The human eye is a critical sensory organ, and any impairment in its function can significantly affect an individual's quality of life. Eye diseases such as glaucoma, cataract, and retinal disorders can lead to severe vision loss if not detected at an early stage, making timely and accurate diagnosis essential for effective treatment and prevention. In this study, an automated eye disease detection system was developed to address the limitations of existing approaches, including insufficient feature representation, lack of interpretability, and high computational complexity. A comprehensive preprocessing strategy was employed to enhance image quality and improve robustness against variations in input data. The proposed approach effectively learned robust and discriminative features for accurate classification of multiple eye diseases while maintaining computational efficiency. In addition, a visualization technique was incorporated to highlight the important regions influencing the model's predictions, thereby improving transparency and supporting better clinical interpretation, which can contribute to enhanced diagnostic confidence and overall clinical performance. The system was trained and evaluated on a multi-class eye disease dataset and demonstrated consistent improvement in classification performance and interpretability compared to conventional methods. The integration of efficient feature learning and visual interpretability enhances the reliability and practical applicability of the system, making it a promising solution for real-world computer-aided eye disease diagnosis.
Professor) et al. (Thu,) studied this question.