Cracking Light on Cataract Detection by Implementing VGG16 Transfer Learning-Based Model on Fundus Images
Gunjan Sharma, Vatsala Anand, Sheifali Gupta
Abstract
Ocular conditions, including cataracts, provide a substantial worry owing to their capacity to result in vision loss if left undetected. The importance of early discovery using fundus imaging cannot be overstated; yet, the process of manual categorization is prone to human error and can be time-consuming. This study introduces a deep learning methodology for automated classifying ocular fundus images into 'cataract' and 'normal'. The model utilized in this study is based on the VGG16 convolutional neural network architecture, which is widely recognized for its effectiveness in image classification tasks. This model's training is conducted using the ODIR dataset, which consists of a total of 1088 images. The results are encouraging: the proposed model demonstrated a high accuracy rate of 96.05%, a precision score of 0.95, a recall value of 0.96 and a loss value of 0.86. The findings of this study not only indicate the viability of employing deep learning techniques for the classification of fundus images but also underscore the potential of these models in supporting healthcare practitioners in the timely and precise identification of diseases. This research represents a significant advancement in the utilization of artificial intelligence within the medical field of ophthalmology, with the primary objective of enhancing patient outcomes by facilitating prompt and accurate diagnoses.