Non-gaussian models of 3-Tesla diffusion-weighted MRI for the differentiation of pancreatic ductal adenocarcinomas from neuroendocrine tumors and solid pseudopapillary neoplasms

Purpose: To assess the MRI performance in differentiating pancreatic ductal adenocarcinomas (PDACs), from solid pseudopapillary neoplasms (SPNs) and pancreatic neuroendocrine tumors (PNETs) using non-gaussian diffusion-weighted imaging models. Methods: This was a retrospective study of patients diagnosed with PDACs (01/2015-06/2019) or with PNETs or SPNs diagnosed (01/2011-12/2019) at our hospital. The lesions were randomized 1:1 to the primary and validation cohorts. The regions of interest (ROIs) were manually drawn on each slice at DWI (b = 1500 s/mm(2)) from 3 T MRI. D (diffusion coefficient), D* (pseudodiffusion coefficient), f (perfusion fraction), distributed diffusion coefficient (DDC), a (diffusion heterogeneity index), mean diffusivity (MD) and mean kurtosis (MK) were obtained. The parameters with largest performance for differentiation were used to establish a diagnostic model. Results: There were 148, 56, and 60 patients with PDAC, PNET, and SPN, respectively. For differentiating PDACs from SPNs, f and MK values were used to establish a diagnostic model with areas under the receiver operating characteristic curves (AUCs) of 0.92 and 0.89 in the primary and validation groups, respectively. For distinguishing PDACs from PNETs, a and MK values were used to establish a diagnostic model with AUCs of 0.87 and 0.86 in the primary and validation groups, respectively. The accuracy rate of the subjective evaluation with the assistance of non-gaussian DWI models for differentiating PDAC from SPNs and PNETs were higher than that of subjective diagnosis alone (P < 0.05). Conclusions: The non-gaussian DWI models could assist radiologists in accurately differentiating PDACs from PNETs and SPNs.