Automatic disease diagnosis using biometric images is a difficult job due to image distortion, such as the presence of artifacts, less or excessive light, narrow vessel visibility and differences in inter-camera variability that affect the performance of an approaches. Almost all extraction methods in the blood vessels in the retina produce the main part of the vessel with no patalogical environment. However, an important problem for this method is that extraction errors occur if they are too focused on the thin vessels, the wide vessels will be more detectable and also artificial vessels that may appear a lot. In addition, when focusing on a wide vessel, the extraction of thin vessels tends to disappear and is low. Based on our research, different treatments are needed to extract thin vessels and wide vessels both visually and in contrast. This study aims to adopt feature extraction strategies with different techniques. The method proposed in segmentation and extraction with three different approaches, namely the pattern of shape, color, and texture. Testing segmentation and feature extraction using STARE datasets with five classes of diseases namely Choroidal Neovascularization, Branch Retinal Vein Occlusion, Histoplasmosis, Myelinated Nerve Fibers, and Coats. Image enhancement on Myelinated Nerve disease Fiber is the best result from the image of other diseases with the highest value of PSNR of 35.4933 dB and the lowest MSE of 0.0003 means that the technique is able to repair objects well. The main significance of this work is to increase the quality of segmentation results by applying the Otsu method. Testing of segmentation results shows improvements with the proposed method compared to other methods. Furthermore, the application of different feature extraction methods will get information on disease classification features based on patterns of suitable shapes, colors, and textures.