Introduction: The World Health Organization estimates that the number of diabetics will increase from 130 million to more than 350 million by the next 25 years. Diabetes can rapidly lead to cardiovascular disorders and a variety of problems in the retina. The adverse effects of diabetes on retina are known as diabetic retinopathy (DR). In this connection, the purpose of this paper is to investigate the diagnosis of spot-shaped red color retinal pathologies, or hemorrhages from retinal colored radiographs through electronic learning and computer, that is automatically, as early as possible.Materials and Methods: A set of 68000 pixels including 35000 hemorrhage and 33000 non-hemorrhage pixels were extracted from 85 colored retinal images. The morphological and Pixel level hemorrhages recognition techniques were used to differentiate these images from other image structures. The retinal lesions were classified into hemorrhage and non-hemorrhage, using a classifier known as decision trees. Finally, the results obtained from this system were compared with those diagnosed by ophthalmologists.Results: In the testing stage, after extracting and classifying the 68000 pixels from retinal images using classifier decision trees formula, this method achieved 98% sensitivity, 97.14% specificity and 97.57% accuracy.Conclusion: The computer-aided diagnosis techniques, the morphological techniques, have a high efficiency and are more precise than the clinical techniques.