Pancreatic cancer detection with a non-contrast MR protocol: is it reliable?

Abstract

Abstract Purpose The pancreatic cancer (PC) is the 4th leading cancer-related death, becoming the second one by 2030, with a 5 year survival rate of 8%. Considering its increased incidence in high-risk categories compared to the general population, we aimed to validate a non-contrast MR protocol, to detect PC in its earliest phase, which could be suitable as a screening tool in high-risk patients. Materials and methods In this retrospective study, we selected 200 patients (> 40 years) from our radiological database, which performed upper abdominal MRI between 2012 and 2017. 100 were negative for pancreatic lesions and 100 positive for pancreatic lesion (< 30 mm). The latter group included: 40 PDAC (pancreatic adenocarcinoma), 42 BD-IPMN (Branch Duct- Intraductal Papillary Mucinous Neoplasm), 10 PNET(pancreatic neuroendocrine tumor), 4 SCN(serous cystic neoplasm), 3 IPS(intrapancreatic spleen), 1 MCN(mucinous cystic neoplasm). Three readers (R1, R2 and R3) with a high, medium and low experience, respectively, analysed, first, the non-contrast MR sequences (single-shot T2w breath-hold, GE T1w FS, DWI and 2D/3D MRCP), and then the standard MR protocol, independently, randomly and anonymously. Readers identified or excluded the presence of pancreatic lesion, in both reading sessions. These results were compared with the histopathological diagnosis, and then divided into 3 different classes of lesions: all lesions, pancreatic adenocarcinoma and solid lesion. Mcnemar’s test was used to compare the results. The inter-observer agreement was determined according to the kappa statistic in both protocols, and then the inter-protocol agreement was calculated. Results The non-contrast MR protocol has reached statistical parameters values ranging between 83% in SE (sensitivity) by R3 and 99% in NPV (negative predictive value) by R1. The standard MR protocol has reported slight increasing statistical parameters compared to those of the proposed one. However, there are not significant statistical differences between the both protocols. The proposed non-contrast MR protocol has reported the highest NPVs in the PDAC group detection (R1: 99%, R2: 99%, R3: 98%). In all groups of lesions, the agreement between the two protocols was excellent for each Reader ranging from 96 to 98%. Conclusion The proposed non-contrast MR protocol showed high PC detection values and a time execution ≤ 20 min. Therefore, it can be proposed as a screening tool in high-risk patients.