���˴�Ƭ

Story written by Dr Sarah Keenihan, AIML

It seems hard to imagine that artificial intelligence (AI) could help patients with rheumatoid arthritis or lung disease. 

However, a type of AI called machine learning offers doctors new ways to review patient scans and other types of clinical information. With careful development, this could open up options for faster, more accurate diagnosis and better treatments for painful, inflamed joints or earlier detection of debilitating respiratory conditions. 

Two new PhD students have recently joined the Australian Institute for Machine Learning (AIML) at the ���˴�Ƭ to advance these exciting possibilities. 

 and  are being supported in their studies through scholarships from GlaxoSmithKline (). 

GSK has put a high priority on AI, machine learning and other advanced technology as part of its .

“Funding these scholarships with AIML at the ���˴�Ƭ is the latest example of how we foster talent at the intersection of medicine, AI and machine learning at GSK in support of our work to help double development success rates of new medicines and vaccines,“ says Kim Branson, Senior Vice President, Global Head of Artificial Intelligence and Machine Learning at GSK. 

Measuring damage in joints

Alix Bird graduated with an MBBS from the ���˴�Ƭ in 2018. For their PhD project, they will be working on machine learning to target rheumatoid arthritis, an autoimmune disease that creates painful swelling in joints. 

In particular, Alix aims to tackle the problem of how to grade damage and immobility in arthritis-affected hand joints, such as knuckles. The process involves judging the extent of bone erosion and narrowing of the spaces between bones. 

“In rheumatoid arthritis, doctors use a scoring system with hand X-rays to assess how damaged joints are, and measure whether drugs are having any impact,” Alix says. 

“This is a super arduous process – it takes 25 minutes for each review, and you need two doctors per X-ray, so 50 minutes at least to assess each scan accurately.” 

Another problem is that grading of X-rays can vary between different doctors. 

“So if we could train a machine learning system to be able to accurately review and grade X-rays according to extent of arthritic damage, this will be better for patients, faster, and more cost-effective for the healthcare system,” explains Alix. 

“It will also help with assessing whether arthritis treatments are having a positive impact.” 

Machine learning becomes accurate through “experience” – in this case, computer systems would be trained to detect arthritis by reviewing accurately-labelled data. 

Alix’s PhD supervisors are AIML’s , GSK’s Dr Ray Shonket and , Rheumatology Unit Director at the Royal Adelaide Hospital. 

Picking up lung damage, early 

Luke Smith’s PhD project will focus on applying machine learning for more accurate and earlier detection of COPD (chronic obstructive pulmonary disease), along with better prognosis of clinical outcomes. 

Patients with COPD experience progressive loss in the ability to breathe, most often linked with conditions such as emphysema and chronic bronchitis. 

“But it’s sometimes difficult to diagnose the different kinds of COPD, and match patients with the best treatment option,” Luke explains. 

“So in my project I’m going to apply machine learning to review images of patient lungs to see we can improve detection – particularly in early cases when the physical changes can be subtle and difficult to identify.” 

“We’d also like to be able to use machine learning to better predict how the disease will develop over time,” Luke says. 

Data from known cases of COPD will be used to train machine learning systems. 

Luke ended up as a PhD student in medical machine learning somewhat by accident. 

“I did electrical engineering and physics as an undergraduate student, and after working for six months I was approached by my supervisors to do postgraduate studies,” Luke says. 

“I see it as a good example of the value of working hard in a field you enjoy, so that when an opportunity comes along you can seize it.” 

Luke’s PhD supervisors are AIML’s , GSK’s Dr Ray Shonket,  (an expert in Respiratory and Sleep Medicine at Flinders University) and the Central Adelaide Local Health Network’s . 

Applying AI to help patients

Professor Lyle Palmer, the leader of AIML’s Machine Learning in Medicine Team, is the principal supervisor of both Alix and Luke’s PhD projects. 

“We are excited to have such fantastic students joining our AIML team – we look forward to watching them and their projects develop over time,” Palmer says.

“We’re thrilled to be working with GSK and hope this is the beginning of a significant and long-term collaboration with a company leading the industry in applying AI and machine learning to help patients.” 

Dr Lauren Oakden-Rayner is involved in both PhD projects as Director of Medical Imaging Research at the Royal Adelaide Hospital, and a researcher in medical AI at AIML. 

“It’s great to be collaborating with GSK to develop new capabilities, especially on such exciting projects which offer the potential to provide better targeted treatment for patients and to accelerate clinical research into these serious diseases,” Oakden-Rayner says. 

If you’re interested in exploring postgraduate opportunities at AIML, find out more on our programs page.