Algorithm 1: Preprocessing step and selecting training patches, the value of S is 100, 75, or 50.

Input: Whole slide images (WSIs) of the digitized prostate biopsy specimens (PBSs). Output: Label the choice patches into the Gleason pattern (GP) classes. Get the histogram equalization. Divide the PBSs into patches, with size S × S pixels. Select convenient training patch Estimate the majority voting for each class in the patch (MV) Calculate two variables for corresponding patch, PC ←Patch Center and BR leftarrow Background ratio If $(MV$==$PC)\&(BR\le 0.5$)        choose the patch Else        Remove the patch