AI improves medical imaging to benefit people with brain ailments

Announced this week by Assistant Minister for Science, Jobs and Innovation the Hon Zed Seselja, the funding comes via the government’s Cooperative Research Centre Project (CRC-P) Program, a competitive, merit-based program supporting industry-led, outcomes-focused partnerships between industry, researchers and the community.

The government’s investment is matched by nearly $2.8 million of cash and in-kind contributions by the University and Brain and Mind Centre’s project partners, including Sydney Neuroimaging Analysis Centre (SNAC) and the I-MED Radiology Network.

“We aim to transform the delivery of neuro-radiology services across Australia,” says the Brain and Mind Centre’s Professor of Neurology, Dr Michael Barnett, who is also a consultant neurologist at Royal Prince Alfred Hospital in Sydney.

“We plan to do this by developing novel, automated algorithms that aid in both the diagnosis and monitoring of brain diseases using magnetic resonance images and CT scans.”

It’s estimated that clinicians misinterpret up to four percent of medical images, a figure likely to be even higher in demanding subspecialties such as neuro-imaging.

“When these algorithms are built they will be deployed on an artificial intelligence (AI) platform that integrates with routine clinical radiology workflows to dramatically improve productivity, enhance reporting accuracy and rapidly identify critical imaging abnormalities,” Professor Barnett says.

The commercial application of AI in the medical imaging industry is currently in its infancy, driven by independent technology companies targeting individual patients, rather than enhancing innovation in the radiology and research-imaging industries.

These tech-companies also lack access to well-characterised clinical populations needed to drive the development of accurate algorithms. 

Sydney Neuroimaging Analysis Centre (SNAC) was established at the Brain and Mind Centre in 2012 to facilitate novel imaging biomarker research and make quantitative analysis of magnetic resonance imaging (MRI) images available to the pharmaceutical industry and researchers undertaking Phase 2-4 clinical trials.

Together with leading University of Sydney experts, SNAC will lead the project’s three-year implementation to develop what Professor Barnett describes as “an artificial intelligence platform and neuro-imaging algorithms based on deep learning ‘artificial neural networks’”.

Deep learning is a collection of machine or computer learning algorithms capable of recognising patterns in data, in this case brain images, without manual labelling or identification of their features.

The University’s project team includes top AI scientist Professor Dacheng Tao, neurologist and academic lead for the biomedical data initiative, Professor Michael Barnett, and multimodal imaging expert Professor Tom (Weidong) Cai.

Project partner, I-MED, is a national radiology provider that processes 4.2 million clinical images annually at more than 200 clinics across Australia. It will supply the bulk of the project’s de-identified imaging and reporting data to inform algorithm development and validation.

The University of Sydney’s Deputy Vice Chancellor, Research Professor Duncan Ivison said the project was a benchmark for how to improve health outcomes.

“I commend the government and project partners for funding this effort to improve diagnostic neuro-imaging for the benefit of people with degenerative brain disorders,” said Professor Ivison.

“Collaboration and multidisciplinary research hold the key to solving our biggest healthcare challenges and this project is a great example of this approach.”