The objective of this study was to analyze the value of artificial intelligence algorithm-based computerized tomography (CT) image combined with serum tumor markers for diagnoses of pancreatic cancer. In the study, 68 hospitalized patients with pancreatic cancer were selected as the experimental group, and 68 hospitalized patients with chronic pancreatitis were selected as the control group, all underwent CT imaging. An image segmentation algorithm on account of two-dimensional (2D)-three-dimensional (3D) convolution neural network (CNN) was proposed. It also introduced full convolutional network (FCN) and UNet network algorithm. The diagnostic performance of CT, serum carbohydrate antigen-50 (CA-50), serum carbohydrate antigen-199 (CA-199), serum carbohydrate antigen-242 (CA-242), combined detection of tumor markers, and CT-combined tumor marker testing (CT-STUM) for pancreatic cancer were compared and analyzed. The results showed that the average Dice coefficient of 2D-3D training was 84.27%, which was higher than that of 2D and 3D CNNs. During the test, the maximum and average Dice coefficient of the 2D-3D CNN algorithm was 90.75% and 84.32%, respectively, which were higher than the other two algorithms, and the differences were statistically significant ( P < 0.05 ). The penetration ratio of pancreatic duct in the experimental group was lower than that in the control group, the rest were higher than that in the control group, and the differences were statistically significant ( P < 0.05 ). CA-50, CA-199, and CA-242 in the experimental group were 141.72 U/mL, 1548.24 U/mL, and 83.65 U/mL, respectively, which were higher than those in the control group, and the differences were statistically significant ( P < 0.05 ). The sensitivity, specificity, positive predictive value, and authenticity of combined detection of serum tumor markers were higher than those of CA-50, CA-199, and CA-242, and the differences were statistically significant ( P < 0.05 ). The results showed that the proposed algorithm 2D-3D CNN had good stability and image segmentation performance. CT-STUM had high sensitivity and specificity in diagnoses of pancreatic cancer.