Manuscript received October 25, 2025; revised November 17, 2025; accepted January 26, 2026; published April 28, 2026.

Abstract—Diabetic Retinopathy (DR) is an ophthalmic complication of diabetes, which is one of the leading causes of vision impairment. Accurate grading of DR severity is essential for early diagnosis and timely intervention. However, manual diagnostic methods are time-consuming, resource-intensive, and error-prone. Contrarily, Computer- Aided Diagnosis (CAD) systems offer a faster, more consistent, and potentially more accurate alternative, while facilitating effective screening and treatment planning. This research proposes the feature selection approach of Adaptive Mutation Probability based Starfish Optimization Algorithm (AMP-SFOA) for automatic diagnosis and classification of DR into different categories. The proposed method is evaluated using two standard dataset, Asia Pacific Tele- Ophthalmology Society 2019 Blindness Detection (ATPOS- 2019) and Diabetic Retinopathy Detection (DDR) based on training and testing. Preprocessing steps, including image resizing, Contrast Limited Adaptive Histogram Equalization (CLAHE), and data augmentation, are performed to enhance the input data. Finally, the Long Short-Term Memory (LSTM) is employed for classifying DR into multi-classes. Experimental results demonstrate that the proposed AMPSFOA approach attains better accuracies of 99.476% and 96.958% on the APTOS-2019 dataset and the DDR dataset respectively, compared to existing methods such as RT2Net and Swin-TransformerV2 Multi-Branch Fine-grained Diabetic Retinopathy Grade Classification (STMF-DRNet).

Keywords—adaptive mutation probability, contrast limited adaptive histogram equalization, diabetic retinopathy, long short-term memory, starfish optimization algorithm

Cite: B. R. Raghu, Janapati Venkata Krishna, and N. Pradeep, "Feature Selection Using Adaptive Mutation Probability Based Starfish Optimization Algorithm in Diabetic Retinopathy Classification," Journal of Image and Graphics, Vol. 14, No. 2, pp. 325-337, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.