Connect with us


How graphics tech could accelerate cancer treatment

Avatar photo



The technology company, NVIDIA is best known for its graphic processing units (GPU) for computer games. However, in recent years the company has turned its attention to healthcare.

Its most recent project was with The Cancer Grand Challenges’ Mutographs team; an international research group with backing from Cancer Research UK.

The research project is looking into somatic mutations; gene changes that occur in cells over an individual’s lifetime and are a primary cause of cancer. Ultraviolet light, radiation, drinking and smoking are all examples of the environmental and behavioural factors that can trigger these gene changes.

Researchers are using AI models to analyse DNA signatures and see if these significant changes in lifestyle can be identified. The end-goal is to better understand the causes of cancer and determine the best type of treatment for patients.

In a collaborative project with the Sanger Institute and Cancer Research UK, NVIDIA embarked on a four-month engineering exercise to help accelerate the research process. The organisations are collectively studying 5,000 tumours across pancreas, colorectal, kidney and two types of oesophagus cancer.

NVIDIA used its GPUs to speed up some of the more time-consuming elements of the research team’s AI framework. Health Tech World spoke to NVIDIA’s health and life science industry business development lead in EMEA, Craig Rhodes, about how its models are aiding the research.

Rhodes says: “The Mutographs team had a very large mathematical problem involving matrix multiplication, so every time they ran the application, it would take 18-30 days for the results to come back. In some instances, that’s fine but in other instances, you might want those results quicker.

“We realised there were a lot of things that we could do to help. Not only taking the code and putting that onto a GPU, but also optimising that code.”

This 30-day turn around gave the research team capacity for 12 analysis runs every year. NVIDIA’s GPU system reduced the analysis time to just seven hours, amounting to a potential 365 iterations of data per year.

Rhodes says: “We’ve always wanted science to be accurate, tested and benchmarked. All of those things are great, but that’s also made it very, very slow.

“Now we’re getting into this rapid, agile world; it’s making science become agile and allows scientific insights to come out faster.

“However, that’s going to create another bottleneck somewhere else in the process. For instance, someone who previously reviewed this data every month will now be reviewing it every day, so they will need a fundamental change to the way they do things and GPUs will be at the heart of that.”

NVIDIA branched into genomics and bioinformatics around two years ago, observing that genomics sequencing processes could be accelerated on a GPU rather than a CPU (central processing unit).

“A couple of years ago, you would sequence a whole genome in 6-8 hours,” says Rhodes, “but now we’re getting that time down to 30-40 minutes.”

At the peak of UK lockdown, Cancer Research UK estimated that 2 million people were waiting for cancer treatment. NVIDIA hopes its GPU and AI technology can help reduce this waiting list by improving pathology image analysis and speeding up genomic sequencing.

Rhodes says: “If we can help with some of the image analysis from pathologies, get that sequencing done faster, and if we can find new information and safely present it back to the clinical teams. In turn, that will help with the amount of patients waiting on cancer treatment.”

NVIDIA’s AI system is said to support pathologists by looking back through historical data that it has seen in previous images and pick out subtleties that the doctor may otherwise have missed.

Pathology images are very large and complex, yet the subtle changes in these images are usually very small.

Rhodes says: “We can now start to look at pathology images in the same way that we might look at an image on a computer game. We can start to determine what is normal and what is abnormal by teaching the system.

“Pathologists who usually see hundreds of these a day can make a clinical decision in five seconds, but what they can’t do is remember the 10,000 images that they saw in the last two years.

“The AI system can. The AI system can look at an image, trawl through all the data that it has seen before, and then start to pick up subtleties, which provides additional information for the pathologist.”

NVIDIA will soon be releasing a white paper about its collaboration with The Cancer Grand Challenges Mutographs project.

The company is also is in discussion with the Sanger Institute about collaborating on other programmes, including the Tree of Life; a project using genome sequencing to investigate the evolution of life, identify materials for new biotechnology and develop tools for biodiversity conservation.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending stories