Research IT Services


Mighty Legion joins the battle against Parkinson’s disease

1 October 2019

High performance computing is helping researchers to solve the mystery of how deep brain stimulation is able to relieve the symptoms of Parkinson's Disease, by enabling them to fit statistical models to huge volumes of data collected from functional MRI experiments.

Magnetic Resonance Imaging

It’s 50 years since man made it to the Moon aided by less computing power than drives a modern iPhone. But in computing – as in so much else – a lot has changed in the past half century. Facilities of once unimaginable scale and sophistication are now routinely unleashed to crack intractable problems and pursue new solutions in every sphere of life. But as capabilities evolve, there’s also a critical need to ensure they can be harnessed as widely and easily as possible so their benefits can reach farther, faster than ever before.

Day in, day out, Research IT Services (RITS) meet this need by making outstanding high-performance computing (HPC) platforms simple to access and use for UCL researchers focused on all kinds of world-shaping work – researchers like Dr Joshua Kahan and Professor Tom Foltynie, who set their sights on helping the millions around the globe with lives affected by Parkinson’s disease.

Advances in Mind

In the quarter century between 1990 and 2015, the number of people with Parkinson’s doubled worldwide. Here in the UK, one in every 350 adults now suffers from a condition caused by the brain’s failure to make enough dopamine, a chemical enabling signals to travel from one nerve cell to another. Not only will the pressures of an ageing population result in the number of Parkinson’s sufferers rising inexorably; a growing slice of healthcare resources will also have to be deployed to treat the condition.

Naturally, the quest for better treatments is relentless. One technique already used to relieve the symptoms of Parkinson’s disease is deep brain stimulation (DBS), where electrical impulses are fired at tiny target areas in the brain. Although it works, why and how it works – and specifically what happens at the level of the neural circuits that underpin the nervous system – isn’t really understood. Yet plugging that knowledge gap is key to enhancing DBS’s effectiveness and perhaps paving the way to new types of treatment.

This was the challenge taken on by Professor Foltynie and colleagues at the UCL Institute of Neurology, working with the National Hospital for Neurology and Neurosurgery and The Wellcome Trust Centre for Neuroimaging. Their aim was to use a tried-and-tested medical imaging technique – functional magnetic resonance imaging (fMRI) – to scan patients’ brains as they were undergoing DBS and performing a range of tasks, and then feed this brain activity data into a series of computer models. The models could then explore what must be happening at neural-circuit level in order to produce the patterns of data collected.

But there was a problem – the colossal volume of data generated. Every scan consisted of thousands of ‘voxels’ (volume pixels), with a 3D movie made for every one of these. Fitting the computer models to this data mountain looked like being a long, laborious, uphill climb.

“Legion allowed us to fit the models almost overnight. It was fast, efficient and painless!” Dr Joshua Kahan

Fresh Thinking

Joshua Kahan takes up the story:

“Using my computer at UCL, it would have taken two to three weeks to fit the models and I wouldn’t have been able to use my machine to work on other things. A friend from The Wellcome Centre suggested using the RITS Legion service. By running jobs simultaneously using cluster computing, it could potentially save us huge amounts of time. So I decided to look into it further.”

Legion is one of three centrally funded HPC platforms made available by RITS for compute-intensive research. But the crunch question for Dr Kahan, who had no previous experience of using a computing cluster like Legion, was this – how easy would it be to use? 

  • Simple Sign-up: “I went onto the RITS website and followed the straightforward directions to set myself up with the Legion service.”
  • Pain-free Preparation: “To fit the series of models to the fMRI data, I wanted to use a popular software package called SPM, written in MATLAB programming language. MATLAB was already installed on Legion so all I had to do was install my version of SPM as well.” 
  • Easy Execution: “The examples listed on the Legion website were extremely useful in helping me create the necessary ‘job submit’ files. I prepared the files for model-fitting locally and uploaded them using a standard file transfer protocol (FTP) client. Once all that had been done, I could submit a job and off it would go!”

Using Legion’s ‘job array’ feature, Joshua submitted batches of up to 150 jobs at a time, knowing that the platform could easily take these in its stride. Email requests for advice or additional support received prompt, clear responses from RITS’s Research Computing Platform Service (RCPS) team. Nor did Joshua have to wait long to see the benefits of unleashing Legion. “It allowed us to fit the models almost overnight. It was fast, efficient and painless!”

Full Speed Ahead

Joshua first used Legion seven years ago. Since then, the platform has been at the heart of the leading-edge work that he and his colleagues have pursued with the goal of generating new insights into DBS – insights that could bring fresh hope to Parkinson’s sufferers. 

“The great thing is, once everything’s been set up in accordance with your needs, it’s easy to reconfigure Legion and use it for any number of further experiments,” he comments. “For example, in our most recently published work, we fitted roughly 600 models to our data, each comprising a final file size of around 10MB. Legion slashed the time it took to do this.”

 What, then, of the future?

“We’ve now collected more data that looks at the longitudinal effects of DBS and can help us refine our proposed model of the brain’s motor system. We’ll certainly be using Legion again to cut the amount of time needed to analyse this data. Quite simply, quicker computing means faster progress with our research and improved prospects that it could help trigger tangible advances in treating Parkinson’s disease.”


Unit of Functional Neurosurgery at UCL

Key Papers: