1. Neural Networks to Classify Diabetic Retinopathy
We used Residual Networks coupled with the cyclic pooling technique for data refinement to train on 35,126 retinal images. ... These retinal screening techniques are crucial for early detection and treatment of the disease, but all require a trained optometrist as well as thousands of dollars of equipment that often are not available in developing countries and rural areas. ... We hope to classify normal eyes from any level of DR, which is a much harder problem due to the subtle differences between normal eyes and those afflicted with mild DR. ... To diagnose the retinal images, we developed...
- Word Count: 874
- Approx Pages: 3
- Grade Level: Graduate