Executive Summary : | Mostly all people have eye diseases, in some minor cases, they go away with their own and the rest require specialist care. The common eye diseases are refractive errors, age-related macular degeneration (ARMD), cataract, diabetic retinopathy (DR), diabetic macular edema (DME), glaucoma, dry eye, amblyopia, strabismus, pathological myopia, choroidal neovascularization (CNV), etc. Globally, at least 2.2 billion people have a near or distance vision impairment as per the World Health Organization (WHO) in 2021. In India, apart from other diseases, dry eye is an emerging threat, which affect children to older age people. In this proposal, the aim is to develop new methodologies which can easily identify eye diseases from ophthalmic images such as fundus imaging, optical coherence tomography (OCT), meibography, etc. The developed methodologies will be an advanced and computationally efficient version of deep learning and transformer approach. For the localization of disease lesions, meibomian glands, optic disc and optic cup, the Panoptic segmentation approach will be developed, which is based on semantic and instance segmentation. Thereafter, the segmentation output of each disease will be used for classification and grading. In medical imaging, accurate segmentation, classification and grading of diseases are most important and thus methodologies will be designed such that they can accurately diagnose disease in real-time. For the assistance of doctors, Graphical User Interface (GUI) software will also be developed which will contain all the developed methodologies for automatic and accurate diagnosis of patient in early stage. Thus, this project will contribute in reducing the blindness and low vision loss among people. |