calculated.

RESULTS. Of 38 lesions, 14 were side-branch intraductal papillary mucinous neoplasms (IPMNs), 12 mixed IPMNs, six mucinous cystic neoplasms, and six nonneoplastic cysts. On histopathology, 26 lesions were nonaggressive (six nonneoplastic cysts, six benign mucinous cystic neoplasms, 14 low-grade dysplasias in IPMNs), whereas 12 lesions revealed aggressive biology (eight moderate-grade dysplasias, four high-grade dysplasias in IPMNs). The sensitivity of MRCP for the detection of morphologic features was better than that of MDCT, but the differences were not statistically significant (p = 0.25–1). Interreader agreement and MDCT–MRI agreement for morphologic features were good to perfect (κ = 0.7–1). The accuracy of MDCT and MRI was higher in classifying cysts as mucinous or nonmucinous than in determining a specific diagnosis (71–84.2% vs 39.5–44.7%, respectively), whereas the accuracy of the two techniques in characterizing cysts into nonaggressive and aggressive categories was similar (MDCT vs MRI, 75–78% vs 78–86%, respectively; p > 0.05).

CONCLUSION. MRI enables more confident assessment of the morphology of small cysts than MDCT, but the accuracy of the two imaging techniques for cyst characterization is comparable. MDCT and MRI have high accuracy in classifying cysts into mucinous and nonmucinous categories and perform similarly in estimating histologic aggressiveness.

Keywords: MDCT, MR cholangiopancreatography, pancreatic cancer, pancreatic cysts

Previous sectionNext section

Advances in cross-sectional imaging technology have enhanced the detection of pancreatic cystic lesions; many of these cystic lesions are small (≤ 3 cm) at detection. The incidence of malignancy in small pancreatic cysts is low, ranging from 0% for invasive carcinomas to 20% for malignancy of any type [14]. Most small lesions are detected incidentally, and the incidence of malignancy in asymptomatic individuals is as low as 3.5% [1, 2]. The morphologic features of cysts such as a solid component or mural nodule, thick septa, and main pancreatic duct (MPD) dilatation are recognized predictors of aggressive biology that often dictate more aggressive management [2, 58]. The superior soft-tissue and contrast resolution inherent to MRI renders MRI better than other noninvasive imaging tests such as CT for evaluating the morphologic features of cysts including cyst communication [911]. Nevertheless, MDCT, because of its high spatial resolution and superior quality of off-axial image displays, is often the first line of investigation for pancreatic lesions and is being increasingly used in the management decisions of these lesions. Despite the prevalent use of MDCT, few articles about studies comparing MDCT and MRI for the evaluation of cystic pancreatic lesions have been published [1215]. Moreover, the value of MDCT and MRI in characterizing small cystic pancreatic lesions and the capacity of these techniques to distinguish nonaggressive from aggressive lesions have not been studied to our knowledge. These issues are clinically significant because the morphologic features of small cysts are important determinants of management.

The purpose of our study was to compare the performance of MDCT and MR cholangiopancreatography (MRCP) in evaluating the morphologic details of small pancreatic cysts and to compare the accuracy of the two techniques for characterizing the lesions into histologic types and predicting aggressiveness and malignancy.

Materials and Methods
Previous sectionNext section
Study Design

Institutional review board approval was obtained for this retrospective study, and informed patient consent was waived. This study was compliant with HIPAA.

A computerized database of radiology reports (Folio Views 4.2, Next Page) was reviewed by a radiology fellow for the time period of January 2000 through December 2007. The search criteria included the terms “pancreas,” “pancreatic,” “cyst,” “cysts,” “cystic,” “CT,” and “MR” to identify patients who underwent both CT and MRI for the evaluation of pancreatic cystic lesions. Our initial search yielded 448 patients, and a review of the images on a PACS workstation (Agfa, version 4.0) for lesions with a diameter of ≤ 3 cm produced a subset consisting of 152 patients. We based inclusion on cyst size measured using digital calipers at imaging because size measurements of surgical specimens and at pathology are usually imprecise because of cyst collapse or rupture. A further restriction was to consider only patients having undergone CT and MRI within a time interval of 90 days followed by histopathologic analysis (surgery or endoscopic ultrasound with guided fine-needle aspiration [FNA]) of the cystic lesion within 6 weeks of imaging.

Thirty patients (17 women, 13 men; age range, 44–86 years; mean age, 68.1 years) fulfilled the study criteria. The other 122 patients were excluded for the following reasons: in three patients, the interval between CT and MRI was more than 3 months, four patients had marginal change in cyst size during the interval between the CT and MRI examinations, 34 patients had clinical or imaging evidence of acute pancreatitis, 26 patients had more than one cyst and the dominant cyst measured more than 3 cm, 16 patients had prior surgical procedures in the pancreas, and 39 patients had no histopathologic proof of diagnosis.

For patients with multiple small pancreatic lesions, only the lesions evaluated pathologically were included, thus resulting in inclusion of 38 small cystic lesions in 30 patients. All of the included patients had undergone CT and MRI with MRCP as part of their clinical evaluation. An upper limit of 90 days between CT and MRI was imposed to avoid the influence of any change in cyst size or morphology over time. Small pancreatic lesions tend to grow slowly and have been reported not to show significant morphologic changes in a 6-month time period [4, 16, 17]. Therefore, we assume that any change was unlikely to occur in a 90-day period, which was confirmed quantitatively and qualitatively between the two studies for all the included patients.

The patients' electronic medical records were reviewed to study clinical details, surgical findings, pathology reports, and information from other pertinent investigations, and these details were recorded in a standardized database. Clinical features that were recorded included symptoms that could be related to pancreatic cysts such as abdominal pain, jaundice, diarrhea with malabsorption, or bloating. When the cysts were discovered on imaging performed for evaluation of other symptoms, they were called incidental.

Imaging Protocol

MDCT examination—All CT examinations were performed on MDCT scanners (4- and 16-MDCT HiSpeed QX/i and LightSpeed scanners, GE Healthcare). CT images from a focused thin-section dual-phase pancreas protocol were available in 17 patients, whereas the other 13 patients had only a single portal venous phase examination. For the dual-phase pancreas protocol, nonionic iodinated contrast material (300 or 370 mg I/mL; 100–150 mL depending on patient weight) was injected IV through a power injector at a rate of 4–5 mL/s; 1.25-mm-thick pancreatic phase images were acquired craniocaudally through the pancreas 40 seconds after contrast injection, and 5-mm-thick portal venous phase images were acquired craniocaudally from the diaphragm to the pubic symphysis 70 seconds after contrast injection.

For the single portal venous phase acquisitions, contrast material was injected at a rate of 2–3 mL/s, and 5-mm-thick portal venous phase images were acquired from the diaphragm to the pubic symphysis 70 seconds after the onset of contrast injection. Reformatted images of 2.5- to 3-mm thickness in coronal planes were reconstructed for all the patients. In patients with a pancreas protocol study (n = 17), image postprocessing was undertaken on a commercially available workstation (Advantage 4, GE Healthcare) by a trained technologist. A standard protocol was used for all patients that included 5-mm coronal and oblique multiplanar reconstructions along the body and tail of pancreas, 1-mm maximum intensity projections, and one or two curved multiplanar reformations along the course of the pancreatic duct.

MR examination—MRI examinations were performed on a 1.5-T system (Signa, GE Healthcare) using a phased-array torso coil. Of the 30 patients, 24 underwent MRI after contrast medium administration, whereas the MRI examinations of the remaining six patients were unenhanced. The following imaging sequences were acquired: coronal (breath-hold) and axial (respiratory-triggered) T2-weighted fast spin-echo (FSE) sequences (TR/TE, 5,000/100; flip angle, 90°; field of view [FOV], 34 cm; slice thickness, 3 mm; matrix, 256 × 192); axial T1-weighted spoiled gradient-echo sequences; and breath-hold in-phase (150/4.2) and opposed-phase (150/2.1) sequences (flip angle, 90°; FOV, 36 cm; slice thickness, 4 mm; matrix, 256 × 160). These sequences were followed by coronal and oblique coronal half-Fourier single-shot FSE breath-hold acquisition for thick-section 2D MRCP (n = 30) (1,650/800; slice thickness, 50–60 mm; FOV, 26 cm; matrix, 256 × 256). Subsequently, fat-suppressed fast spoiled gradient-echo dynamic breath-hold MR images (n = 24) (150/2; flip angle, 70°; FOV, 36 cm; slice thickness, 4 mm; matrix, 256 × 160) through the pancreas after injection of 0.1 mmol of gadopentetate dimeglumine (Magnevist, Bayer HealthCare) per kilogram of body weight were obtained. The acquisitions included the arterial phase at 20 seconds, portal venous phase at 70 seconds, and equilibrium phase at 180 seconds after injection of IV contrast material. Thin-section respiratory-triggered 3D MRCP (n = 15) (3,750/650; slice thickness, 1.5 mm; FOV, 40 cm; matrix, 256 × 256) was performed after contrast-enhanced acquisitions.

Image analysis—The CT and MR images were reviewed independently by two subspecialty-trained radiologists (reviewer 1 with 15 years and reviewer 2 with 8 years of experience) on a PACS workstation. The radiologists were aware of patient age and sex but were blinded to the clinical findings and other observations made at surgery, endoscopic ultrasound, and pathology. To minimize interpretation bias, the CT examinations were reviewed first; an interval of 4 weeks then elapsed before they reviewed the MR examinations. A template was designed to record the cyst morphology details.

Each reader recorded the number, size, and location of each cyst along with morphologic features such as septa (thin-hairline or thick [i.e., thicker than hairline]), wall thickness, mural nodule, calcification, pancreatic duct communication, and MPD dilatation. In patients with multiple cysts, the morphologic features were recorded separately for each cyst. Other features such as biliary duct dilatation attributable to the cystic lesion, enlarged peripancreatic lymph nodes (short-axis diameter > 10 mm), vascular involvement, and metastases were recorded.

Based on the published imaging criteria for cyst morphology and characterization [3, 8, 13, 1821], the lesions were first classified as mucinous or nonmucinous. Reader confidence for this classification was rated on a 5-point scale: 1, definitely nonmucinous; 2, probably nonmucinous; 3, indeterminate; 4, probably mucinous; or 5, definitely mucinous. A specific diagnosis was then assigned based on established criteria in the literature [13, 18, 20, 22], and reader confidence for that diagnosis was rated as follows: 1, definitely not; 2, possibly not; 3, indeterminate; 4, probably; or 5, definitely. The lesions were finally categorized as benign or malignant, and diagnostic confidence was assigned using the following scale: 1, definitely benign; 2, probably benign; 3, indeterminate; 4, probably malignant; or 5, definitely malignant. The following features, previously described in the literature [3, 57, 14, 23, 24], were considered to assign cysts to the malignant category: thick wall, thick septations, mural nodule, MPD dilatation of > 8 mm, vascular encasement, peripancreatic lymphadenopathy, and metastases.

MRI examinations were also independently evaluated systematically, as described for CT. All the imaging findings were recorded on a defined template and were later transferred to the standardized database for analysis by another investigator.

Standard of Reference

Histopathologic evaluation of surgically resected cysts served as a standard of reference for confirming the morphologic features, diagnosis, and cyst biology in 21 patients (29 cysts). For the remaining nine patients (nine cysts), endoscopic ultrasound served as an alternate reference standard for confirming the morphologic features, and endoscopic ultrasound with FNA with fluid analysis served to confirm the diagnosis and presence of malignancy [25, 26]. The time interval between imaging and pathologic evaluation varied from 10 to 42 days (mean, 23 days).

Surgery and Histopathologic Analysis

The type of surgical procedure performed was recorded. A single blinded gastrointestinal pathologist with more than 8 years of experience reviewed all specimens and slides. Gross pathologic features of the cysts, including septa, mural nodules, and ductal communication, were noted. Microscopically, the lesions, including muci nous cystic neoplasms, were categorized as benign or malignant. The intraductal papillary muci nous neoplasms (IPMNs) were categorized according to the World Health Organization classification system as low-grade dysplasia (adenoma), moderate-grade dysplasia, high-grade dysplasia (carcinoma in situ [CIS]), or invasive carcinoma (malignant) [27]. When more than one type of biologic activity was noted at the cellular level in an IPMN, the lesion was considered to belong to the most biologically aggressive category. The presence of malignancy in the resected lymph nodes was recorded.

Endoscopic Ultrasound with FNA

One of three experienced gastroenterologists with expertise in endoscopic procedures performed the endoscopic ultrasound examinations with FNA using a radial or linear scanning endoscope (GFUM-20, Pentax) at 7.5 and 12.0 MHz. The cystic lesions were evaluated for diameter, septa, wall thickness, mural nodule, internal debris, ductal communication, vascular involvement, and lymphadenopathy. Cyst fluid was aspirated using a 19- or 22-gauge needle with a stylet from a transgastric approach. The cyst fluid was evaluated for cytology, amylase, and the tumor marker carcinoembryonic antigen.

Statistical Analysis

The morphologic features detected on MDCT and MRCP by both readers were noted and interreader agreement (kappa value) was calculated using statistical software (SAS, version 9.1.3, SAS Institute). Agreement of MDCT and MRCP for the detection of each morphologic feature was recorded, and kappa values were calculated using the same statistical software. Differences in sensitivity of MDCT and MRCP for the detection of various morphologic features were calculated using the McNemar test. Differences in sensitivity of recognition of various morphologic features with and without the use of the dual-phase pancreatic protocol CT were computed using the chi-square test. The accuracy of both readers for classifying lesions as mucinous or nonmucinous, assigning a specific histologic diagnosis, and categorizing lesions as nonaggressive or aggressive with MDCT and MRI was evaluated.

Based on reader confidence on MDCT and MRCP for cyst characterization as nonaggressive or aggressive, a parametric receiver operating characteristic (ROC) curve was constructed using ROCKIT software (version 0.9.1 beta, Metz CE et al., University of Chicago, Chicago, IL) for reviewers 1 and 2 to calculate accuracy (area under the ROC curve [Az]). The lesions with biologic activity of moderate-grade dysplasia, high-grade dysplasia (CIS), and invasive carcinoma on pathology were included in the aggressive category, whereas all the benign lesions and those with low-grade dysplasia (adenoma) were included in the nonaggressive category. Because of sample size limitations, ROC curve analysis for histologic diagnosis was not performed. The predictive values of MDCT and MRCP for determining malignancy and benignity were also calculated. A p value of < 0.05 represented statistical significance.

Previous sectionNext section

The time interval between CT and MR examinations varied from 1 to 90 days (mean, 30.1 days). The cysts ranged in size from 9 to 30 mm (mean, 16.8 mm). Furthermore, seven lesions were ≤ 10 mm, 23 were 11–20 mm, and eight were 21–30 mm. Five patients had more than one pathologically verified cyst: Three patients had three cysts, and the other two patients had two cysts. Seventeen cysts were located in the proximal pancreas (head, uncinate, or neck), whereas the remaining 21 were in the distal pancreas (body and tail). Of the 21 surgically treated patients, Whipple procedure was performed in 11 patients, seven underwent distal pancreatectomy, and three underwent middle pancreatectomy. The type of resection was based on lesion location, multiplicity, MPD involvement, and findings at frozen-section analysis with regard to the presence and extent of malignancy. The samples for endoscopic ultrasound with FNA performed in nine cysts in patients who did not undergo resection were reported to be adequate.

Pathologically, of 38 cysts, 14 (36.8%) were side-branch IPMNs, 12 (31.6%) were mixed (side-branch and main duct type) IPMNs, six (15.8%) were mucinous cystic neoplasms, and six (15.8%) were nonneoplastic cysts, either a pseudocyst or retention cyst (Table 1). Of 38 cysts, 12 (31.6%) were benign, 14 (36.8%) revealed low-grade dysplasia, eight (21.1%) had moderate-grade dysplasia, four (10.5%) had high-grade dysplasia (CIS), and none revealed invasive carcinoma (Table 1). High-grade dysplastic (CIS) changes were detected in mixed IPMNs, whereas low- and moderate-grade dysplastic changes were noted in both side-branch and mixed IPMNs (Table 1). Cysts were detected incidentally in 21 of the 30 patients (70%), whereas the remaining nine patients (30%) reported abdominal discomfort. Among the 21 patients with incidentally detected lesions, seven had benign lesions, seven had low-grade dysplasia, five had moderate-grade dysplasia, and two had high-grade dysplastic (CIS) changes. Of the nine symptomatic patients, three had benign lesions, four had low-grade dysplasia, one had moderate-grade dysplasia, and one patient had a high-grade dysplastic (CIS) lesion. None of the patients had invasive malignancy.

TABLE 1: Pathologic Distribution of Small Pancreatic Cystic Lesions

The gross features of the cysts in 38 patients, as described on pathology or endoscopic ultrasound, the gross features identified by readers on MDCT and MRI, agreement between reviewer 1 and reviewer 2, and agreement between MDCT and MRI are listed in Table 2. Good to perfect interreader agreement was established for MDCT (κ = 0.7–1) and MRI (κ = 0.72–1) for detecting cyst features. Similarly, good to perfect agreement (κ = 0.71–0.92) between MDCT and MRI for various cyst features confirmed on pathology or endoscopic ultrasound was noted (Figs. 1A, 1B, 1C and 2A, 2B). We discovered MRI to be more sensitive than MDCT for the assessment of cyst morphologic features such as septa, mural nodule, and especially ductal communication (sensitivity of MRI vs MDCT, 91%, 33.3%, 100%, respectively, vs 73.9%, 0%, 85.7%); however, these differences were not statistically significant (p = 0.25–1) (Table 3 and Figs. 3A, 3B and 4A, 4B). In nine patients with endoscopic ultrasound as the standard of reference, endoscopic ultrasound and MRI detected septa in six of nine cysts, and MDCT detected septa in five of nine cysts (Fig. 5A, 5B, 5C).

TABLE 2: Morphologic Features of 38 Small Pancreatic Cysts in 30 Patients on MDCT and MRI With MR Cholangiopancreatography (MRCP) with Pathologic Correlation

TABLE 3: Sensitivity of MDCT and MRI and of Dual-Phase and Single-Phase MDCT Protocols for the Detection of Various Morphologic Features of Small Pancreatic Cysts

There were false-positive detections of morphologic features on both MDCT and MRI. For septations, there were nine false-positives each with MDCT and MRI; for ductal communication, 11 and 15 false-positives, respectively, were detected with MDCT and MRI and one false-positive each with CT and MRI for mural nodule (Fig. 4A, 4B). The sensitivity of MDCT was higher when a dedicated dual-phase technique (85.7%, 0%, 87.5%, respectively) was used for scanning in comparison with a single-phase protocol (55.6%, 0%, 71.4%), although the difference was not statistically significant (p = 0.09–1) (Table 3).

For lesion characterization, the accuracy of both techniques and both readers was higher for classifying lesions into mucinous and nonmucinous types than for providing a specific diagnosis (Table 4). Considering the lesions with aggressive biology (moderate-grade dysplasia, n = 8; high-grade dysplasia [CIS], n = 4), the incidence of aggressiveness was 12 of 38 (31.6%). Both MDCT and MRCP showed a comparable accuracy for stratifying lesions into aggressive and nonaggressive groups. With ROC curve analysis for nonaggressive versus aggressive lesions, the Az for MDCT for reviewer 1 was 0.78 compared with 0.86 for MRCP (p = 0.34). Likewise, the Az for MDCT for reviewer 2 was 0.75 and that for MRCP was 0.78 (p = 0.72) (Fig. 6A, 6B). The positive predictive value (PPV) for aggressiveness and negative predictive value (NPV) for nonaggressiveness were better with MRCP than MDCT (Table 5).

TABLE 4: Accuracy of MDCT and MRI for Small Pancreatic Cyst Characterization

TABLE 5: Predictive Values of MDCT and MRI for Classification of Small Pancreatic Cysts Into Aggressiveness and Nonaggressiveness Categories

View larger version (197K)
Fig. 1A 77-year-old man with carcinoma in situ in intraductal papillary mucinous neoplasm. Axial thin-section contrast-enhanced MDCT image reveals lobulated cystic lesion in uncinate process with thick septations (arrows).

View larger version (158K)
Fig. 1B 77-year-old man with carcinoma in situ in intraductal papillary mucinous neoplasm. Contrast-enhanced fat-saturated T1-weighted MR image corresponding to A reveals similar features. Arrows point to thick septations.

View larger version (155K)
Fig. 1C 77-year-old man with carcinoma in situ in intraductal papillary mucinous neoplasm. Photograph of gross specimen from pathologic evaluation confirms presence of septations (arrow) and also revealed small mural nodule (not shown) that was not detected on CT or MRI.

Thick septations on imaging were associated with aggressive histology. Five lesions (one side-branch IPMN and four mixed IPMNs) had thick septations on MDCT and MRCP. Of these, four were aggressive and one was nonaggressive: Two mixed IPMNs revealed high-grade dysplasia (CIS), two mixed IPMNs revealed moderate-grade dysplasia, and the only side-branch IPMN revealed low-grade dysplasia (Fig. 1A, 1B, 1C and Table 6). Except the low-grade side-branch IPMN, none of the other nonaggressive lesions had thick septations; in fact, 15 of 26 nonaggressive lesions had thin septations. None of the 12 unilocular lesions on imaging were high grade or invasive, but four IPMNs (two mixed IPMNs and two side-branch IPMNs) exhibited low- or moderate-grade dysplastic changes.

TABLE 6: Relationship Between Morphologic Imaging Features and Histologic Aggressiveness

Three mural nodules detected on pathology in three cysts (mixed IPMN, pseudocyst, retention cyst) were not identified on MDCT. MRCP identified the nodule in a pseudocyst, which was characterized as debris at pathology (Fig. 7A, 7B). Although MDCT identified a mural nodule in one cyst (IPMN), that finding was confirmed to be thick mucous on pathology (Table 6).

MPD dilatation of > 8 mm was observed in three patients with mixed IPMN, and all three lesions were aggressive—i.e., two lesions showed high-grade dysplasia and one, moderate-grade dysplasia. MPD dilatation detected in a patient with a pseudocyst and in another patient with a mucinous cystic neoplasm was due to mass effect (Table 6).

View larger version (169K)
Fig. 2A 78-year-old woman with benign side-branch intraductal papillary mucinous neoplasm. Axial contrast-enhanced MDCT image (A) and axial T2-weighted MR image (B) reveal cystic lesion in neck of pancreas communicating with main pancreatic duct through narrow channel (arrow), which was also confirmed on pathology. Communication is more confidently visualized on MRI than CT.

View larger version (126K)
Fig. 2B 78-year-old woman with benign side-branch intraductal papillary mucinous neoplasm. Axial contrast-enhanced MDCT image (A) and axial T2-weighted MR image (B) reveal cystic lesion in neck of pancreas communicating with main pancreatic duct through narrow channel (arrow), which was also confirmed on pathology. Communication is more confidently visualized on MRI than CT.

Considering the subcategories of cysts by size (1–10, 11–20, and 21–30 mm), benign and various grades of dysplasia were found in all the subgroups, but high-grade dysplastic changes were not noted in lesions smaller than 10 mm. No identifiable morphologic difference was noted on imaging in lesions reported to harbor low- or moderate-grade dysplastic changes on pathology.

Previous sectionNext section

Increasingly frequent imaging detection of small pancreatic cysts and the dilemma of their management are well recognized [13, 28, 29]. Characterization of small pancreatic cysts is an important clinical objective for selecting the appropriate management. Observation can be considered for small pancreatic lesions that are incidentally detected in middle-age or elderly individuals if there are no suspicious morphologic features. Surgery is often recommended for symptomatic lesions, cysts enlarging during observation, cysts having complex morphology, or cysts detected in younger patients (< 50 years old) [16, 30]. MDCT and MRI with MRCP are the two commonly used noninvasive techniques for imaging the pancreas. Although MRI is a well-recognized tool for the detection and characterization of cystic pancreatic lesions, MDCT, with its esthetically pleasing multiplanar image displays, has compared well with MRI in several studies [2, 11, 13, 31].

Because of the inherent benefits of high spatial and temporal resolution, superior image display, and routine availability, MDCT is now being more prevalently used for the evaluation of pancreatic lesions including cysts, resulting in a large accumulating radiologic experience. Prior studies have compared the ability of CT and MRI in evaluating pancreatic cystic lesions irrespective of size [11, 12, 32]. To our knowledge, no studies have compared the accuracy of MDCT and MRI for characterizing small pancreatic cysts despite the fact that these lesions are now being more commonly encountered and are more likely to be considered as candidates for watchful waiting. This study offers an opportunity to address and understand several aspects of the imaging analysis of small pancreatic cysts with the two frequently used techniques which, in turn, can affect patient management.

View larger version (195K)
Fig. 3A 84-year-old man with side-branch intraductal papillary mucinous neoplasm. Axial contrast-enhanced MDCT image reveals cystic lesion in uncinate process, but intralesional septa are not distinctly visualized.

View larger version (128K)
Fig. 3B 84-year-old man with side-branch intraductal papillary mucinous neoplasm. Coronal single-shot fast spin-echo thick-slab MR cholangiopancreatography image reveals intralesional septa (large arrow) and communication with main pancreatic duct (small arrow). Pathology confirmed presence of septa and moderate-grade dysplastic changes.

View larger version (193K)
Fig. 4A 76-year-old man with benign side-branch intraductal papillary mucinous neoplasm. Coronal reformatted contrast-enhanced MDCT image reveals cystic lesion in body of pancreas with no obvious communication with main pancreatic duct.

View larger version (146K)
Fig. 4B 76-year-old man with benign side-branch intraductal papillary mucinous neoplasm. Coronal single-shot fast spin-echo thick-slab MR cholangiopancreatography image reveals very narrow communication channel (arrow) between lesion and main pancreatic duct; intralesional septa are also noted. Neither septa nor communication was reported on pathologic evaluation.

View larger version (173K)
Fig. 5A 75-year-old woman with mucinous cystic neoplasm in uncinate process. Axial thin-section contrast-enhanced MDCT image reveals cystic lesion within uncinate process with thin septa (arrow).

View larger version (127K)
Fig. 5B 75-year-old woman with mucinous cystic neoplasm in uncinate process. Axial fat-saturated T2-weighted MR image reveals intralesional septa (arrow) as seen on MDCT; additional septa are seen on right side of lesion (arrowhead).

View larger version (167K)
Fig. 5C 75-year-old woman with mucinous cystic neoplasm in uncinate process. Endoscopic ultrasound image confirms presence of septations (arrow and arrowhead).

Agreement between MDCT and MRI for detecting various pathologically confirmed morphologic features of small cysts was good to excellent. Compared with CT, MRI indeed allowed more reliable detection of cyst morphology, such as thin septa and mural nodules, and MRCP had an added advantage in detecting ductal communication. However, the differences in their performances were not statistically significant. Taking a similar approach in evaluating renal cystic lesions, Israel et al. [33] reported slight benefits with MRI over CT in detecting the internal features of renal cysts. In our study, despite a higher sensitivity of MRI than CT for detecting individual morphologic features, the overall accuracy of the two techniques for classifying lesions into histologic types (mucinous vs nonmucinous and specific diagnosis) was equivalent. Visser et al. [12] made similar observations in a subcohort of 12 patients with pancreatic cysts of varying sizes; they found that MRI did not offer any significant advantage over CT in characterizing cystic pancreatic masses. A study undertaken almost two decades ago showed similar benefits with the use of MRI for evaluating cyst features (except calcification) but found no substantial advantage in the diagnostic performance of MRI over CT [32]. Imaging technology has, however, advanced dramatically over the past decade and overcome several of the initial limitations. In addition, our understanding of these lesions also has improved over time. Indeed, in a recent study, investigators reported the performance of CT and MRI for evaluating IPMNs to be comparable [11].

View larger version (12K)
Fig. 6A Receiver operating characteristic curves for MDCT versus MRI categorization of small pancreatic cysts (n = 38) into nonaggressive and aggressive categories by reviewers. Gray curve represents data for MRI, and black curve represents that for MDCT. Dashed line is reference line. Reviewer 1.

View larger version (12K)
Fig. 6B Receiver operating characteristic curves for MDCT versus MRI categorization of small pancreatic cysts (n = 38) into nonaggressive and aggressive categories by reviewers. Gray curve represents data for MRI, and black curve represents that for MDCT. Dashed line is reference line. Reviewer 2.

In our study, there were false-positive diagnoses for septations and ductal communication on both MDCT and MRI. We postulate that these false-positives are due to the limitations of pathology. Thin septa can easily rupture during handling of a specimen or when a cyst is cut open for specimen processing. A high concordance established between MDCT and MRI findings and endoscopic ultrasound findings for cyst morphologic features supports our hypothesis. Similarly, cyst communication may not be identified if the channel is very small or is blocked by mucous or secretions. Furthermore, the pathologist's primary concerns more often are to provide a diagnosis and to determine the presence or absence of dysplasia and malignancy rather than to elucidate each morphologic feature. Because of the discussed limitations, the specificity of imaging for evaluating morphologic features could not be calculated.

View larger version (184K)
Fig. 7A 61-year-old man with infected pseudocyst. Axial thin-section contrast-enhanced MDCT image reveals smooth-walled cystic lesion in body of pancreas with unremarkable features and was called nonneoplastic cyst.

View larger version (135K)
Fig. 7B 61-year-old man with infected pseudocyst. Axial T2-weighted MR image reveals at least two definite mural nodules (arrows), resulting in diagnosis of mucinous lesion. Pathology revealed pseudocyst with infected debris adherent to wall of cyst.

In our study, the accuracy of both techniques in categorizing lesions as mucinous or nonmucinous was better than that in predicting a specific diagnosis for small lesions. A similar observation was made by Visser et al. [12] for cystic lesions regardless of their size and imaging technique. This result is expected and is due to morphologic overlap and variability in imaging appearances of various pathologic subtypes. Premalignant changes are more common in mucinous lesions [34, 35], and we found a similar trend in the small lesions as well. Hence, reliable distinction of mucinous from nonmucinous cysts is more relevant for patient management than prediction of specific histology.

The prevalence of high-grade dysplastic changes (4/38, 10.5%) and invasive carcinoma (0/38, 0%) was found to be relatively low in the current study. However, with the inclusion of moderate-grade dysplastic lesions (8/38) in the aggressive category, the prevalence of aggressiveness would be 31.6% (12/38). The rationale behind inclusion of moderate-grade dysplastic lesions in the aggressive category is that although the evolution and rate of progression into a more aggressive biology are not yet settled, these lesions are certainly premalignant [4, 3436]. On the other hand, mucinous lesions with low-grade dysplastic changes (adenomas) parallel other benign lesions in evolution [27]. Moreover, with CT and MRI being screening imaging tools, the detection of aggressiveness should tend toward higher sensitivity than specificity. Also, the relatively high incidence of aggressiveness in our study compared to previous reports [2, 4] could be attributed to selection bias because only patients who underwent MDCT and MRI and had confirmation of diagnosis on surgery or endoscopic ultrasound were included in this study. Confirmation of cyst morphology and suspicious features on two imaging tests could have led to an aggressive approach in these patients. Patients with pancreatic cysts who were followed up with only imaging possibly had less suspicious cyst morphology or unrelated clinical symptoms. Hence, this cohort is not representative for calculating the incidence of malignancy for small pancreatic cysts.

We found MDCT and MRI to be equally accurate in characterizing small cystic lesions into nonaggressive and aggressive categories with a reasonably high accuracy. A similar observation encompassing cysts unrestricted in size was made by Visser et al. [12]. Imaging features such as thick septations, significant MPD dilatation (> 8 mm), and mural nodules that have been noted in biologically aggressive lesions should result in a more aggressive management approach. Although the association of these variables with malignancy has already been shown in the literature [4, 7, 11, 37], our study reinforces this association and the applicability of the same criteria for small cystic lesions also. Although the presence of a solid component in a cyst is considered a strong indicator for malignancy [3, 7, 14], in our study only one lesion with a solid component (IPMN) revealed high-grade dysplasia (CIS), and this solid component was undetected on both MDCT and MRI. Proteinaceous material, thick secretions, or mucin found in benign lesions, both mucinous and nonmucinous lesions, can imitate a mural nodule on imaging. Because MRI is marginally better than MDCT in evaluating cyst morphology, cystic lesions with suspicious or dubious features on MDCT might benefit from evaluation with MRI. MRI may contribute not only by detecting or confirming ominous features such as a mural nodule or thick septa, but also by establishing the diagnosis of an IPMN by detecting ductal communication, which can be better visualized on MRCP especially on T2-weighted images [4].

There were several limitations in our study. Because of the retrospective nature of this study and very specific inclusion criteria, the possibility of an inadvertent bias cannot be ruled out. The relatively small sample size is another limitation of this study; therefore, we cannot use our data for assessing the incidence of malignancy in small pancreatic cysts. In addition, a dedicated thin-section pancreatic protocol CT examination with high-quality multiplanar display was not available in all patients; this conceivably could have further augmented the sensitivity of CT for the detection of morphologic features. However, this inconsistency in our CT protocol provided us an opportunity to study the limitations of a single-phase CT protocol in evaluating small pancreatic cysts. Finally, surgical confirmation and detailed pathologic evaluation were not available in nine patients; in those patients, findings on endoscopic ultrasound and FNA provided our reference standard.

In conclusion, MDCT and MRI have comparable accuracy in characterizing the histologic aggressiveness of small pancreatic cysts (≤ 3 cm). MDCT, with its wide availability, fast scanning, and lack of need of special operator skills, can thus be advantageously used for the evaluation, observation, and management decision making of these lesions. A dedicated thin-section dual-phase technique improves the diagnostic performance of MDCT for assessing cyst morphology and should be the preferred approach for evaluating small cysts with CT, although MRI using MRCP can be resorted to in cases with doubtful or suspicious features or can be used alternatively instead of CT. The accuracy of both modalities is best when considered in evaluating morphologic features reflecting histologic aggressiveness, a factor that assists in guiding management.

Address correspondence to D. V. Sahani ().

We gratefully acknowledge Elkan Halpern for assisting in statistical analysis.

Previous sectionNext section
1. Fernández-del Castillo C, Targarona J, Thayer SP, Rattner DW, Brugge WR, Warshaw AL. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003; 138:427 –433; discussion 433–434 [Google Scholar]
2. Sahani DV, Saokar A, Hahn PF, Brugge WR, Fernandez-Del Castillo C. Pancreatic cysts 3 cm or smaller: how aggressive should treatment be? Radiology 2006; 238:912–919 [Google Scholar]
3. Megibow AJ, Lombardo FP, Guarise A, et al. Cystic pancreatic masses: cross-sectional imaging observations and serial follow-up. Abdom Imaging 2001; 26:640–647 [Google Scholar]
4. Kirkpatrick ID, Desser TS, Nino-Murcia M, Jeffrey RB. Small cystic lesions of the pancreas: clinical significance and findings at follow-up. Abdom Imaging 2007; 32:119–125 [Google Scholar]
5. Sugiyama M, Izumisato Y, Abe N, Masaki T, Mori T, Atomi Y. Predictive factors for malignancy in intraductal papillary-mucinous tumours of the pancreas. Br J Surg 2003; 90:1244 –1249 [Google Scholar]
6. Wiesenauer CA, Schmidt CM, Cummings OW, et al. Preoperative predictors of malignancy in pancreatic intraductal papillary mucinous neoplasms. Arch Surg 2003; 138:610–617; discussion 617–618 [Google Scholar]
7. Allen PJ, Jaques DP, D'Angelica M, Bowne WB, Conlon KC, Brennan MF. Cystic lesions of the pancreas: selection criteria for operative and nonoperative management in 209 patients. J Gastrointest Surg 2003; 7:970 –977 [Google Scholar]
8. Procacci C, Biasiutti C, Carbognin G, et al. Characterization of cystic tumors of the pancreas: CT accuracy. J Comput Assist Tomogr 1999; 23:906 –912 [Google Scholar]
9. Balci NC, Semelka RC. Radiologic features of cystic, endocrine and other pancreatic neoplasms. Eur J Radiol 2001; 38:113 –119 [Google Scholar]
10. Irie H, Honda H, Aibe H, et al. MR cholangiopancreatographic differentiation of benign and malignant intraductal mucin-producing tumors of the pancreas. AJR 2000; 174:1403 –1408 [Abstract] [Google Scholar]
11. Taouli B, Vilgrain V, O'Toole D, Vullierme MP, Terris B, Menu Y. Intraductal papillary mucinous tumors of the pancreas: features with multimodality imaging. J Comput Assist Tomogr 2002; 26:223 –231 [Google Scholar]
12. Visser BC, Yeh BM, Qayyum A, Way LW, McCulloch CE, Coakley FV. Characterization of cystic pancreatic masses: relative accuracy of CT and MRI. AJR 2007; 189:648 –656 [Abstract] [Google Scholar]
13. Song SJ, Lee JM, Kim YJ, et al. Differentiation of intraductal papillary mucinous neoplasms from other pancreatic cystic masses: comparison of multirow-detector CT and MR imaging using ROC analysis. J Magn Reson Imaging 2007; 26:86 –93 [Google Scholar]
14. Sahani DV, Kadavigere R, Blake M, Fernandez-Del Castillo C, Lauwers GY, Hahn PF. Intraductal papillary mucinous neoplasm of pancreas: multidetector row CT with 2D curved reformations—correlation with MRCP. Radiology 2006; 238:560–569 [Google Scholar]
15. Fukukura Y, Fujiyoshi F, Hamada H, et al. Intraductal papillary mucinous tumors of the pancreas: comparison of helical CT and MR imaging. Acta Radiol 2003; 44:464–471 [Google Scholar]
16. Handrich SJ, Hough DM, Fletcher JG, Sarr MG. The natural history of the incidentally discovered small simple pancreatic cyst: long-term follow-up and clinical implications. AJR 2005; 184:20–23 [Abstract] [Google Scholar]
17. Lahav M, Maor Y, Avidan B, Novis B, Bar-Meir S. Nonsurgical management of asymptomatic incidental pancreatic cysts. Clin Gastroenterol Hepatol 2007; 5:813–817 [Google Scholar]
18. Kim YH, Saini S, Sahani D, Hahn PF, Mueller PR, Auh YH. Imaging diagnosis of cystic pancreatic lesions: pseudocyst versus nonpseudocyst. RadioGraphics 2005; 25:671–685 [Google Scholar]
19. Sahani DV, Kadavigere R, Saokar A, Fernandez-del Castillo C, Brugge WR, Hahn PF. Cystic pancreatic lesions: a simple imaging-based classification system for guiding management. RadioGraphics 2005; 25:1471 –1484 [Google Scholar]
20. Morana G, Guarise A. Cystic tumors of the pancreas. Cancer Imaging 2006; 6:60–71 [Google Scholar]
21. Curry CA, Eng J, Horton KM, et al. CT of primary cystic pancreatic neoplasms: can CT be used for patient triage and treatment? AJR 2000; 175:99 –103 [Abstract] [Google Scholar]
22. Planner AC, Anderson EM, Slater A, Phillips-Hughes J, Bungay HK, Betts M. An evidence-based review for the management of cystic pancreatic lesions. Clin Radiol 2007; 62:930–937 [Google Scholar]
23. Choi BS, Kim TK, Kim AY, et al. Differential diagnosis of benign and malignant intraductal papillary mucinous tumors of the pancreas: MR cholangiopancreatography and MR angiography. Korean J Radiol 2003; 4:157 –162 [Google Scholar]
24. Kawamoto S, Lawler LP, Horton KM, Eng J, Hruban RH, Fishman EK. MDCT of intraductal papillary mucinous neoplasm of the pancreas: evaluation of features predictive of invasive carcinoma. AJR 2006; 186:687 –695 [Abstract] [Google Scholar]
25. Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology 2004; 126:1330 –1336 [Google Scholar]
26. Moparty B, Logrono R, Nealon WH, et al. The role of endoscopic ultrasound and endoscopic ultrasound–guided fine-needle aspiration in distinguishing pancreatic cystic lesions. Diagn Cytopathol 2007; 35:18 –25 [Google Scholar]
27. Hruben RH, Pitman MB, Klimstra DS. Intraductal neoplasms. In: Hruban RH, Pitman MB, Klimstra DS, eds. AFIP atlas of tumor pathology. 2007; Series 4:71 –110 [Google Scholar]
28. Napoleon B, Lefort C. Incidental discovery of cystic lesions of the pancreas. Gastroenterol Clin Biol 2007; 31:210–215 [Google Scholar]
29. Spinelli KS, Fromwiller TE, Daniel RA, et al. Cystic pancreatic neoplasms: observe or operate. Ann Surg 2004; 239:651 –657; discussion 657–659 [Google Scholar]
30. Tanaka M, Chari S, Adsay V, et al. International consensus guidelines for management of intraductal papillary mucinous neoplasms and mucinous cystic neoplasms of the pancreas. Pancreatology 2006; 6:17–32 [Google Scholar]
31. McNulty NJ, Francis IR, Platt JF, Cohan RH, Korobkin M, Gebremariam A. Multi-detector row helical CT of the pancreas: effect of contrast-enhanced multiphasic imaging on enhancement of the pancreas, peripancreatic vasculature, and pancreatic adenocarcinoma. Radiology 2001; 220:97 –102 [Google Scholar]
32. Minami M, Itai Y, Ohtomo K, Yoshida H, Yoshikawa K, Iio M. Cystic neoplasms of the pancreas: comparison of MR imaging with CT. Radiology 1989; 171:53–56 [Google Scholar]
33. Israel GM, Hindman N, Bosniak MA. Evaluation of cystic renal masses: comparison of CT and MR imaging by using the Bosniak classification system. Radiology 2004; 231:365–371 [Google Scholar]
34. Kawamoto S, Horton KM, Lawler LP, Hruban RH, Fishman EK. Intraductal papillary mucinous neoplasm of the pancreas: can benign lesions be differentiated from malignant lesions with multidetector CT? RadioGraphics 2005; 25:1451–1468; discussion 1468–1470 [Google Scholar]
35. Procacci C, Carbognin G, Biasiutti C, Guarise A, Ghirardi C, Schenal G. Intraductal papillary mucinous tumors of the pancreas: spectrum of CT and MR findings with pathologic correlation. Eur Radiol 2001; 11:1939 –1951 [Google Scholar]
36. Salvia R, Fernandez-del Castillo C, Bassi C, et al. Main-duct intraductal papillary mucinous neoplasms of the pancreas: clinical predictors of malignancy and long-term survival following resection. Ann Surg 2004; 239:678 –685; discussion 685–687 [Google Scholar]
37. Sakorafas GH, Sarr MG, van de Velde CJ, Peros G. Intraductal papillary mucinous neoplasms of the pancreas: a surgical perspective. Surg Oncol 2005; 14:155–178 [Google Scholar]

Recommended Articles

Comparative Performance of MDCT and MRI With MR Cholangiopancreatography in Characterizing Small Pancreatic Cysts

Full Access, , , , ,
American Journal of Roentgenology. 2013;200:343-354. 10.2214/AJR.12.8862
Abstract | Full Text | PDF (877 KB) | PDF Plus (1009 KB) 
Full Access, , , , ,
American Journal of Roentgenology. 2007;189:648-656. 10.2214/AJR.07.2365
Abstract | Full Text | PDF (888 KB) | PDF Plus (1023 KB) 
Full Access, ,
American Journal of Roentgenology. 2011;196:W668-W677. 10.2214/AJR.10.4378
Citation | Full Text | PDF (979 KB) | PDF Plus (986 KB) 
Full Access, , , , , , , ,
American Journal of Roentgenology. 2008;191:802-807. 10.2214/AJR.07.3340
Abstract | Full Text | PDF (656 KB) | PDF Plus (959 KB) 
Full Access, , , ,
American Journal of Roentgenology. 2007;189:657-661. 10.2214/AJR.07.2772
Citation | Full Text | PDF (83 KB) | PDF Plus (178 KB) 
Full Access, , , , , ,
American Journal of Roentgenology. 2010;195:947-952. 10.2214/AJR.09.3985
Abstract | Full Text | PDF (724 KB) | PDF Plus (782 KB)