International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 11 Issue: 05 | May 2024
www.irjet.net
p-ISSN: 2395-0072
A Comparative study for Detection of stages of diabetic retinopathy of diabetic patients using machine learning Darshit Jivrajani 1, Riya Kaku 2, Sandip Panchal 3 1Research Scholar, Dr. Subhas University 2Assistant Professor, B.H.Gardi College, GTU
3Assistant Professor, Dr. Subhas University, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - People in their middle years who work are most
great performance and high adaptability. Deep learning algorithms, also referred to as deep neural networks, are essentially an offspring of artificial neural network architectures with a larger number of hidden layers. Convolutional Neural Networks (CNNs) are a kind of feed forward network and one of the most popular deep neural network models. A synopsis of popular and well-respected CNN models is provided in this presentation. A comparative assessment of these models with various parameter considerations is also included.
commonly suffering from diabetic retinopathy. Due to inadequate resources, diabetic retinopathy can be tough to identify in remote regions. Many scientific and medical techniques are available to For the most part, the disease is examined for and identified by retinal fungal imaging. Here, the goal is to be capable to automatically screen images for retinopathy caused by diabetes. This will be achieved via developing an application for smartphones that makes use of fundus images and a model developed using machine learning to evaluate an eye image and calculate the patient's level of blindness. This strategy will help shorten the time required for diabetic retinal disease screening.
Abhishek Samanta [2] - Automated detection of diabetic retinopathy using convolutional neural networks on a small dataset - Patients with either type 1 or type 2 diabetes may experience a condition known as diabetic retinopathy. The condition must be detected early on since complications like glaucoma, vitreous hemorrhage, and retinal detachment can impair vision. Both non-proliferative and proliferative diabetic retinopathy are the two main stages of the disease. This work presents a transfer learning based CNN architecture on color fundus photography that recognizes classes of Diabetic Retinopathy from hard exudates, blood vessels, and texture, and performs reasonably well on a much smaller dataset of skewed classes of 3050 training images and 419 validation images. With its limited processing capability, this model can function rather well in small realtime applications, as it is incredibly lightweight and durable, which could expedite the screening process. Google Colab was used to train the dataset. With four classes—I, No DR, ii) Mild DR, iii) Moderate DR, and iv) Proliferative DR—we trained our model. We obtained a Cohens Kappa score of 0.8836 on the validation set and 0.9809 on the training set.
Key Words: Diabetic retinopathy, Machine learning, Fundus photographs, CNN, Android application.
1. INTRODUCTION An excessively elevated blood sugar level can lead to diabetes, a chronic organ disease. Individuals suffering from diabetes may get diabetic retinopathy, a major cause of vision loss resulting from chronic damage to the blood vessels in the eyes. It is common to categorize diabetic retinopathy as either nonproliferate or proliferate. Blood vessels start to degenerate and more fluid enters the eye during the first stage of DR, or NPDR. In this phase, there are three different structures that are visible: hemorrhages (small area of blood streaming into the retina), exudates (drop of fatty tissues), and micro aneurysms (red dot). Proliferate diabetic retinopathy is a condition where there is a reduction in blood flow due to the constriction of blood vessels in the retina. PDR reduces eyesight in both the center and periphery.
Thippa Reddy Gadekallu [3] - Deep neural networks to predict diabetic retinopathy - A major contributor to blindness in the elderly is diabetic retinopathy, which has emerged as a major global health concern in recent years. Retinal fungal imaging is mostly used for the majority of the disease's detection, while there are other scientific and medical methods for screening and diagnosing it. The current study employs deep neural networks based on principal component analysis. Classification of the collected features from the diabetic retinopathy dataset using a network model and the Grey Wolf Optimization (GWO) technique. Selecting the best parameters for training the DNN model is made possible by the use of GWO. In this study, the diabetic
1.1 Literature review Sanskruti Patel [1] - A Comprehensive Analysis of Convolutional Neural Network Models - Image, speech, and text processing are just a few of the many domains that deep learning—an emerging subject of machine learning—applies to with high success rates. In comparison to conventional machine learning techniques, experiments demonstrate its
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 2052