ObjectivesTo investigate the diagnostic value of different diffusion-weighted MRI (DWI) models between two Gaussian DWI models including mono-exponential and bi-exponential, and the non-Gaussian kurtosis model in poorly differentiated pancreatic ductal adenocarcinoma.MethodsSubjects comprised 52 patients with poorly differentiated pancreatic ductal adenocarcinoma which had been confirmed by surgery. All patients underwent DWI (1.5T, multi-b values: 0, 50, 100, 150, 200, 500, 800, 1000, 1 500, 2 000s/mm2). Mean values of DWI-derived metrics ADCstandard, ADCslow, ADCfast, f, MD, MK and ADCstandard were calculated from regions of interest in all tumours and non-tumorous parenchyma and compared. ANOVA and Mann Whitney U test was used to compare the MRI paremeters. ROC was used to evaluate the diagnostic efficiency.ResultsMean ADCstandard, ADCfast, f and MK values showed significant differences between tumours and non-tumorous parenchyma (P<0.05). AUC for ADCstandard, MD, ADCfast and f were 0.705, 0.665, 0.648, 0.614, respectively. The ROC curve integrated with ADCstandard and MD had better diagnostic efficiency (AUC was about 0.754).ConclusionsADCstandard, ADCfast, f and MK values can differentiate tumours from non-tumorous parenchyma. The combination of Gaussion distribution model and non-Gaussion distribution model has the potential to increase the diagnostic accuracy of DWI in patients with pancreatic ductal adenocarcinoma.