![Professor Stacy Carter, the lead author of the study on using AI in mammograms. Picture supplied Professor Stacy Carter, the lead author of the study on using AI in mammograms. Picture supplied](/images/transform/v1/crop/frm/HcD9H4nNcktxiWcmkEEpQD/71d8b75f-d75f-4dbf-b5d6-f5adc95d7ebb.jpeg/r0_218_8192_4824_w1200_h678_fmax.jpg)
If you've ever laughed at a funny automatic caption, or struggled to get the self-serve check-out to correctly recognise your shopping bag, it makes sense that you might be a little suspicious of the increasing role of artificial intelligence in our daily lives.
Subscribe now for unlimited access.
or signup to continue reading
And, when you're talking about detection of the world's most prevalent cancer, you really don't want an AI glitch to stuff something up.
Which is why University of Wollongong researchers have been looking into how women feel about robots doing the first read of mammograms, with the breast screening system already primed to be aided - and made faster and more accurate - by AI.
The lead author of a new UOW study on breast screening Professor Stacy Carter said the health care sector was rightly more conservative than others when it comes to using AI.
"Everyone is talking about artificial intelligence now, since ChatGPT, and in a lot of sectors it's already very widely used," she said.
"But in health care... there's a lot at stake and we don't want to use AI in a way that might harm people or that might damage people's trust in the health system."
There's masses of data - a huge digital dataset of images that have been taken during people's routine mammograms... and there's also a pretty straightforward decision to be made.
- Professor Stacy Carter
However, she said there were many areas of health - especially those with lots of good quality data and a pretty simple decision or recommendation to be made - which could be improved by AI.
"Breast screening is one of those areas," she said.
"There's masses of data - a huge digital dataset of images that have been taken during people's routine mammograms... and there's also a pretty straightforward decision to be made.
"It's not diagnostic, it's about sorting people who may possibly have a condition from those that don't."
A recently published study led from UOW's Australian Centre for Health Engagement, Evidence and Values (ACHEEV) found that women would support the use AI in breast screening, as long as there are safeguards in place and people remain part of the process.
How would AI be used in breast screening?
Led my Prof Carter, the researchers discussed with women how AI might be used in breast screening, highlighting that at the moment, up to three people were involved in assessing mammograms.
"Every single person's mammograms is definitely looked at by two, and if two people don't agree then a third will look at them to adjudicate," Prof Carter said,
"One of the motivations for trying to find a way to responsibly use AI and breast screening in future is that breast screening services often struggle to find enough expert humans.
"When you don't have enough humans, it just slows down the amount of time that it takes to provide the report back to women, and that means women are waiting for longer between having their screening and getting their result because there's not enough people."
She said her study had shown women wanted humans to remain involved in the screening process, especially in the physical process of taking the scan and also playing a role in assessing them and being accountable.
She said women surveyed wanted humans to provide a final back-up in decision making about the scans, especially in borderline cases.
"Humans have this ability, a discernment or an intuition where they can look at an unusual case or something that's hard to interpret and say, 'oh hang on a minute, what's going on here?," Prof Carter said.
"Whereas AI is actually really quite bad at what's called edge cases - the cases that are not like most of the cases it was trained on."
"Women also talked about the importance of human responsibility and having humans that were in some way responsible for the decision, which meant you could get an explanation."