Our previous study reported a method of using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry to analyze the association between abnormal fucosylation of serum glycoproteins and the progression of hepatitis B virus (HBV)-associated hepatocellular carcinoma (HCC). In the present study, the aforementioned method was improved by focusing on fucosylated glycoproteins <10 kD, classification models were established and blind tests were performed on an enlarged sample size (n=299). According to the present results, the classification models had a sensitivity and specificity of 74.31 and 76.32%, respectively, to identify HCC among all serum samples, 81.65 and 83.08%, respectively, to distinguish HCC from HBV-associated cirrhosis and chronic hepatitis Band 88.99 and 84.62%, respectively, to distinguish HCC from HBV-associated cirrhosis. When combined with α-fetoprotein (AFP) measurements (AFP >20 ng/ml), the sensitivity and specificity of the models were significantly elevated to 80.73 and 87.37%, 87.16 and 90.00%, and 92.66 and 93.84%, respectively. In addition, the HBV-HCC vs. HBV-cirrhosis classification model was used to analyze serum samples collected from 9 patients with cirrhosis 1 year before they were diagnosed with HCC, and from 6 patients who had cirrhosis but developed no signs of HCC for the following 3 years. The model identified 7 patients (77.78%) with no significant clinical symptoms of HCC, and gave no false positive results, demonstrating that the classification models established in the present study may be useful for the early diagnosis of HCC. After isolation and purification, two proteins with differential expression were identified as isoform 1 of inter-α-trypsin inhibitor heavy chain 4 precursor, and thymosin β-4-like protein 3. These may be used as candidate markers for HCC diagnosis. Additionally, the present study indicates that defucosylation of serum glycoproteins may occur during the development and progression of HCC.