MR Quantitative 3D Shape Analysis Helps to Distinguish Mucinous Cystic Neoplasm from Serous Oligocystic Adenoma

PURPOSE We aimed to assess the performance of quantitative 3D shape analysis in the differential diagno- sis of pancreatic serous oligocystic adenoma (SOA) and mucinous cystic neoplasm (MCN). METHODS Four hundred thirty-two patients diagnosed with serous cystic neoplasms (SCNs) or MCNs were retrospectively reviewed from August 2014 to July 2019 and finally 87 patients with MCNs (n = 45) and SOAs (n = 42) were included. Clinical data and magnetic resonance morphologic fea- tures with 3D shape analysis of lesions (shape sphericity, compacity, and volume) were recorded and compared between MCNs and SOAs according to the pathology. Univariable and multivari- able regression analyses were used to identify independent impact factors for differentiating MCN from SOA. RESULTS The age of MCN patients was younger than SOAs (43.02 ± 10.83 years vs. 52.78 ± 12.31 years; OR = 0.275; 95% CI: 0.098-0.768; P = .014). MCN has a higher female/male ratio than SOA (43/2 vs. 27/15; OR = 40.418; 95% CI: 2.704-604.171; P = .007) and was more often located in the distal of pancreas (OR = 31.403; 95% CI: 2.985-330.342; P = .004). Shape_Sphericity derived from 3D shape analysis was a significant independent factor in the multivariable analysis and the value of MCN was closer to 1 than SOA (OR = 35.153; 95% CI: 5.301-237.585; P < .001). Area under the receiver operating characteristic curve (AUC) of Shape_Sphericity was 0.923 (optimal cutoff value was 0.964876). CONCLUSION Shape_Sphericity in combination with age, sex, and location could help to distinguish MCN from SOA.