New study results show promise for AI in breast cancer diagnosis and screening

A study published in the journal Nature has looked at the role Artificial Intelligence (AI) could play within breast cancer diagnosis and screening.

A mammogram uses low-energy X-rays to allow the human breast to be examined and is often used in breast cancer screening and diagnosis. However, according to the American Cancer Society, radiologists miss around 20% of breast cancers in mammograms and 50% of all of the women who are screened over a ten year period will receive a false positive result.

How are mammograms currently read?

In the UK mammograms are read by two radiologists. In the case of a disagreement, a third is consulted.

What did the study published in the journal Nature test?

The research compared a computer which was created by Google’s AI experts with healthcare professionals for screening mammograms. It did this by comparing the computer system’s performance on tens of thousands of mammograms with the actual results from a set of 25,856 previously screened mammograms in the UK.

What were the study outcomes?

The study showed the AI system could identify cancers with a similar degree of accuracy to radiologists. It also showed that the number of false positive results, where tests are wrongly classified as normal, could be reduced by using AI.

Are these findings the first of its kind?

No. Computer-Aided Detection (CAD) systems are used within mammography clinics. However, according to Connie Lehman from Harvard’s Massachusetts General Hospital, CAD programs are trained to identify things human radiologists can see, but with AI, which has the potential to train computers to spot cancers based on actual results of thousands of mammograms, she believes the technology can, “exceed human capacity to identify subtle cues that the human eye and brain aren’t able to perceive.”

What’s next?

More research is needed as well as the relevant regulatory approvals.