AI could be the key to transforming breast cancer screening

Image 2

Pictured: Dr Helen Frazer, Clinical Director of St Vincent’s Hospital Melbourne’s breast screening service with a team of radiologists in the assessment clinic at St Vincent’s Hospital Melbourne’s BreastScreen.

World-leading research using a series of computer algorithms to read breast x-rays (mammograms) aims to revolutionise breast cancer screening and improve breast cancer detection rates.

The research is part of a unique project being jointly developed by St Vincent’s Institute of Medical Research, St Vincent’s Hospital Melbourne, the University of Melbourne, the Australian Institute of Machine Learning at the University of Adelaide and BreastScreen Victoria.

Preliminary research was published recently in the Journal of Medical Imaging and Radiation Oncology.

Heading up the BRAIx project is Dr Helen Frazer, Clinical Director of St Vincent’s Hospital Melbourne’s breast screening service – one of the largest breast screening services in Victoria.

“We believe the use of AI models will be a significant step in the transformation of screening enabling more accurate detection when used in conjunction with a radiologist examining every mammogram,” Dr Frazer says.

August marks a key milestone for the $2.26 million study, funded by the Commonwealth Government Medical Research Future Fund, as researchers embark on a new stage trialling AI models against human readers to analyse the mammograms of 500,000 Victorian women screened as part of the BreastScreen Victoria program.

Dr Frazer describes it as an exciting step forward, especially as this is one of a few studies only testing algorithms in a real-world setting.

Stepping up accuracy

The current breast screening service available to Australian women is a highly effective public health program that is successfully reducing deaths from breast cancer. However, some women experience interval cancers that were not detected at the time of the screening but are detected prior to their next screening appointment in two years.

In a small number of cases, women can be recalled for further tests following their breast screen to explore an area of concern. Most of the women who attend an assessment appointment will be found not to have breast cancer.

“AI holds promise to pick up on mammogram data at a pixel level the human eye can’t see. Where we hope to make some really great inroads is with earlier detection of the high-grade cancers enabling earlier treatment and improved survival, as well as avoiding unnecessary recalls for assessment,” Dr Frazer says.

Lessening the worry and wait

Service delivery speed is also expected to improve by using AI models in the breast screening program. Currently women wait two weeks for their mammogram result and 95 per cent of them are normal.

Another focus will be to examine the potential AI models can offer to better predict the risk of breast cancer.

“This way, we will be able to identify women with a higher risk and their screening can be tailored by changing the frequency and the modality used,” Dr Frazer says.

Benefitting from collaboration

The BRAIx project is being developed by a cross-disciplinary research team of clinicians, data scientists, epidemiologists and biostatisticians, who are participants in ACMD – Australia’s first collaborative, hospital-based biomedical engineering research centre, located at St Vincent’s Hospital Melbourne.

ACMD’s specially tailored environment provides researchers with a collaborative base to investigate and accelerate their study swiftly and effectively in an effort to tackle this real-life medical challenge with a strong viable solution.