Subimages that are classified incorrectly. According to classification results (RBF kernel, C = 211, γ = 2-5), false positives (a)-(c) and false negatives (d)-(f) are shown. Regarding (a), Γ (P, E)/P ratio and Γ (E, W) are high regarding other negatives in training dataset. Therefore, (a) is classified as negative, which actually is positive. The subimage (f) has angiogenesis because of a small pericytes regions is surrounded by endothelium around lower-left corner of image. Mistakenly, this image was classified as negative.