Objective: The magnetic resonance imaging (MRI) features of intestinal-type periampullary carcinoma (IPAC) and pancreatobiliary-type periampullary carcinoma (PPAC) were compared and analyzed to discuss the optimal diagnosis scheme. Method: Preoperative MRI images of 59 patients (32 males, 27 females, aged 37-80 years) diagnosed with periampullary carcinoma (PAC) confirmed by surgery and pathology in Nanjing Drum Tower Hospital from January 2017 to July 2020 were retrospectively analyzed. The patients were divided into 21 cases in the IPAC group (11 males, 10 females) and 38 cases in the PPAC group (21 males, 17 females) according to histopathological results. The conventional MRI plain scan signs included in the analysis include lesion morphology, the largest diameter of the lesion, lesion location, duodenal papilla morphology, plain scan lesion signal (with the normal pancreatic signal as reference), diffusion weighted imaging (DWI) signal. Magnetic resonance cholangiopancreatography (MRCP) image signs include the dilatation of common bile duct and main pancreatic duct and quantitative analysis of their diameter, the presence of a round filling defect in the distal end of the common bile duct, the morphology of common bile duct stenosis, the dilatation of lateral branches around the obstructed pancreatic duct, the ductal sign, the distance from the end of the obstructed common bile duct to the duodenal papilla, the distance from the end of the obstructed pancreatic duct to the duodenal papilla, and the angle of the pancreaticobiliary duct. The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic efficacy of single meaningful factors. The MRI features of PAC were summarized according to the significant single factor indicators and were classified into 5 image types. And the diagnostic efficacy of the classification criteria for pathological subtypes of PAC was evaluated by the ROC curve. The DeLong test was used to compare the area under the ROC curve (AUC) of multiple diagnostic methods. Results: In univariate analysis, there were statistically significant differences between IPAC and PPAC in lesion location, duodenal papilla morphology, the circular filling defect in the distal end of the common bile duct, the distance from the obstructed pancreatic duct to the duodenal papilla, the angle of the pancreaticobiliary duct, and lesion signal characteristics on plain T2WI fat suppressant images (all P<0.05). Among the 5 types of MRI images, IPAC is mostly manifested as duodenal papillary nodules(15/21,71.4%), while PPAC is more manifested as pancreatic mass type(18/38,47.4%), thickened common bile duct wall type(9/38,23.7%) or ampullary mass type(9/38,23.7%). Both IPAC(2/21,9.5%) and PPAC(0,0) rarely showed the nodular type of common bile duct lumen. In the DeLong test of the significant univariate index(lesion location, duodenal papilla morphology, the circular filling defect in the distal end of the common bile duct, the distance from obstructed pancreatic duct to duodenal papilla, the angle of the pancreaticobiliary duct, and lesion signal characteristics on plain T2WI fat suppressant images) and the 5 classification of MRI images, the AUC of the 5 classifications of MRI images was 0.932(95%CI:0.867-0.997), which was higher than that of any of the significant univariate indexes (all P<0.05). In addition, the 5 classifications of MRI images have the same high diagnostic power as the logistic regression analysis model(P>0.05). Conclusions: The 5 classification of MRI images can improve the accuracy of differential diagnosis of IPAC and PPAC before surgery, and the diagnostic efficiency is better than any single factor meaningful index and comparable to that of the logistic regression analysis model.