A non-invasive approach for diagnosing diabetic retinopathy using stacked ensemble deep learning mechanism incidental towards chronic kidney disease

Chronic kidney disease (CKD) is progressive in several parts of the world. The prevalence of CKD is rapidly increasing due to dis-‎eases such as hypertension and diabetes, the two major causes of CKD, and due to the global aging population. Non-invasive methods play ‎an important role in identifying CKD. Retinal photography nowadays is a source that provides vital information related to the eye as well ‎as systemic vasculature and acts as a non-invasive diagnostic test. The physiological, developmental, and pathogenic pathway information ‎shared by the eye and the kidney seems to be similar. With the advent of Deep Learning, the detection of CKD and diabetic retinopathy is taking new ‎horizons. In this paper, a novel mechanism is proposed for identifying Diabetic Retinopathy, which highly influences CKD progression. ‎Also, CKD prediction is done using retina images. To accomplish a notable detection, A lightweight and efficient Transfer Learning based ‎architecture, “RIX Net,” has been employed in classifying diabetic retinopathy severity as well as CKD. Five datasets of various sizes with ‎respect to Diabetic Retinopathy and one corresponding retinal dataset to CKD have been used to measure the performance of the proposed ‎architecture. The architecture performed well with around 90% performance in all cases‎.