Tech will soon reduce deaths from breast cancer
November 5, 2019
More than 600 women die from breast cancer each year in New Zealand. With an AI augmented screening programme, many of
these deaths might be preventable according to the AI Forum of New Zealand’s latest research report.
The programme could pay for itself, given that the cost of breast cancer in New Zealand is more than $126 million per
year, with $45,000 spent per diagnosed case.
By 2021, the cost of cancer care is predicted to rise by 20 percent. Once proven, such an approach to screening could
then be generalized to other cancers, the AI Forum’s executive director Emma Naji says.
AI can be used to predict breast cancer, potentially five years in advance, diagnose it from medical images with
comparable ability to radiologists, and decrease the error rate of pathologists by 85 percent, Naji says.
“Early detection of breast cancer dramatically improves the outcome for the patient. In New Zealand, improved screening
for breast cancer over the past 20 years has reduced deaths from the disease by 27 percent.
“If AI can further augment and improve breast cancer screening by even 10 percent this would result in saving 60 women’s
lives each year and reduce the cost to the taxpayer by $12.6 million a year.”
The need for early diagnosis underpins breast cancer screening. AI driven predictive analytics could mean that screening
in New Zealand is offered according to risk and sparing those at very low risk from false positive results and
The report says 12 percent of women will develop breast cancer in their lifetime, but this does not mean that each woman
has a 12 percent risk.
New Zealand has an opportunity to link and leverage our comprehensive health datasets and use this big data to train
intelligent predictive models for breast cancer, Naji says.
“Combining local and national datasets such as primary health records, electronic health records, the cancer registry,
mortality collections and adding new datasets over time such as a biobank, could yield a national-level predictive model
for breast cancer risk, able to provide far more granular and relevant predictions for high disease risk.
“This model could be used by GPs and DHB oncology services to enable intensive monitoring and prevention or pre-emptive
treatment of women at high risk for breast cancer.
“It could also reduce the screening frequency of those with vanishingly low probability of cancer, thereby minimising
discomfort, anxiety around false positive results and freeing up resources.
“Such a solution relies on AI foundations such as skilled AI talent with domain knowledge and expertise building
predictive models in health and medicine.
“It also leans on applying AI to disease predictions and ensuring that data, algorithms and models employed are
thoroughly evaluated and reviewed against ethical principles.”