A new deep learning model from Beijing Jiaotong University, known as the Local-Global Scale Fusion Network (LGSF-Net), enhances the diagnosis of eye diseases such as cataract, diabetic retinopathy, and glaucoma. This hybrid AI model merges convolutional neural networks and transformer architectures to effectively analyze fundus images, achieving 96% classification accuracy with minimal parameters. Its lightweight design allows for practical application in resource-limited clinical settings, potentially revolutionizing automated ophthalmic diagnosis and management.
1. LGSF-Net developed by Beijing Jiaotong University enhances eye disease diagnostics. 2. Combines CNNs and transformers for fundus image analysis. 3. Achieved 96% classification accuracy with 18.7K parameters. 4. Superior to ResNet50 and ViT in identifying diabetic retinopathy. 5. Lightweight design suited for resource-limited settings. 6. Future research may test on larger, imbalanced datasets. 7. Potential to improve early detection of vision-threatening diseases.
Listen Tab content
