Deep Learning Approaches for Cataract Detection Using Fundus and Real-Time Naked Eye Imaging

Cataracts are the most common eye disease and the leading cause of blindness worldwide. The eye ailment known as cataracts results in clouded lenses, leading to vision that appears foggy or frosted. People with cataracts often struggle with tasks like reading, driving, or recognizing faces due to the impaired clarity of their vision. Effective treatment and prevention of vision loss depend on the early detection of cataracts. The proposed system is a deep learning-based approach to detect cataracts from the naked eye and fundus images that can help ophthalmologists diagnose cataracts earlier, more accurately, and efficiently. Both fundus and naked eye images are processed by the system to detect cataracts. The system offers a dependable and automated solution for cataract detection, enabling timely diagnosis and treatment. Its performance is on par with that of ophthalmologists, highlighting its potential as an effective and trustworthy tool for automated cataract detection. The proposed model employs two distinct deep learning approaches. For identifying cataracts using naked eye images, a standard CNN architecture was utilized, achieving an accuracy of 94.26%. For detecting cataracts in fundus eye images, a Deep Convolutional Neural Network (DCNN) with the VGG19 architecture was implemented, demonstrating a notable accuracy of 96.78%. This model holds substantial promise, offering a non-invasive and accessible method for cataract screening. It is particularly helpful in environments with limited resources, where access to expert eye care is frequently limited.