Pancreatic ductal adenocarcinoma: CT texture analysis

Purpose: The aim of this study is to evaluate the presence of metastasis through CT texture analysis in patients with pancreatic ductal adenocarcinoma (PDAC). Material and Methods: 3 cohorts of patients were retrospectively enrolled: 1218 with anatomopathological diagnosis of PDAC; 1383 received surgery of head pancreatic cancer; 1260 with head pancreatic cancer extracted from our radiological archive. Essential for each one were arterial and venous phases in CT examination available in DICOM format; clinical and biological parameters including CA19.9, CEA, CA125, AST, ALT, GGT and bilirubin levels were recorded. Texture features were extracted from 2D regions of interest (ROIs) based on unenhanced, arterial- and portal-phase CT images and compared with the presence of metastasis. Clinical features and texture-derived parameters were compared; receiver operating characteristic (ROC) curves were constructed for the features that showed a significant difference between groups. Kruskal–Wallis and ANOVA tests were used to assess differences of texture parameters between metastatic (M+) and non-metastatic (M−) groups. Results: 321 patients (167 males, 154 females; age range, 30–87 years) were included. Several features were different with statistical significance. Arterial and venous HU_min and HU_mean were significantly minor in M+ (p = 0.00 for all). Arterial and venous SHAPE_Volume_(ml) were significantly greater in M+ (p = 0.00 both). Furthermore, significant differences were found with p <0.05 in other 7 first-order features, in 3 venous GLCM features, in 12 GLRLM features, in 6 NGLDM features, and in 9 GLZLM features. Conclusion: CT texture analysis could predict PDAC metastasis through several parameters with statistically significant differences between M+ and M− groups.