AI-Driven Diabetic Retinopathy Detection Using ILWOA-Enhanced Extreme Learning Machine on EyePACS and APTOS Datasets

Introduction: A serious consequence of diabetes, diabetic retinopathy (DR) damages the retinal tissues besides can cause severe vision loss or complete blindness in some people. The only way to slow or stop the worsening of this ailment is to catch it early. The minor signs of DR might be difficult to discern on one's own, making early detection a challenge. An innovative approach to DR detection is shown. Objectives: In this study, which utilises an Extreme Learning Machine (ELM) that has been optimised utilising an Improved Logical Whale Optimisation Algorithm (ILWOA). In order to improve classification accuracy, convergence speed, and robustness, the ILWOA adjusts the parameters of the ELM model. Methods: Two benchmark datasets, EyePACS and APTOS, are used to test the methodology, making sure the model can be used for various real-world data circumstances. Overcoming issues like overfitting and sensitivity to parameter initialisation, the suggested method uses sophisticated optimisation techniques to circumvent the shortcomings of conventional ELM models. The ILWOA-optimized ELM achieves better performance metrics and surpasses existing approaches in detecting different stages of DR, according to the experimental data. The approach is also flexible and scalable, so it can handle massive ophthalmic datasets. Conclusions: This study makes a substantial influence to the area of automated healthcare diagnostics by developing an AI-driven system for the identification of diabetic retinopathy that is dependable, efficient, and scalable. To further enhance the proposed framework's applicability in clinical applications, future efforts should focus on studying real-time implementation and multi-modal data integration.