Undiagnosed and unoperated cataracts are the leading cause of blindness in the world. This problem is especially severe in developing countries such as China and India. One of the mains reasons is poor access to diagnostic care in low resource regions.
What it does
CatSpotter provides a cheap, easy and integrated way to identify cataracts and refer patients. It facilitates telemedicine by automated-diagnosis.
How we built it
We used Sketch to build up the UI prototype of the smartphone app, used pytorch to build a state-of-art deep learning model of computer vision for cataract detection, used sketchup to build the 3D demo, and use python to crawl cataract images from the open website. Clinicians to classify the images.
Challenges we ran into
How to take photos of cataracts.
How to obtain the training data to improve prediction.
Accomplishments that we're proud of
Automated detection algorithm
What we learned
What's next for CatSpotter
We hope to provide free subscription for everyone interested to raise awareness about cataracts. Moving forward we would like to build a network of patients, regional (rural) doctors, specialists and health care provider institutes. In the long term we hope to perfect our algorithm to deliver a more accurate cataract detection system that can be widely used in low resource settings.