Objective To establish a serum protein fingerprinting technique with a pattern matching algorithm to distinguish Dukes A stage from colorectal cancer of Dukes B,C,D stage. Methods Serum samples were collected from both reserved group and test group, each of the groups comprised 10 patients of colorectal cancer in Dukes A stage and 68 patients in Dukes B,C,D stage. The sera of the reserved group was combined with the surface of the IMAC3 proteinchip. Then the data of SELDI TOF MS was read and analyzed by BioMarker Wizard software and BioMarker Pattern software to get a classificatory pedigree tree, which is a standard configuration that can distinguish the sera of colorectal cancer patients of Dukes A from colorectal cancer patients of Dukes B,C,D, and the standard was confirmed by double blind test in the test group. Results At the M/Z values of 8 320 Da, 8 604Da, 8 867Da and 15 872Da, the protein contents in reserved group are obviously different between the two classes of patients. The accuracy of classification was 97 4%(76/78), corresponding sensitivity was 100%(10/10) and corresponding specificity was 97 1%(66/68). By double blind examination in test group, the corresponding accuracy was 100%(10/10), the corresponding sensitivity was 94 1%(64/68) and the corresponding specificity was 94 1%(64/68). Conclusion Colorectal cancer of Dukes A can be quickly and exactly diagnosed by this method with high sensitivity and specificity. That will be widely used in clinical application