ABSTRACT
Objective To investigate the feasibility of the CT texture analysis (CTTA)in differentiating autoimmune pancreatitis (AIP)from pancreatic ductal adenocarcinoma (PDAC).Methods 25 patients with AIP and 31 patients with PDAC who confirmed by pathological or clinical underwent pretreatment three-phase contrast-enhanced CT were enrolled.Histogram parameters (mean CT values,median CT values,25 th,75 th percentile CT values,skewness,kurtosis,entropy and uniformity)were derived from CT images through texture analysis.The differences of histogram parameters between AIP and PDAC groups were compared.ROC and AUC were used to evaluate the diagnostic efficacy of histogram parameters in differentiating AIP from PDAC.Results The values for mean CT values,median CT values,25 th,75 th percentile CT values and uniformity of AIP were significantly higher than those of PDAC group,while the values for entropy of AIP were significantly lower than those of PDAC group in arterial phase,portal phase,and delay phases (all P＜0.05). There were no significant differences in kurtosis and skewness between AIP and PDAC groups (all P＞0.05).The uniformity in portal phase achieved the optimal diagnostic accuracy in differentiating AIP from PDAC (AUC＝0.973 ),the cutoff value was 0.797,the corresponding sensitivity and specificity were 92% and 9 6.8%,respectively.Conclusion CTTA can be used as a quantitative analysis method for differential diagnosis between AIP and PDAC,providing a reference for clinicians to select therapeutic schedules.