Diabetic retinopathy is one of the complications of diabetes leading to blindness. Early detection through periodic screening of retinal images can help reduce the long-term effects and blindness by as much as 90%. However, its resource demands are high, requiring eye doctors and specialist equipment.
The Sydney Informatics Hub along with Assoc. Prof. Ravi Seethamraju from the University of Sydney Business School, and with Prof. Krishna Sundar and Dr. Shashank Garg, both from the Digital Innovation Lab at the Indian Institute of Management Bangalore, want to design a low-cost solution to this problem. Artificial intelligence technologies, in particular deep learning, could offer a way to implement real-time screening at relatively low cost and take retinal screening to the diabetic patient in primary care. The researchers needed an algorithm to help classify the disease based on photographs of patient eyes, so have used the a coding challenge as a platform to crowd source the collaboration to help work toward a solution. 24 teams submitted algorithms to the hosted Kaggle competition.
The most common approaches used in the competition started with pre-trained models (e.g. EfficientNet, DenseNet, InceptionV3) with additional neural network layers optimised for the competition dataset. Competitors were open to use any tools and methods they wanted, as it turned out every entrant used the Python programming language. In more detail, there was a fairly even split between using the PyTorch and Keras/TensorFlow deep learning libraries.
The prize recipients were acknowledged at the Innovation Week awards night this week along with other innovators and entrepreneurs.
The prize winners for the Diabetic Retinopathy Innovation Coding Challenge were:
1st (“DR Prize”) - $7500: Simon Cai - A data scientist working with the Children’s Medical Research Institute.
2nd (“Kaggle Prize”) - $2000: Mike Li - A research associate at Sydney University’s Centre for Complex Systems.
3rd (“Artemis Prize”) - $500: Patrick Hao and Andrew Lee - Undergraduate students at Sydney University studying computer science and financial mathematics.
The Sydney Informatics Hub thanks all our participants, as well as our partners in the Sydney Innovation Hub for helping make the Coding Challenge such a great success. We look forward to next year’s challenge!
Have a problem you think would make for a good challenge? Drop us a project request or get in touch firstname.lastname@example.org.