Background: Vascular invasion is closely related to the prognosis of hepatocellular carcinoma (HCC). Increasing evidence suggests that miRNAs can serve as biomarks to predict prognosis in various tumors. Thus, the aim of this study was to develop a novel, vascular invasion-related miRNA signature for prediction of HCC prognosis.Methods: Differentially expressed miRNAs (DEMs) between HCC samples with vascular invasion and without vascular invasion were obtained from GSE67140. MiRNAs expression profiles and clinical information for 344 patients were collected from The Cancer Genome Atlas database, and the patients were randomized (1:1) to training set and validation set. LASSO regression model was employed to identify survival-associated DEMs and establish risk score in the training set. Moreover, risk score was verified in the validation set. And nomogram based on risk score and clinical information was constructed to improve the prediction of prognosis. Meanwhile, four online tools were used to predict target genes and enrichment analysis was utilized to explore the biological pathway of the miRNAs.Results: A novel three-miRNA signature was screened including hsa-mir-210, hsa-mir-149 and hsa-mir-99a, and risk score was established for HCC prognosis prediction. Patients were divided into the low-risk group and high-risk group according to risk score. High-risk group both have higher hazard of death compared with low-risk group in training set and validation set. And the 5-year AUC of risk score were 0.718, 0.674 and 0.697 in training set, validation set and the total set, respectively. The C-index of nomogram was 0.724, and calibration curves showed nomogram had high concordance to predict 1- ,3- , and 5-year survival rates among HCC patients. Furthermore, enrichment analysis identified several tumor-associated pathways including Ras signaling pathway, PI3K-Akt signaling pathway, and MAPK signaling pathway and so on, which may contribute to explain the potential molecular mechanisms of above miRNAs.Conclusion: This study developed a risk assessment model based on three miRNAs, which could accurately predict the prognosis of HCC.