Detection of Diabetic Macular Edema in Retinal OCT Images with Firefly-Algorithm Selected Features

In modern hospitals, computer algorithms are widely implemented to diagnose the disease in patients using the medical data collected using a recommended protocol. Recently, the diabetes is human group is gradually rising and the early stage can be predicted by screening the eye. Diabetic-Macular-Edema (DME) is one of the harsh diseases and the untreated disease will initiate various eye abnormalities including the vision loss. In clinics, the retinal image recorded with a chosen imaging system is considered to inspect the disease. In this work, Optical Coherence Tomography (OCT) images are considered for the assessment of the DME in the OCT images. The proposed scheme consist the following stages; (i) Image collection and resizing, (ii) Retinal section extraction using joint thresholding and segmentation, (iii) Feature extraction and dominant feature selection with Firefly-Algorithm, (iv) Binary classification and validation. In this work, 500 numbers of Normal/DME class images with pixel dimension 256 × 256 ×1 pixels are considered and a binary classification with a 5-fold cross validation is implemented. The proposed technique helped to get a detection accuracy of 95% with K-Nearest Neighbors (KNN) classifier.