Improved eyecare using GPUs
Posted: 12 Nov 2015 | 13:49
A recent MSc project at EPCC has paved the way for improved diagnosis of eye-related conditions.
The quality of the image analysis was originally restricted by the short timeframe available in a clinical setting. Stefan showed in his class-medal winning dissertation "Accelerated Retinal Image Registration” that the computational performance can be improved many times over through tailoring of the software for exploitation of modern highly-parallel NVIDIA Graphics Processing Units (GPUs), combined with algorithmic optimisations.
Dr Jano van Hemert from Optos, who co-supervised the project along with myself, stated: “This opens up a range of exciting options for Optos which will potentially lead to significant healthcare advantages for patients."
Stefan has now started a PhD at Karlsruhe Institute of Technology: another industrial collaboration, this time with Audi. He will continue to work on software development for parallel hardware such as GPUs, with the new aim of analysing images from on-board cameras to enable the next generation of automated cars.
After completing the taught courses, students work on a three-month individual project leading to a dissertation. Dissertation projects may be either research-based or industry-based with an external organisation, with opportunities for placements in local companies.
Find out more
Image: John Liu, Flickr