Abstract
Early detection of cataract is crucial for thwarting visual impairment worldwide, and the utilization of automated cataract detection through medical images has shown increased growth for several years. The automated detection model comprises image processing, feature extraction and classification process to ensure accurate identification of infected cataract eye images. However, the recent detection model endures various challenges, including complex computing requirements, feature redundancy, inadequate precision, generalization and less data diversity. To overcome these challenges, a Chaotic Adaptive Poplar-Bacteria Optimization (Cha-PO) based Cataract VisionNet (CVNet) method is proposed to enhance diagnostic accuracy and operational efficiency. The Cha-PO model is specifically used for optimal feature selection of fundus images by reducing the dimension of the images, which ensures acute diagnostic data preservation. CVNet model used for classifying the cataract images by applying the deep hierarchical learning mechanisms alongside optimized network parameters to boost accuracy levels and operational reliability. The proposed approach is validated using the Eye Cataract Kaggle dataset, and it outperforms the traditional models with 99.10% accuracy, 99% precision, 99.21% recall and 99.10% of F1-Score. With a 99s execution time, it requires fewer computational resources than other baseline models, making it suitable for medical diagnosis.