Integrative Approach for Early Cataract Detection using Whale Optimized Convolutional Gated Recurrent Neuronet

Abstract

Background: The development and visual impairment caused by cataracts, the primary cause of blindness in many other nations, must be addressed with regular screening and prompt treatment. Vision distortion is a typical result of eye illnesses such as cataracts. The greatest method of reducing the risk and preventing blindness is early and accurate cataract identification. Research attention has recently been drawn to artificial intelligence-based cataract detection methods.

Purpose: The research enhances diagnosis accuracy and care timeliness, improving clinical outcomes for patients with detected cataract illness.

Methods: The Kaggle Cataract Disease detection dataset is utilized in the research for detection purposes. Resize and color adjustment are used to improve the quality of the images and make feature extraction smoother afterward. Global Contrast Normalization (GCN) is used during the preprocessing step. Utilize the Oriented FAST and rotated BRIEF (ORB) algorithm, and Efficient Net to extract features in images of cataract illness, ensuring a reliable and efficient method for identifying distinguishing characteristics. The research proposed a Whale Optimized Convolutional Gated Recurrent NeuroNet (WOCGRN) to improve diagnosis accuracy and care timeliness, thereby improving clinical outcomes for cataract patients. This novel model combines the spatial learning power of Convolutional Neural Nets (CNNs) with Whale Optimization's Gated Recurrent Unit layers to fine-tune the research model.

Result: Furthermore, the model has been designed to concentrate on symmetrical areas of interest within images, enhancing its sensitivity to microscopic structural alterations linked to cataract disease. Using the Python tool and comparative analysis demonstrates that the model outperforms existing methods. Compared to previous models, the proposed WOCGRN model achieves superior performance metrics: 99.49% accuracy, 99.00% precision, 98.00% recall, 98.00% F1 score, 97% sensitivity, and 96% specificity.

Conclusion: These results underscore the potential of integrative approaches in developing a robust diagnostic tool for the early detection of cataract diseases, ultimately contributing to improved patient outcomes.