Figure 1.

Study schematic flow chart. The marker panels used in the model generation consist of 11 candidate biomarkers and 2047 combinations. After combining age and gender variables to each panel, the combination is applied to the five classification algorithms. Selection criteria for optimal biomarker combinations are as follows; (1) the top 10% out of the 2047 sets, (2) sets containing CEA and CA19-9, (3) minimal difference between the training set and validation set, (4) excellent performance independent of the linear and non-linear methods.