Machine Learning-Based Diabetic Retinopathy Detection Using Iris Texture Feature Extraction

This study aims to detect retinal complications of diabetes, a serious condition affecting diabetic patients and harming the retinal veins and arteries, which are the light-sensitive part at the back of the eye. Diabetic retinopathy (DR) has various stages, from mild to critical, potentially causing vision problems. Early detection of DR is essential for timely treatment to prevent vision loss. Previous research primarily used complex methods to classify only binary DR and No DR cases using complex deep learning methods. However, a lightweight machine learning model is proposed to accurately categorize different DR stages. The system takes retinal fundus images through steps, including enhancing the images, extracting features using Gray-Level Co-occurrence Matrix (GLCM) and Gabor filters, and reducing dimensionality with principal component analysis (PCA). In the multi-class scenario, the Light Gradient Boosting Machine (LGBM) classifier achieves 92.83% accuracy, enhancing the ability to detect retinopathy effectively using optimized machine learning techniques. The study holds potential clinical applications, and future steps involve improving datasets and model robustness.