A Novel Machine Learning Approach for Diabetic Retinopathy Detection via Feature Importance Analysis

Authors: Janhvi Arvindbhai Chauhan, Dhaval Maheshkumar Modi

Country: India

Full-text Research PDF File:   View   |   Download

Abstract: The human eye condition known as diabetic retinopathy damages the retina of the eye and can lead to total blindness. To prevent total blindness, diabetic retinopathy must be identified early. Diabetic retinopathy is identified by physical examinations such as optical consistency tomography, pupil dilation, and visual acuity testing. However, it may have an impact on the patients and is time-consuming. In light of these implications, this study uses a machine learning system to identify diabetic retinopathy in the human eye. The suggested approach uses classification algorithms on a number of characteristics of an existing Diabetic Retinopathy dataset, such as optical disk diameter, lesion-specific characteristics (microaneurysms, exudates), or the presence of hemorrhages. After that, the traits were taken out and applied to the ultimate decision-making process to determine whether diabetic retinopathy was present. The suggested approach made use of logistic regression and decision trees. For the prediction, use a support vector machine. The suggested approach outperformed the current efforts by achieving 89% accurate outcomes. This further demonstrates the vastness of the suggested approach.

Keywords: Hard Exudates, Logistic Regression, Feature Importance, Support Vector Machine

Paper Id: 232519

Published On: 2023-06-09

Published In: Volume 11, Issue 3, May-June 2023

Share this