Abstract: Objective To explore the value of computer-aided diagnosis of hepatic carcinoma with digital subtraction angiography (DSA) images by back propagation (BP) neural network. Methods DSA images of 32 cases with hepatic carcinoma were analyzed. According to the analysis results, computer-aided diagnostic model by using BP neural network was established. Then 100 random cases (55 cases without lesion, 45 cases with lesion) were input to verify diagnosis accuracy of the model. Results Computer-aided diagnostic model of hepatic carcinoma was established successfully. One hundred random sample data were analyzed by the model. The true positive rate was 91%, and the true negative rate was 95%. Conclusion BP neural network can be used as an assisted method for DSA diagnosis of hepatic carcinoma.