Abstract:
CAD systems for helping Medical diagnostic are spread worldwide and are quite famous. The
primary objective of these systems is to provide support to the doctors and officials in the process of
diagnosis and to deliver the best services to the patients. Similar systems are also being used for the
detection of disease such as diabetic macular edema in some advanced countries. Diabetic macular
edema can be categorized as an abnormality of retinal function which can lead to severe vision loss
and affects the whole human visual system at its worst stage. In this research study, an automated
system for assisting the ophthalmologist for detection and classification of diabetic macular edema
is proposed. The automated system will work on the digital retinal images captured from specified
devices and will try to isolate the macula from other components of human visual system from the
retinal images. Once the macula is available, a Gaussian Mixture Model (GMM) classifier will
mark the possible regions as exudates or non-exudates. The classifier then works on the binary map
and the system then classifies the retinal image as one of the pre-determined stages of macular
edema. For evaluating the effectiveness of the proposed system, statistical and comparative study
has been performed on the proposed system and the recently proposed and published researches of
the same domain by utilizing the freely available retinal image databases online. The effectiveness
of the system is measured by evaluating the improvements in system’s performance and accuracy.
