Relative accuracy of CT and MRI in the differentiation of benign from malignant pancreatic cystic lesions.

Aims To assess the diagnostic accuracies of multidetector computed tomography (MDCT) and magnetic resonance imaging (MRI) for differentiating benign from malignant lesions and suggesting the specific diagnoses for pancreatic cystic lesions, and to assess whether review of both MDCT and MRI is beneficial. Materials and methods Patients with various neoplastic and non-neoplastic pancreatic cystic lesions that were identifiable by biopsy or surgery, who underwent both MRI and MDCT ( n  = 63), were retrospectively reviewed by three reviewers. The likelihood of malignancy was recorded on a five-point scale, and a specific diagnosis was given. ROC analysis was performed and the sensitivity, specificity for the characterization of malignancy, and the accuracy of specific diagnoses were calculated. Results MDCT and MRI yielded comparable results for the characterization of malignancy (Az: 0.639, 0.735, 0.806 for MDCT and 0.732, 0.753, 0.792 for MRI, for each reviewer). The accuracies of specific diagnosis based on MDCT or MRI were 61.9 versus 55.6% for reviewer 1; 76.2 versus 76.2% for reviewer 2; and 65.1 versus 61.9% for reviewer 3. There was a trend toward better prediction of malignancy (Az: 0.787, 0.745, 0.849 for each reviewer), and better accuracy in suggesting a specific diagnosis (77.8, 73, and 73% for each reviewer) for MDCT + MRI over MDCT or MRI alone, although it was statistically significant for one reviewer in the comparison of MDCT versus MDCT + MRI for the prediction of malignancy, and MRI versus MDCT for suggesting a specific diagnosis. Conclusions MDCT and MRI have equivalent accuracy for characterizing pancreatic cystic lesions as benign or malignant, and suggesting a specific diagnosis. Combined review of MDCT and MRI was not significantly better but may have the potential to improve diagnostic accuracy in equivocal cases.