AI can find answers to childhood disease, with the right data and researchers.
BIG data is playing an increasingly important role in the microscopic world of disease research, with developments in artificial intelligence and machine learn- ing driving significant advances in early diagnosis capabilities.
For researchers at the Nedlands-based Telethon Kids Institute, AI and machine learning are key tools in their work, providing the capability to crunch large quantities of data in a short timeframe.
Such is the efficacy of AI doing some of the heavy lifting in terms of data collation, and the analysis of machine learning (which builds on data and learns to identify patterns), that researchers at the institute can take their projects down new avenues of enquiry.
Timo Lassmann, a leading computational biologist at the institute, has been using these technologies to better diagnose the causes of rare diseases by ‘learning’ the relationship between the observable disease symptoms and the genes turned on and off in particular tissues.
Put simply, Associate Professor Lassmann and his team are working directly to diagnose incredibly rare diseases and to improve the wellbeing of children suffering from cancers.
Telethon Kids Institute director of research services and innovation, Paul Watt, said AI extracted meaning from data and sent research in new directions, with direct benefits to projects currently undertaken at the institute.
“This approach avoids the potential for bias in setting up research questions, which are limited by current theories that may be wrong or not complete,” Professor Watt told Business News, pointing to the emergence of ‘data-driven discovery’.
“There will also be more emphasis on multi-layered machine learning, deep learning, together with synthetic data, which is a tool to gain powerful medical insights from transformed clinical data, while respecting patient privacy by eliminating the potential for identification.”
A practical example of the technology making a real-world difference is seen in the case of a project under way at the Walyan Respiratory Research Centre, based at Telethon Kids Institute.
Centre researchers have discovered an algorithm that could predict whether structural lung damage might develop over time in kids with cystic fibrosis.
“This Telethon Kids innovation was patented, and the approach partnered with European [Thirona] and Western Australian [Resonance Health] machine-learning diagnostics companies, in order to help develop scalable automated diagnostic software to better help manage treatment in kids and young adults living with this disease,” Professor Watt said.
A deeper understanding of how the barriers to infection in the lung are maintained is also a focus at the institute, overseen by Tom Iosifidis and biostatistician Yuliya Karapievitch, with graduate researcher Alphons Gwatimba.
“Most surfaces of the human body, including the lungs, are lined with epithelial cells,” Professor Watt said.
“Epithelial cells form a layer that operates as the first line of defence against the external environment.
“Viruses are parasitic lifeforms that can attach to tight junctions and compromise their structure and function, leading to the infection of the host cell.
“The goal of this challenge is to analyse images of airway epithelial cells to identify cells that have had their surrounding tight junctions compromised.
“The technology used involves deep-learning algorithms for image recognition [scans], image segmentation and pattern recognition.”
Powerful data sets are used to train machine-learning algorithms to recognise signs of disease and congenital disorders.
For example, the application of machine learning analysis to the way babies move in their first few months of life can reliably predict cerebral palsy in at-risk children.
Before the uptake of AI, these movements could only be assessed by a handful of specially trained professionals.
Telethon Kids Institute research director Cath Elliott is leading a national collaboration of researchers at The ORIGINS Project in a study called Early Moves.
The study is using specialist AI expertise from Deakin University’s A2I2 Institute (including machine learning) to develop a model that will allow a computer to conduct these assessments using videos taken by parents at home on their smartphones.
Hundreds of families have already signed up to share videos of their babies for the study.
“Machine learning in health care requires high-quality data sets,” Professor Watt said.
“But it also requires data specialists. “A key challenge we face in Western Australia’s not-for-profit research institutes is access to data scientists with sufficient experience in the big data and machine-learning fields, given the competition from other industries such as mining and oil and gas.
“We’ve been fortunate to recruit some world-class scientists with this experience, but it is important that we build capacity in this area. It’s why we’re collaborating with universities and other centres of excellence with the help of powerful networks such as the WA Data Science Innovation Hub.”
Professor Watt said the suggestion that big data meant big risk, particularly around data security, was a misconception, since these risks could now be addressed.
“Many people may not be aware that clinical data can be effectively deidentified and pooled to create transformed meta-data sets,” he said.
“These transformed data sets can be used for machine learning to gain key health insights, while effectively eliminating the potential for breaching of privacy of personal medical data at the individual level, which can remain confidential within the health system.
“This approach to privacy has en- abled countries like Israel to accelerate their approach to innovation in the digital health arena.
“Most recently, these secure and comprehensive electronic medical record systems also allowed some of the best real-time reporting of the safety of the Pfizer COVID vaccine as it was rolled out in Israel.”