UNSW researchers develop AI-driven breath test for early detection of deadly silicosis

A new method for diagnosing silicosis simply by analysing a person’s breath is being developed by physicians and scientists from UNSW Sydney, paving the way to potential early detection of the deadly disease.

According to a study published in the Journal of Breath Research by Professor William Alexander Donald and Conjoint Professor Deborah Yates, the revolutionary test could transform the way silicosis is diagnosed and provide patients with signs of trouble ahead of irreversible lung damage appearing.

The test combines mass spectrometry and artificial intelligence to rapidly detect silicosis from breath samples, providing a non-invasive diagnostic tool for at-risk workers that delivers test results in minutes.

This compares with traditional methods of detection such as X-rays and CT scans that usually find silicosis at later stages of the disease.

“Our study shows that the AI-driven model accurately distinguished silicosis patients from healthy individuals based on their breath profiles, providing a reliable tool for early detection,” says Donald, the lead researcher from UNSW’s School of Chemistry.

“This suggests that breath testing could be a practical tool for large-scale worker screening and early intervention.”

Silicosis, a lung disease caused by inhaling tiny crystalline particles of silicon dioxide, has become a major occupational health concern in Australia, with cases now extending beyond engineered stone workers to those in tunnelling and construction.

While the federal government has banned engineered stone, new cases have emerged from other high-risk industries which the UNSW researchers say creates an urgent need for better diagnostic tools.

In their study, the researchers analysed breath samples from 31 silicosis patients and 60 healthy controls, confirming the test’s ability to differentiate between affected and unaffected individuals with high accuracy.

To be tested for silicosis using the new technology, participants breathe into a bag. The breath content then gets pushed to a mass spectrometer which detects all the different molecules present.

“In human breath, there are thousands of organic molecules that you breathe out,” says Donald.

“Our instrument can make a profile of someone’s breath, and then we feed that into an artificial intelligence algorithm that’s really good at finding patterns.

“In this case, it’s looking for patterns in the organic compounds that are present in the breath of people in the early stages of silicosis. And we’re getting very high accuracies, like over 90 per cent accuracy, for just such a simple, non-invasive breath test.”

The entire breath sampling and analysis process takes less than five minutes per patient, which the researchers say makes it suitable for routine screening of at-risk workers.

"By combining advanced chemical analysis techniques with AI, it means we can now detect silicosis in minutes rather than waiting for irreversible lung damage to appear on scans," says Donald.

The next step for the researchers is to validate the test among larger cohorts before it can be implemented as a routine screening tool.

The instrument is a compact benchtop system, occupying less than a cubic metre, making it feasible for installation in clinical settings for point-of-care testing. While breath samples are currently collected in clinics and transported to the lab for analysis, future implementation could enable direct, on-site testing.

“We’ve now installed this testing at a second site to test its effectiveness with hundreds of at-risk workers, including coal miners, to further validate its effectiveness,” says Donald.

The ultimate aim of the researchers is to refine the system to a point where it is able to differentiate silicosis in patients from other lung diseases.

Yates, who as a clinician has seen many cases of silicosis, describes the disease as both incurable and totally preventable as its cause is excess silica dust exposure.

“There are three stages of silicosis,” says Yates.

“It can be difficult to diagnose such patients, especially in the early stages of disease. Sometimes a biopsy is needed, which is invasive and expensive.  

“So, it is crucial to detect affected workers early and remove them from further silica exposure in order to stop the progression of their disease.”

Silica exposure can lead to several other diseases including lung fibrosis, chronic obstructive pulmonary disease, lung cancer, and autoimmune diseases such as rheumatoid arthritis and scleroderma.

“Breath testing is a new and simple technique for patients and doctors but has been valuable in other lung diseases such as asthma,” says Yates.

“It could provide the non-invasive method needed to monitor such workers for development of silicosis. Although this research is yet in its early stages and will need further development, it provides a breath of hope for the future.”