The document discusses a proposed method for the automatic detection of non-proliferative diabetic retinopathy (DR) using fundus images, targeting early diagnosis to reduce blindness caused by the disease. The developed algorithm achieved a sensitivity of 96.3% and specificity of 95.1% in identifying lesions such as microaneurysms and hemorrhages. The research emphasizes the importance of automated systems in aiding ophthalmologists and improving the screening process for diabetic retinopathy.