Automated algorithm for retinal image exudates and drusens detection, segmentation, and measurement

Exudates and drusens detection and measurement from the retina background makes a significant impact on the diagnosis of retinal pathologies. These diseases usually appear as cotton wall spots, yellowish exudates and drusens (macula degeneration). Information about illness severity can be inferred by the measurement of the sizes of the exudates and drusens and comparing them to the retina background size. In this paper, we have proposed robust algorithm for automatic exudates and drusens detection, segmentation, and measurement on 2D retinal images. The applied methods we suggest for exudates and drusens measuring are mathematic (labeling function) and numerical methods. Numerical methods have more sophisticated calculation steps and can be used to approximate more complicated area of exudates using a poly Area function. In our algorithms, prior to measuring exudates and drusens (AMD), a preprocessing takes place, in which first exudates and drusens detection and segmentation were implemented. For these implemented processes, we applied preprocessing operations, including image filtration, correction of non-uniform illumination, and color contrast enhancement, and then the combined approaches for image segmentation and classification were implemented using: two methods of texture, an adaptive threshold, and morphological operators. Moreover, we introduce methods to eliminate the optic disc completely for exudates detection and measuring. After applying these approaches to a number of images provided from ophthalmologists as well as Drive database, this automated diagnostic algorithm resulted in more accurate yields of exudates and Drusens detection and measurements especially for low intensity and less color contrast images from non-dilated eye pupils. This automated algorithm helps ophthalmologists monitor the progression of diabetic retinopathy.