The effectiveness of electronic method in detecting retinal microaneurysms in patients suffering from diabetic retinopathy

• Main Authors: Ali Shaeidi, Farhad Kiyanfar
• Format: Article
• Language: English
• Published: Shiraz University of Medical Sciences 2010-12-01
• Series: Interdisciplinary Journal of Virtual Learning in Medical Sciences
• Subjects: electronic learning; region based microaneurysm detection; classification; decision trees; diabetic retinopathy
• Online Access: https://ijvlms.sums.ac.ir/article_45993_663924cdd42cbb5968e50c2d1d6d996e.pdf

Introduction: Microaneurysms (MAs) are one of the types of retinal lesions. MAs are one of the early signs of diabetic retinopathy (DR), which is one of the leading causes of vision loss. The presence of MA on the surface of the retina can severely affect the eye. Detection of MAs in a large number of images generated by screening images, is a very time consuming, cost – effective, and erroneous process. Therefore it is better to use a modern computer method for the detection of MAs. The aim of this study is to compare the different comparison types of automatic images of the retina in patients suffering from diabetes in order to quickly and accurately categorize and diagnose red colored pathology of the retina (microaneurysms), and the diagnosing diabetic retinopathy (DR) in primary stages. Materials and Methods: In this study, the region-based MAs diagnostic method was used for differentiating between these pathologies from other regions, and we also used classifiers such as decision trees, support vector machine and Bayesian networks. Therefore, the extracted data from Bayesian networks, must be trained electronically and prepare themselves for diagnosis and detection in the next stages. Results: In the methods, the sensitivity and the features of the categorization method of the mashin bordare poshtiban was 97.5% and 99.9% the …. Shabakiye method, 96.3% and 99.5%, and the decision tree categorization method, 100% and 98. Conclusion: After comparing the mentioned results, the best diagnosis method was the application of the C4.5 categorizer, and this method is based on decision making trees. This method was completely automatic with the help of a computer and analyzes and diagnosis the images without the interference of the physician. Compared to the clinical diagnosis methods it is more accurate.