22 March 2019

Brave new world for cancer diagnosis

Artificial intelligence could be used to review – and possibly one day even make – cancer diagnoses if a team of WA researchers, led by PathWest anatomical pathologist Jeremy Parry, is successful in training a specialised computer to detect abnormalities in lymph nodes.

The bold new project, which featured on Channel 7 News (external site) is one of 10 local research projects awarded funding in the latest round of the Department of Health-funded Research Translation Projects (RTP).

Dr Parry and his co-researchers will take a computer capable of “deep learning” and teach it to detect changes in lymph nodes that may or may not be malignant. With later refinements, the computer could potentially discern malignant from benign changes.

The research team, which includes experts in artificial intelligence from Murdoch University’s College of Science, Health, Engineering and Education will “teach” the computer using digitised whole-slide scans of lymph node tissue collected from Western Australian patient samples.

Dr Parry hopes that through the process of deep learning – in which the computer learns to recognise patterns within data it has already analysed – the computer might eventually be able to detect nuanced early indicators of cancer.

Dr Parry’s says the aim of his project is not to replace pathologists in analysing samples, but to assist in the review and validation of their findings.

In a second part of the project, Dr Parry and his team will assess the value of using digitised whole-slide scans of tissue samples across the WA health system.

The present system for examining tissue samples involves putting them on glass slides so that they can be viewed under a microscope.

“If we need a second opinion we must physically transport the slide to wherever the person is, which could be at another hospital – or even in another state,” Dr Parry explained.

“However if we take the sample on the slide and then scan it using our digital whole-slide pathology scanner, we have access to an image that we can send anywhere in the world and which can be viewed instantaneously.”

Dr Parry says that digitisation would add an extra step to the processing of these samples, but has potential benefits which include improved flexibility for information sharing, reduced time and cost of transporting slides and improved storage and preservation of images.

The RTP program, now in its twelfth year, is designed to encourage research and the translation of research outcomes into effective healthcare policy and practice.

The RTP program highlights how research can improve patient outcomes yet at the same time enhance efficiency and cost effectiveness in the public health system.

The full list of RTP 2018 recipients are:

Coordinating Principal Investigator (CPI)

CPI Institution
Project Title
A/Professor Glenn Arendts
People dying of and with dementia: using an emergency department visit as a positive opportunity for palliative dementia care
Clin A/Professor Susan Benson
West Australian SMART Application of Blood culture Initiative (WASABI): improving the management of patients with serious infection and reducing low value care
Dr Andrew Davies
Dr Andrew Davies
Pilot study of a mobile Homeless Outreach Dual Diagnosis Service (HODDS)
Professor Graham Hall
Telethon Kids
Success in the operating theatre: multidisciplinary pre-operative briefings for efficiency, patient safety and staff engagement
A/Professor Tim Inglis
Sepsis FASTtrack; streamlining the diagnosis and management of sepsis  
Professor Alexander John
The clinical and economic benefits of early use of clozapine in first-episode schizophrenia
Dr Andrew Martin
Telethon Kids on behalf of CCHR UWA
FeBRILe3 - Fever, Blood cultures and Readiness for discharge in Infants Less than 3 months old
Dr George O'Neil
Australian Medical Procedures Research Foundation
Reducing hospital admission costs through the development of antibiotic infusion systems for the home and hospital
Dr Jeremy Parry
Application of deep-learning artificial intelligence algorithms to help diagnose