Figure 2

Plot of the final neural network classifier. (A) Please note that due to clarity, only the ten most important features (as per Garson importance) out of 101 are shown. Red lines indicate positive weights, blue lines negative weights. The thicker the line, the stronger the weight. (B) Representative examples (central two-dimensional slices of the respective cluster centers) of the 9 most important LBP textures. Please note that for illustrative purposes, only the central two-dimensional slice within each three-dimensional patch is shown for each texture. For the network analysis, three-dimensional textures have been used.