AI Could Spot Aggressive Breast Cancers Sooner, Study Finds

Artificial Intelligence Could Help Detect Aggressive Breast Cancers Earlier

Artificial intelligence (AI) has the potential to detect aggressive breast cancers earlier and alleviate the burden on overworked doctors, according to the results of a large Swedish breast screening trial. The study, published in The Lancet, analyzed nearly 100,000 women aged 40 to 80 who participated in Sweden’s national breast screening programme between April 2021 and December 2022.

Mammography is widely regarded as the most effective method for early detection of breast cancer and reducing mortality. European guidelines recommend that scans be reviewed independently by two radiologists. However, researchers have found that between 20 and 30 percent of breast cancers are diagnosed between routine screening appointments. These so-called “interval cancers” tend to grow faster, are more advanced at diagnosis, and are harder to treat.

How Was the Trial Carried Out?

In the trial, women were randomly assigned to either standard screening with two radiologists or an AI-supported approach. In the AI-supported group, AI flagged suspicious areas on mammograms, and a radiologist made the final decision.

The AI-supported group recorded 11 fewer interval cancers over two years—a 12 percent reduction compared to standard screening—and more cancers were detected during routine screening. Study researcher Jessie Gommers from Radboud University Medical Centre explained that the AI acted as a support tool rather than a replacement. She noted that “the AI highlighted suspicious areas in a mammogram that helped radiologists during interpretation,” while “the final recall decision was still made by the radiologists.”

This approach also reduced the workload for radiologists. Earlier findings from the trial showed a 44 percent reduction in radiologist workload, which researchers say could be particularly valuable in health institutions facing staff shortages.

Gommers emphasized the importance of the drop in interval cancers, stating that “you want to have as few of those interval cancers because they are known to be more aggressive and lead to poorer patient outcomes.” One in 20 women will be diagnosed with breast cancer during their lifetime, making it the most common form of cancer for women.

In 2022, an estimated 2.3 million women were diagnosed with breast cancer worldwide, and around 670,000 died from the disease. In Europe alone, the World Health Organization (WHO) recorded nearly 558,000 new cases that year.

Will AI Replace Human Doctors?

While AI-assisted mammography is already being introduced in parts of Sweden and Denmark, researchers stress that it is not ready to replace human oversight and that more evidence is necessary. Independent experts have also urged caution about wider rollout across the globe.

Dr Olga Oikonomidou, the principal investigator and breast cancer research lead at the University of Edinburgh’s Cancer Centre, said the trial focused on whether AI could help radiologists work “quicker and more efficiently,” rather than replacing them. She warned that AI systems differ widely and must be tested with the same rigour as medicines, and that they are not a low-cost alternative.

Benefits of AI in Breast Cancer Screening

• Early Detection: AI can identify suspicious areas on mammograms that might be missed by human radiologists.
• Reduced Workload: By assisting radiologists, AI can significantly reduce the time required for each scan.
• Improved Accuracy: AI can enhance the accuracy of cancer detection, leading to better patient outcomes.
• Cost Efficiency: While not a low-cost solution, AI can potentially reduce long-term healthcare costs by catching cancers earlier.

As the use of AI in healthcare continues to expand, its role in breast cancer screening represents a promising step forward. However, it is crucial to ensure that these technologies are rigorously tested and integrated into existing medical practices in a way that complements, rather than replaces, human expertise.