Comparative Analysis of CNN Models for Retinal Disease Detection

Machine learning, with all its predictive capabilities has leant its services to the field of healthcare extensively. One of the many use cases of ML in healthcare is eye disease detection. Eye diseases have been rampant in society for a long time now. Diseases like Retinoid defects, Cataract, and Glaucoma are some of the more common issues in the sector now. Because of the significant nature of the issue, it's early detection and diagnosis is extremely important. It can otherwise lead to terminal after-effects like vision loss. Transfer Learning is one more approach which has been utilized in the sector of eye disease detection. In this paper, a comparative analysis has been carried out on a multi-image fundus dataset, which analyses and compares the performance of 5 transfer learning models on account of metrics like accuracy, precision and recall score. This paper exploits the usage of models like VGG16, MobileNetV2, DenseNet121, ResNet50, EfficientNetB0, ConvNeXtXLarge, and CNNs. The selected model will contribute to the sector and lend its predictive abilities to improve detection rate for optical diseases.