Performance analysis of machine and deep learning techniques for early glaucoma and stargardt disease detection

Abstract

Glaucoma and Stargardt disease are vital diseases that are concerned with the retinal portion of the eye. These diseases cause blindness when not accurately detected at an early stage. Many researchers performed their research on early stage detection of Glaucoma and Stargardt disease and accurate detection was not achieved with in less time. Image segmentation methods are also used to divide the images to accomplish disease detection in minimal time. In addition, classification methods are employed to categorize the retinal fundus images for disease detection. However, the error rate was not minimized in the classification process. To solve these problems in disease detection, novel classification methods are introduced in this research for achieving effective disease detection at an early stage. newlineIn the first research work, Adaptive bilateral Filterative Morlet Histogram Thresholding Segmentation based Multi-Layer Log-Linear Classification (ABFMHTS-MLLC) is designed to determine the glaucoma disease at an early stage. Retinal fundus images are considered to identify the glaucoma disease. Initially, adaptive bilateral filter is employed to ignore the unwanted noise from the input image for increasing the quality of input images. Then, the image that is without noise is forwarded to extract the pertinent features such as color, texture and intensity of retinal fundus image by using Morlet Transformation. Fuzzy Histogram Thresholding Segmentation is used to separate the images into several regions. Log-Linear analysis is lastly used to classify the retinal fundus images into normal or glaucomatous for accurate disease detection. newline newline