Characterization of cystic pancreatic masses: Relative accuracy of CT and MRI

OBJECTIVE. The objective of our study was to determine the role and relative accuracy of CT and MRI in the characterization of cystic pancreatic masses. MATERIALS AND METHODS. We retrospectively identified 58 patients with histopathologically proven cystic pancreatic masses at our institution who underwent preoperative CT (n = 40), MRI (n = 6), or both (n = 12). Two radiologists independently recorded their leading diagnoses with levels of diagnostic certainty (0-100%), their estimates of overall likelihood of malignancy (0-100%), and the morphologic characteristics of the tumors. Data were analyzed to determine relative accuracy in the diagnosis of malignancy, relationship between diagnostic certainty and accuracy, and frequency of malignancy in unilocular thin-walled cysts smaller than 4 cm. RESULTS. Twenty-one (36%) of 58 masses were malignant. CT and MRI were equally accurate in establishing the diagnosis of malignancy (area under the receiver operating characteristic curve [A_z] = 0.91 and 0.85 for reviewers 1 and 2 at MRI vs 0.82 and 0.76 at CT, respectively; p > 0.05). The leading diagnosis given by reviewers 1 and 2 was correct in 46% (32/70) and 43% (30/70) of the studies, respectively. When reviewer diagnostic certainty was 90% or more, the corresponding values were not significantly (p > 0.05) improved at 55% (12/22) and 48% (10/21), respectively. Two (15%) of 13 unilocular thin-walled cysts smaller than 4 cm were frankly malignant. CONCLUSION. CT and MRI are reasonably and similarly accurate in the characterization of cystic pancreatic masses as benign or malignant; limitations include a substantial rate of misdiagnosis even when reviewer certainty is high and a moderate frequency of malignancy in small morphologically benign-appearing cysts.