Speaker: Joe Mellor, University of Edinburgh Usher
Institute
Abstract:
In Scotland, all people with diabetes aged 12 and upwards
are offered regular screening in which a single-field image is taken from each
eye in the diabetes eye screening (DES) programme.
As part of the "Using Deep Learning on Retinal
Images to Predict Complications and Therapeutic responses in Type 1
Diabetes" project, fundus images acquired from DES were linked to medical
records using the National Diabetes Research Platform.
I will talk about some of the prediction tasks that we
tackled in this project, how we applied deep learning models to solve them, and
how we evaluated the models for their intended use cases.
Speaker bio:
Joseph Mellor is a
Research Associate at the Usher Institute and a member of the Diabetes Medical
Informatics and Epidemiology Group led by Prof. Helen Colhoun and Prof. Paul
McKeigue, as well as a member of the Bayesian and Neural Systems Group founded
by Prof. Amos Storkey. He was AI lead of the JDRF-funded project "Using
Deep Learning on Retinal Images to Predict Complications and Therapeutic
responses in Type 1 Diabetes.”