Objective To explore the value of AKR1B10 combined with GPC-3 in improving the sensitivity and specificity of immunohistochemical diagnosis of hepatocellular carcinoma (HCC). Methods The microarray including 75 HCC and adjacent tissues was subjected to immunohistochemistry detection of AKR1B10 and GPC-3 expression. A Logistic regression diagnostic model was established using the results of tissue microarray (training group). The ROC curves (the receiver-operating characteristic curve) and area under the curve (AUC) were used to evaluate the sensitivity and specificity of AKR1B10, GPC-3 or their combination. The Logistic regression diagnostic model was validated with 200 HCC and adjacent tissues (testing group). Results For the training group, the AUC values of AKR1B10, GPC-3, and AKR1B10 combined with GPC-3 were 0.773, 0.800, and 0.931, respectively. The sensitivity of AKR1B10 and GPC-3 were 56% and 61.3%, respectively, and their specificity was both 98.7%. AKR1B10 combined with GPC-3 yielded a sensitivity of 88.0% and a specificity of 97.3%.For the testing group, sensitivity and specificity of AKR1B10 combined with GPC-3 were 97.0% and 96.5%, respectively. Conclusion AKR1B10 combined with GPC-3 can greatly improve the sensitivity and specificity of HCC immunohistochemical diagnosis, and it should be used when necessary in addition to the routine pathological assessment.