Classification of age-related macular degeneration using convolutional-neural-network-based transfer learning

• DOI: 10.1186/s12859-021-04001-1
• Publication Date: 2021-11-08T15:05:43Z
• Authors: Yao-Mei Chen; Wei-Tai Huang; Wen-Hsien Ho; Jinn-Tsong Tsai
• Keywords: Hyperparameter; Transfer of learning

To diagnose key pathologies of age-related macular degeneration (AMD) and diabetic edema (DME) quickly accurately, researchers attempted to develop effective artificial intelligence methods by using medical images.A convolutional neural network (CNN) with transfer learning capability is proposed appropriate hyperparameters are selected for classifying optical coherence tomography (OCT) images AMD DME. perform learning, a pre-trained CNN model used as the starting point new solving related problems. The (parameters that have set values before process begins) in this study were algorithm affect speed quality. During training, different CNN-based models require (e.g., optimizer, rate, mini-batch size). Experiments showed that, after (8-layer Alexnet, 22-layer Googlenet, 16-layer VGG, 19-layer 18-layer Resnet, 50-layer 101-layer Resnet) successfully classified OCT DME.The experimental results further VGG19, Resnet101, Resnet50 had excellent performance