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

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Ophthalmology

Background:

• Accurate diagnosis of age-related macular degeneration (AMD) and diabetic macular edema (DME) is crucial.
• Current diagnostic methods can be time-consuming.
• Developing AI for rapid and precise diagnosis using medical images is an active research area.

Purpose of the Study:

• To develop and evaluate artificial intelligence (AI) methods for classifying optical coherence tomography (OCT) images.
• To accurately diagnose key pathologies of AMD and DME.
• To leverage transfer learning with convolutional neural networks (CNNs) for improved diagnostic performance.

Main Methods:

• Proposed a CNN with transfer learning capability for OCT image classification.
• Utilized pre-trained CNN models as a starting point for new models.
• Selected appropriate algorithm hyperparameters (optimizer, learning rate, mini-batch size) to optimize learning speed and quality.
• Experimented with various CNN architectures including Alexnet, Googlenet, VGG, and Resnet.

Main Results:

• Transfer learning enabled CNN models to successfully classify OCT images of AMD and DME.
• Tested models demonstrated effective classification capabilities.
• Performance varied across different CNN architectures.

Conclusions:

• VGG19, Resnet101, and Resnet50 models, after transfer learning and with optimized hyperparameters, showed excellent capability in classifying OCT images of AMD and DME.
• AI-powered image analysis holds significant promise for diagnosing macular diseases.
• Further research can refine these models for clinical application.