• Build a dataset of fundus images using open data (kaggle and other sources)
• Develop and train a neural network to predict stage of the DR from zero (healthy eye) to the 4th (almost lost eye).
• Package the neural network into the docker container with json api.
• Develop a system to clean, fix mistakes and to append new data to the datasets faster.
• Develop a model to detect low-quality images before grading the DR stages.
Example image from kaggle dataset
A solution to do diabetic retinopathy diagnostic in photos of eyes fundus using artificial neural networks. This service helps ophthalmologists to monitor more people who should be diagnosed to mitigate blindness.
It will significantly reduce the ophthalmologist's time spent on image grading.