Silent diseases are those which symptoms are difficult to detect in early stages, for this reason, in most cases this type of diseases are detected when their treatment is very complex and risky. However, they can change the eye appearance as the disease progresses. Therefore, the development of an algorithm able to detect silent diseases based on ocular fundus images is relevant to give appropriate treatment. This paper presents the classification of normal ocular fundus (control class) and three silent diseases: glaucoma, pathological myopia, and cataracts. These classification classes were selected because they present the most significant differences between each other in ocular fundus images. In order to make the results more accurate, it was necessary to use images with the same format, and so, the images were preprocessed. After having performed a robust set of experiments for the classification methods and descriptors, the best performing method was SVM, as the classification method, combined with HOG, as the descriptor. The final method had a precision of 0.61, a recall of 0.477 and an F-score of 0.517. In conclusion, the developed method showed some difficulties classifying glaucoma, however, the classification of pathological myopia seems to be appropriate compared with methods found in literature.