Optimized Deep Cnn for Robust Dr Detection and Classification Using Remora Tuna Swarm Algorithm

Diabetic Retinopathy (DR) is a serious eye disease linked to Diabetes Mellitus, a condition causing high blood sugar levels that damage the retina's blood vessels, often leading to vision impairment or blindness if untreated. DR is a major cause of blindness in adults, especially in developing countries, with a growing global prevalence. Early and accurate detection is crucial to prevent irreversible damage, yet traditional manual diagnosis is time-intensive and costly. This study proposes a novel deep learning approach to enhance DR detection accuracy through the integration of Deep Convolutional Neural Network (DCNN) and Remora Tuna Swarm Optimizer (RTSO) algorithm, aiming to improve diagnostic efficiency and outcomes for at-risk individuals. To enhance image quality, gaussian filtering is employed for preprocessing, effectively reducing noise while preserving critical retinal details. Segmentation of retinal features is carried out using Sparse Fuzzy C-Means (SFCM) clustering, providing precise region extraction. The optimized DCNN, tuned with RTSO(DCNN-RTSO), classifies DR with high accuracy. The model's performance is evaluated using key metrics, achieving accuracy, Recall and Specificity of 98.12 %, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 2. 5 6 \%}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 8. 9 \%}$</tex> respectively. This hybrid technique is a promising tool for the early diagnosis of DR in clinical settings because of its strong classification performance.