Purpose. Improving the efficiency of CT in the differential diagnosis of mass-forming pancreatitis (MFP) and pancreatic ductal adenocarcinoma (PDAC) making a diagnostic model based using a combination of texture features and contrast enhancement features.Methods and materials. 45 patients with histologically confirmed non-metastatic locally advanced PDAC and 13 patients with MFP where underwent CT examination with contrast enhancement. For each group, the ratio of the densities of intact pancreatic tissue and tumors, the relative tumor enhancement ratio (RTE) in all enhanced phases of CT, 94 texture features for each phase of the study were calculated and compared. The selection of predictors in the logistic model was carried out in 2 stages: 1) selection of predictors based on one-factor logistic models, the selection criterion was padj <0.2; 2) selection of predictors using LASSO-regression after standardization of variables. The selected predictors were included in a logistic regression model without interactions.Results. There were statistically significant differences in 14, 17, 4 out of 94 for the unenhanced, arterial, and venous phases of the study, respectively (p < 0.05). After selection, the final diagnostic model included the texture features CONVENTIONAL HUQ2 and DISCRETIZED HUQ1 for the unenhanced phase, DISCRETIZED HUQ1 and GLRLM RLNU for the arterial phase, DISCRETIZED Skewness for the venous phase, RTE for the delayed CT phase. The diagnostic model was built showed an accuracy of 81% in the diagnosis of MFP.Conclusion. We have developed a diagnostic model, including textural parameters and contrast enhancement features, which allows preoperatively distinguish MFP and PDAC, the developed model will increase the accuracy of preoperative diagnosis.
Read full abstract