Differentially private transfer learning for medical image generation to support open research

Our proposal will focus on data synthesis for retina-based disease diagnoses, such as Glaucoma detection. Glaucoma is one of the leading causes of irreversible but preventable blindness in working-age populations. In 2013, 64.3 million people aged 40–80 years were estimated to suffer from glaucoma, while this number is expected to increase to 76 million by 2020 and 111.8 million by 2040. This project aims to generate privacy-preserving synthetic ophthalmological medical image data, i.e., fundus images, through the use of public data, differential privacy and machine learning-based generative models.

Faculty Supervisor:

Xiaoxiao Li;Mijung Park

Student:

Partner:

Hoffmann-La Roche Limited

Discipline:

Computer science

Sector:

Manufacturing; Professional, scientific and technical services; Wholesale trade

University:

The University of British Columbia

Program: