Whole-tumor histogram analysis of non-Gaussian distribution DWI parameters to differentiation of pancreatic neuroendocrine tumors from pancreatic ductal adenocarcinomas

Purpose: To evaluate the utility of volumetric histogram analysis of monoexponential and non-Gaussian distribution DWI models for discriminating pancreatic ductal adenocarcinoma (PDAC) and neuroendocrine tumor (pNET). Materials and methods: A total of 340 patients were retrospectively reviewed. Finally, 62 patients with histopathological confirmed PDAC (n = 42) and pNET (n = 20) were enrolled in the study. All the patients accepted magnetic resonance imaging (MRI) at 3 T (including multi-b value DWI, 0-1000 s/mm(2)). Isotropic apparent diffusion coefficient (ADC), true molecular diffusion (Dt), perfusion-related diffusion (Dp), perfusion fraction (1), distributed diffusion coefficient (DDC) and alpha (a) were obtained from different DWI models. Then, mean value, median value, 10th and 90th percentiles were obtained from histogram analysis of each DWI parameter. Results: Histogram metrics derived from ADC, Dp, f and DDC were significantly lower in PDAC than pNET group (P < 0.05). In contrast, histogram metrics derived from a were observed significantly higher in the PDAC than pNET group (P < 0.05). No significant difference was found in Dt (P >= 0.05) between PDAC and pNET patients. Among all parameters, f-median had the highest diagnostic performance (AUC 0.91, cutoff value 0.188, sensitivity 97.62%, specificity 80%). Conclusions: f-Median derived from IVIM DWI model may be potentially more valuable parameter than ADC, Dp, DDC and a for discriminating PDAC and pNET. Histogram analysis based on the entire tumor was an emerging and valuable tool.