Analysis of Retinal Images for Accurate Drusen Extraction

Abstract

Medical systems based on state of the art image processing and pattern recognition techniques are very common now days. These systems are of prime interest to provide basic health care facilities to patients and support to doctors. Age related Macular degeneration is the leading cause of severe vision loss in people over age 60. It occurs when the small central portion of the retina, known as the macula, deteriorates and can be a source of significant visual disability. Drusen represents the only visible sign of Macular degeneration in patients. However, to diagnose the “bright lesions “associated with age-related macular degeneration (AMD), namely drusens, lesions must be differentiated from exudates, the “bright lesions” associated especially with diabetic retinopathy, which can have similar appearance. In this thesis, we propose image processing and pattern recognition based system to automate the detection of drusens and the system is capable of differentiating among these different lesion types, it assists the ophthalmologists in early detection of the disease. The proposed system applies basic image pre-processing steps to enhance the lesions present in the given images. On the basis of these enhanced images, the system creates the features vector for bright lesion type. Bright lesions and background pixels are classified through KNN classifier. Feature vector for the bright regions are than computed for exudate and drusen discrimination. At the end SVM Classifier detects and differentiates among different lesion types by using given feature set. The statistical analysis is performed on publicly available standard retinal image databases