CNN Architecture for Disease Identification from Retinal Images

The rise of artificial intelligence is changing the shape of many different fields, and ophthalmic medicine is no exception. Convolutional neural networks (CNNs) have achieved remarkable results in the assisted diagnosis of ophthalmic diseases. However, they face challenges such as limited labeling and small dataset sizes. In recent years, self-supervised learning methods based on the vision Transformer have been introduced to help solve some of the problems plaguing the medical imaging space. However, the self-attention mechanism of these methods suffers from high computational complexity, and the dependence on large-scale datasets limits their application in practice.

 

Researchers at Arizona State University in conjunction with a collaborator at the Mayo Clinic, have developed a method based on a CNN architecture for disease identification from fundus images. This method efficiently utilizes a large amount of unlabeled data, which reduces the need for expensively labeled data and lowers computational costs. By pre-training the method with self-supervised learning on a dataset of more than 170,000 retinal fundus images, the model has learned the representational features of retinal fundus images. The performance of the pre-trained model was further validated by applying it to the classification tasks of AD, PD and other retina-associated diseases.

 

This method not only demonstrates how effective self-supervised pre-training can be in a model’s understanding medical images, but also provides the framework for applying deep learning techniques to other medical image analyses.

 

Potential Applications

Benefits and Advantages

For more information about this opportunity, please see

 

For more information about the inventor(s) and their research, please see