Design and Development of a Cataract Detection Robot Using Deep Convolutional Neuron Networks

A cataract is an eye disease that is the most common cause of blindness in Thailand and many countries around the world. In Thailand, 80% of all eye diseases can be prevented or cured. It is highly effective if treated early, but is seen as a condition that tends to develop gradually and become more severe over time, resulting in most patients being complacent until facing severe consequences later. The organizers have seen the problems and importance of the patient's admission for cataract treatment. Therefore, it intends to study the process of sorting out various abnormalities in cataract screening used for initial corneal evaluation and apply it to machine learning to train datasets using LeNet-Convolutional Neural Network (LeNet-CNN) and Support Vector Machine (SVM). The results obtained from both models were compared for efficiency and accuracy.

The experimental results showed that LeNet-CNN gave 96% accuracy, 95% sensitivity, and 96% specificity, while SVM had 92% accuracy, 91% sensitivity, and 94% specificity. This means that both processes had relatively high accuracy, especially LeNet-CNN, which had higher accuracy than SVM. It can be concluded that the cataract abnormalities were isolated using a deep convolutional neural network and were able to correctly sort them out. It must be precise and efficient. It can be used to diagnose cataracts from photos for initial corneal assessment for patients to be aware of abnormalities and be treated promptly.