Aberdeen AI Trial Helps Physicians Find Breast Cancer

An ongoing trial at the Aberdeen Royal Infirmary is examining whether artificial intelligence (AI) could assist radiologists in examining thousands of mammograms each year.

The pilot helped detect early-stage breast cancer in Jun, a healthcare assistant and trial participant, who is now undergoing surgery as a result.

A mammogram is a low-level x-ray used in breast cancer screening to monitor and detect changes that are too small to be seen or felt.

In Scotland, more women were invited to participate in routine breast cancer screenings in the three years to 2022, but fewer radiologists reviewed the results.

AI, the technology that makes computers perform specific tasks that normally require human intelligence, is already widely used in a variety of industries.

Concerns from prominent experts that AI could lead to human extinction have recently made headlines, but the technology’s more realistic reality is already being shown in the medical field.

The potential to speed up the drug and disease discovery process means that many scientists and doctors see AI as a powerful tool to collaborate with rather than replace doctors. means

The number of screen-detected breast cancers in women of all ages increased to 1,830 in Scotland between 2021 and 2022, according to Public Health Scotland (PHS) data.

Of the average radiologist reviewing 5,000 mammogram exams each year, 250-300 patients are recalled, of whom 30-40 require close attention.

“That number could miss a cancer,” Dr Gerald Lipp, clinical director of the North East Scotland Breast Cancer Screening Programme, told BBC Crick.

The NHS Grampian’s Gemini Project (a partnership between the NHS, the University of Aberdeen and the private sector) was mentioned in Scotland’s AI Strategy to be launched in 2021.

Kheiron Medical Technologies developed the AI ​​model Mia used in the exam, and Microsoft provided cloud computing services to support it.

Dr. Lipp and other radiologists are piloting AI as an additional check at the end of a mammogram scan review, as regulations set by the National Screening Council currently prohibit automatic deployment of AI in screening. doing.

After Dr. Lipp explained how the AI ​​tool helped identify areas of concern, June, a trial participant who had had similar surgery before, took a piece of breast tissue for testing. underwent a biopsy to remove the

He showed BBC Click how the software works using anonymized mammogram results.

“What we’re looking at now is a woman who had a mammogram on her left and right side. We’re looking for differences,” he said.

With the click of a button, radiologists can view and see the differences between the two scans identified by AI.

Dr. Lip pointed out one area circled in the AI ​​software and identified it as the primary area of ​​concern.

“With screenings, you want to pick things up when they’re small, before they get big,” he added.

A few weeks after the biopsy, June told Crick that using AI instead of human eyes made the process less cumbersome.

“You know your image is on the screen and people are looking at it,” she says. “On the other hand, with artificial intelligence, you no longer have the feeling that someone is watching you.”

The results of June’s biopsy led her to undergo surgery again.

“The biopsy showed that I had early-stage cancer, and this time it was definitely found earlier,” June said. “But I’ve had a mastectomy in the past, so I’m going to have one.

“This is not the type of treatment I would like to receive, but at the same time I am relieved that the infection has been confirmed.”

A large review of breast cancer screening programs in Scotland published last year suggests that a large number of radiology and specialist physician staff are approaching or reaching retirement age.

He said the service was at risk of remaining “vulnerable”, especially with the decline of “superleader” radiologists who rely on the service to review large numbers of results.

In a Scottish government review, two radiologists are required to read and report the results, so replacing one human reader with AI “covers half of the burden of reading about 1.72 million images per year. We can.”

But will the technology being tested in Aberdeen eventually be replaced by human staff?

“Whether the goal of this evaluation is to replace one of the radiologists, to read as part of a routine mammogram, or to improve cancer detection as a safety measure, the best way to work with AI is to I think the way is to figure out what the ‘net is,” Dr. Lipp said.

Kheiron co-founder Peter Kecskemethy said the technology will be used by more than 30 NHS trusts across the UK, making it accessible to millions of people.

And the UK trust is already looking at how AI can support better, faster outcomes for breast cancer patients, and the technology will continue to play a big role in saving doctors’ lives. It is considered.