Objective: To retrospectively elucidate the findings useful in determining the tumor grade of pancreatic neuroendocrine tumors (PNETs) by combined assessment of magnetic resonance (MR) and dynamic computed tomography (CT) images.

Methods: Eighty-nine patients with PNETs (96 lesions) were included, and classified as G1, 59; G2, 29; and G3, 8 lesions. Image analysis included lesion diameter, shape, enhancement pattern on arterial phase (AP) and delayed phase CT images, calcification, cystic portion, main pancreatic duct dilatation, signal-intensity on T1-, T2-weighted MR images, and appearance of apparent diffusion coefficient (ADC).

Results: Significant differences among G1, G2, and G3 groups were noted in tumor maximal diameter (p < 0.0001), shape (p < 0.0001), enhancement pattern on AP image (p < 0.0001), cystic portion (p = 0.012), and ADC finding. In multivariate analysis, ADC finding was the independent factor (p = 0.002). The combination findings of low ADC ratio (ADC value of the lesion/ADC value of the parenchyma <0.94), not homogeneous hyper-attenuation, lobulated shape, and hyper-intensity on T2-weighted image were suggestive of G2 or G3 with a probability of 100%. Conversely, all lesions with high ADC ratio and small size (≤25 mm) belonged to the G1 group.

Conclusions: Combined assessment of MR and CT findings could improve the prediction of tumor grading in PNETs.