Empirical Evaluation of Retinal Disease Identification Based on Deep Learning Assisted Image Classification Strategy

This study introduces a Hybrid Neural Network (HNN) model for the early detection of retinal diseases like Diabetic Retinopathy, Cataract, and Glaucoma. The model integrates EfficientNet as a feature extractor with an SVM classifier, enhanced by attention mechanisms and multi-scale feature fusion. Tested on the “Eye Diseases Classification” dataset, it achieved 98.90% accuracy, surpassing other models such as ResNet-50 (86.75%), DenseNet-121 (88.42%), and Attention ResNet (95.40<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>). The model also demonstrates excellent performance with Precision (98.50<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), Recall (98.70<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), and Fl-Score (98.60<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>), along with a fast inference time of 12.3ms per image. This approach provides an efficient, high-accuracy solution for retinal disease classification, ideal for real-time clinical use.