Diabetic retinopathy is a condition that affects people with diabetes and can lead to vision loss or blindness. There are four stages to this condition.
For my project, I would like to train neural networks (possibly with a CNN) to detect and classify diabetic retinopathy using photographs of retinas with and without hemorrhages. As an extension, I would also like to train other image processing neural networks on this dataset and compare their accuracies. Then, I would like to possibly predict the stage and progression of retinopathy, and assess the accuracy of the model in classifying retinopathy and predicting the stage/progression of the disease. The extension of this project may encounter some difficulties surrounding the existence and public availability of such a dataset.
- Find/clean a dataset of images.
- Train a model to classify images of diabetic retinopathy.
- Test the model with varying levels of noise.
-
- Retrain the model to detect and predict stages (dependent on whether a big enough dataset exists)
- Discuss
Discuss the application of AI to Medicine and research the downsides of ML classification versus specialists.