Added value of multiplanar reformations to axial multi-detector row computed tomographic images for the differentiation of macrocystic pancreas neoplasms: receiver operating characteristic analysis

Abstract

Purpose: To retrospectively assess the added value of multiplanar reformations (MPR) to axial multi-detector row computed tomographic (MDCT) images in differentiating macrocystic pancreas neoplasms.

Materials and methods: Approval from the institutional review board was obtained. Two radiologists retrospectively reviewed axial CT images of 48 pathologically proven pancreas macrocystic neoplasms with and without MPRs. They were asked to determine the presence of pancreatic duct (PD) communication with the lesions and whether the lesion is an intraductal papillary mucinous neoplasm (IPMN) or not on a 5-point confidence scale and to record the specific diagnoses and their confidence. Radiologists' performances for determining PD communication and lesion differentiation using axial CT with and without MPRs were evaluated using receiver operating characteristic analysis. To determine the accuracy of the specific diagnoses, Fisher exact and Mann-Whitney U tests were used. Interobserver agreement was also analyzed.

Results: With the addition of MPRs, receiver operating characteristic analysis revealed a tendency toward improved determination of PD communication and better differentiation between IPMN and non-IPMN. However, a significant difference was found only in reviewer 2 for the determination of PD communication (P = 0.009). Diagnostic accuracy of specific diagnoses was also improved; however, the differences were not significant. Specific diagnoses were more confidently made with the addition of MPRs than with axial images alone, and a significant difference was seen for reviewer 2 (P < 0.001). Furthermore, substantial interobserver agreement was achieved with the addition of MPRs, whereas fair or substantial agreement was noted with axial images alone.

Conclusions: The addition of MPRs to axial CT images may improve diagnostic performance and decrease interobserver variability of MDCT for the determination of PD communication with macrocystic pancreatic neoplasms and differentiation between IPMN and non-IPMN.