AUTOMATIC DIABETIC RETINOPATHY DETECTION AND ESTIMATION FROM MULTISPECTRAL IMAGES USING MACHINE LEARNING ALGORITHMS

摘要

#$%^&*AU2020100868A420200813.pdf#####AUTOMATIC DIABETIC RETINOPATHY DETECTION AND ESTIMATION FROM MULTISPECTRAL IMAGES USING MACHINE LEARNING ALGORITHMS ABSTRACT In our day to day life, eyes play a significant role in human activities. Diabetes has become one of the most significant general medical issues to date. Diabetic retinopathy, which is otherwise called a diabetic eye illness that influences up to 80% of every patient who has diabetes, necessarily proposes visual deficiency. Diseases like exudates, hemorrhage, glaucoma, diabetic, and retinopathy microaneurysm, can be discovered in more advanced stages by employing retinal images. Diabetic Retinopathy classification is a standard and tedious method that needs a qualified ophthalmologist to look at and estimate the digital fundus photographs of the retina. Spectral imaging is used in numerous fields of industry and scientific research as it consists of both spatial and spectral data. The approach of using spectral image enhancement methods is to promote the diagnostic performance of medical image technologies similar to retinal imaging. Computer machine learning more advanced techniques, for example, Convolutional Neural Networks (CNNs), have risen as a viable tool in medical image analysis for the detection and classification of disease in various ways progressively. Machine learning and Image processing techniques help to diagnose the different disorders before utilizing the retinal image. The retinal image is employed to recognize diabetes in the beginning stages by estimating retinal blood vessels collectively. The main aim of this invention is to identify disorders in retinal images using machine learning techniques. Results show that the neural network is superior to the other methods for vessel classification.