Detecting Eye Diseases from Oct Images Through Deep Learning

About 23% Indians suffer from eye disorders. The main risk causing loss of vision includes Choroidal Neovascularization (CNV), Diabetic macular edema (DME), Multiple drusen , Normal retina . Diagnosis of the above eye conditions can be done with the imaging technique Optical Coherence Tomography (OCT). It allows the retina specialist to visualize the retina and take different slices early. Manual diagnosis of different diseases by ophthalmologists is very time-consuming and prone to errors due to ocular anomalies. Early detection of eye diseases will prevent severe complications such as total blindness. Though the current technique for eye diseases uses fundus photographs to train deep learning models, the method seems ambiguous from the perspective of understanding data since its final results do not inspire confidence. A way out would be the extraction of relevant information from medical Ocular Coherence Tomography (OCT) images through a deep learning model. This requires the use of OCT images followed by a convolutional neural network (CNN), and feature extraction for the early identification of diseases. The proposed work shows the efficiency of VGG16 and RESNET-50 MODELS in predicting various eye disorders at an early stage using OCT images. The expected results showed higher accuracy of the proposed model compared to other existing systems. The proposed model classifies retinal OCT images into four categories: NORMAL RETINA, DME, CNV, and DRUSEN.