Breast screening technology developed with the help of artificial intelligence (AI) can detect abnormalities that would otherwise be missed with current screening methods, the University of Aberdeen writes in a paper. press release.
- The AI detected anomalies that would have been missed by current scanning methods;
- The program reminded a third of women with cancer between checkups.
Pioneering work by the University of Aberdeen, NHS Grampian and Kheiron Medical Technologies involved analyzing 220,000 mammograms from more than 55,000 people to determine how well an AI tool could detect breast cancer – from research to clinical application.
Khyron Medical Technologies Ltd.’s developed AI software called ‘Mia’ was evaluated by a research team led by Professor Leslie Anderson, Head of Health Data Science at the University of Aberdeen, as part of the Industrial Center for Research on Artificial Intelligence in Digital Diagnostics (iCAIRD) programme. Dr Clarice de Vries, a researcher in radiology imaging at the University of Aberdeen, led the data analysis.
Identifying overlooked cancers
The analysis found that Mia was successful in identifying potentially missing cancers, known as interval cancers, that are caught between scans. The team found that Mia would have suggested 34.1 percent of women diagnosed with cancer be re-evaluated between check-ups. With current screening, these cancers are not detected until women show symptoms.
Dr. De Vries explains: “Currently, two experts review each mammogram and decide whether a person should be invited for further examination. If the two experts disagree, a third expert makes the final decision. Like any human expert, Mia can examine a mammogram and give an opinion About whether this person should be invited for further examination.
“We’ve shown that once AI adjusts to the local environment, it can have huge benefits for clinicians and especially for people at high risk of cancer.
Improving health care
Radiology consultant Dr. Gerald Lieb was instrumental in getting Mia into clinical practice: “When Mia was originally applied to NHS Grampian data, she was very sensitive – recommending that women be recalled for further investigation when this was not necessary. However, Mia’s performance improved significantly when They were adapted to local conditions and technology, allowing as few women as possible to be recalled while maintaining a high cancer detection rate.
“Until now, it was not clear whether AI tools developed elsewhere could be used in different places and testing centers. We now know that there are risks in simply using an AI tool developed elsewhere and deploying it locally. You should test the tool on local data first to make sure it works as expected,” Lieb said.
Professor Roger Staff, Head of Imaging Physics at the NHS Grampian, added: “This is an important study that outlines the steps required to bring this technology to service. While the results suggest that the technology does not work immediately, it has the potential to bring significant health and operational benefits to the service.” .