High Performance Computing for Early Cancer Detection
The Department of Psychology and the Research Computing team at the University of York leverage High Performance Computing (HPC) and Machine Learning (ML) to seek out ‘ nonlocalizable global markers of cancer’ through the rigorous testing and refinement of neural network models on public cloud.
Excerpt from the Alces Flight client success story series.
The Department of Psychology and the Research Computing team at the University of York have taken on the challenge to understand how clinician’s minds often notice the “nonlocalizable global markers” of cancer in in x-ray imaging or mammography and if, through the use of neural networking models, they can replicate this ability in order to increase capability for early cancer detection. To achieve this goal the team leverages HPC and machine learning via extensive testing and refinement of neural networking models in cancer diagnosis.
“Our Research Computing team at the University of York strives to make HPC available to all research subjects. Thanks to our work with the Alces Flight team we are continuing to expand these services into public cloud, utilising the Flight CCV toolset to allow researchers to seamlessly engage with supercomputing resources at the right time, scale, and cost that suits their project needs.”
— Dr. Emma Barnes, University of York
“The human mind is a wonderful thing,” says Karla Evans, Associate Professor, at the University of York. “Our minds have this uncanny ability to see patterns in every input, be it speech or images. However, through years of seeing and comparing thousands of images and tuning the visual system to the natural occurring patterns we have this ability also to pick up important differences, including signals that something might be wrong. The question is, can we train machines to help us find those signals faster and at scale so we can do something about it? This is the premise behind our cancer research project — utilising HPC and ML to locate markers that trained clinicians notice at scale so we have the potential to treat people much sooner.”
Building this early cancer detection project meant the University of York’s HPC services needed to undergo a public cloud expansion. “We want our researchers to have access to the right resources at the right time,” explains Dr Emma Barnes, “In order to achieve their goals the team at the Department of Psychology needed to work interactively with their models at scale. We turned to Microsoft Azure to add this functionality and the team at Alces Flight to bring this capability into our already existing HPC service portfolio.”
To read the full success story (no registration required) visit this page.
To learn more about how Alces utilises an technology-backed approach to building and managing HPC clusters and environments visit us at: