Predictive value of multi-detector computed tomography for accurate diagnosis of serous cystadenoma: Radiologic-pathologic correlation

Abstract

To identify multi-detector computed tomography (MDCT) features most predictive of serous cystadenomas (SCAs), correlating with histopathology, and to study the impact of cyst size and MDCT technique on reader performance. The MDCT scans of 164 patients with surgically verified pancreatic cystic lesions were reviewed by two readers to study the predictive value of various morphological features for establishing a diagnosis of SCAs. Accuracy in lesion characterization and reader confidence were correlated with lesion size (<or= 3 cm or >or= 3 cm) and scanning protocols (dedicated vs routine). 28/164 cysts (mean size, 39 mm; range, 8-92 mm) were diagnosed as SCA on pathology. The MDCT features predictive of diagnosis of SCA were microcystic appearance (22/28, 78.6%), surface lobulations (25/28, 89.3%) and central scar (9/28, 32.4%). Stepwise logistic regression analysis showed that only microcystic appearance was significant for CT diagnosis of SCA (P = 0.0001). The sensitivity, specificity and PPV of central scar and of combined microcystic appearance and lobulations were 32.4%/100%/100% and 68%/100%/100%, respectively. The reader confidence was higher for lesions > 3 cm (P = 0.02) and for MDCT scans performed using thin collimation (1.25-2.5 mm) compared to routine 5 mm collimation exams (P > 0.05). Central scar on MDCT is diagnostic of SCA but is seen in only one third of SCAs. Microcystic morphology is the most significant CT feature in diagnosis of SCA. A combination of microcystic appearance and surface lobulations offers accuracy comparable to central scar with higher sensitivity.