Source: Radio New Zealand
AFP
Artificial intelligence could soon be used to screen for breast cancer in the public system.
Experts say using AI as a second set of eyes for mammograms could make better use of our radiologists – as long as its trained on the right data.
Typically, the tool works by assessing a mammogram for tumours, using past examples of positive and negative scans to make a decision on whether a tumour is present, not present, or unclear, meaning the patient should get a second scan.
A Swedish study of 100,000 women found a higher rate of early detection, and a UK study of ten thousand women found it increased detection generally by more than 10 percent.
Just last month, the New Zealand government put out a request for information on the topic, assessing the market. The tender closed on 4 March.
Breast cancer is the most common cancer affecting women in New Zealand, killing more than 650 each year, and more than 3700 diagnosed, according to the Breast Cancer Foundation.
Dr Mehdi Shahbazpour from the Breast Cancer Foundation explained in New Zealand, two radiologists looked at each scan.
If AI could replace one of those, it could alleviate some of the pressure on a stretched workforce.
“Our shortages of radiologists are real, they’re impacting screening, they’re impacting treatment and diagnosis. We need to implement a solution fast – next one to two years is our preferred timeline.”
Recalls – where a second screening was required as the first was unclear – added additional load, he said, and overseas AI had been shown to reduce recall rates, and improve accuracy.
Studies had also shown it was better at catching “interval cancers” – that is, those which appeared between scans, usually because of symptoms like lumps, and usually signalled a more aggressive form of cancer.
The timeline for introducing a new system was variable, with a lot of unknowns.
“We want New Zealand to be a fast follower,” Shahbazpour said. “Don’t try to reinvent the wheel, go with trusted models that have been implemented elsewhere.”
Dr Karaitiana Taiuru, an expert in AI and data governance, said if we were to adopt a tool developed overseas, it was important to train it on local data.
Research has shown characteristics like breast tissue density vary between ethnicities. Māori and Pacific women have typically more dense breast tissue than, for example, an American woman’s.
“The risks there are that the AI could just not see a cancer,” Taiuru said. “Or it might see something that is not a cancer, but say it is a cancer.”
That local dataset was ready-made, with the country’s history of mammograms already on-file.
Taiuru said as long as it was fed the right data, AI could actually remove bias.
“The AI doesn’t care if you’re Māori, or if you’re Asian, or if you’re European.”
He said people should be able to opt out of having their data fed to AI if they chose to, including for cultural reasons.
While the government scouts for information, private companies are a step ahead, with some already working to introduce AI tools.
Shayne Hunter, chief digital officer for RHCNZ, the parent company of Auckland, Pacific and Bay Radiologies, said for them, its arrival was imminent.
He said they planned to use it for private diagnostic and screening work, catering for women who fell outside the criteria of the public system and chose to pay.
“We have a capacity issue in the health system in that we don’t have enough breast radiologist, and so part of this is about creating capacity to meet the demand,” Hunter said.
Ongoing clinical governance would be key, he said, to make sure clinicians didn’t come to rely on AI, or become disinclined to challenge its results.
“The key thing about AI is there is still a human in the loop,” he said. “We still have a human that will absolutely check every image.”
Nicholas Knowlton, a senior research fellow at the University of Auckland, said it was clear the tool was effective.
“Should the public be paying for these things, is the question, and how much should we be paying? If we take a Pharmac-type of approach and demand high value for money, yeah, why not?”
Were there any risks? Not with the tech itself, anyway.
“The model is decision support,” he said. “It’s like spellcheck on your document when you’re writing. It says ‘This word’s spelled wrong,’ you look at and go, ‘No, that’s right. This is what I meant to say.’ It’s helping, it’s not making clinical decisions for us.”
“I think the risk is dragging our feet and missing cancers that can be detected earlier.”
Sign up for Ngā Pitopito Kōrero, a daily newsletter curated by our editors and delivered straight to your inbox every weekday.
– Published by EveningReport.nz and AsiaPacificReport.nz, see: MIL OSI in partnership with Radio New Zealand
LiveNews: https://livenews.co.nz/2026/03/25/artificial-intelligence-could-soon-be-used-to-help-screen-for-breast-cancer/