On the extraction of retinal area from SLO images using RBFN classifier and its comparison to the optimally trained ANN classifier

Abstract

Retinal diseases can be detected earlier by scanning laser ophthalmoscope (SLO). Automated detection of retinal area from the SLO images is a crucial task. With the invention of screening technology, the large retinal part must be imaged for better diagnosis of the retinal diseases. During the process of imaging, artifacts (eyelashes and eyelids) are come along with the retinal part. So removal of artifacts is a big challenge. In this chapter, true retinal area is extracted from the SLO image based on machine learning approach for better diagnosis of retinal diseases. To reduce the complexity of an image, the image is divided into group of pixels which is based on its compactness, colour and regional size then that group of pixels are called Superpixel. Then, classifier is used to classify the true retinal area and artifact. The results provide better performance with 96% accuracy. Comparison is performed with respect to accuracy and computational time between ANN and RBFN.